id	sid	tid	token	lemma	pos
fcis-30297	1	1	frontiers	frontier	NOUN
fcis-30297	1	2	in	in	ADP
fcis-30297	1	3	computing	computing	NOUN
fcis-30297	1	4	and	and	CCONJ
fcis-30297	1	5	intelligent	intelligent	ADJ
fcis-30297	1	6	systems	system	NOUN
fcis-30297	1	7	issn	issn	VERB
fcis-30297	1	8	:	:	PUNCT
fcis-30297	1	9	2832	2832	NUM
fcis-30297	1	10	-	-	SYM
fcis-30297	1	11	6024	6024	NUM
fcis-30297	1	12	|	|	NOUN
fcis-30297	1	13	vol	vol	NOUN
fcis-30297	1	14	.	.	PROPN
fcis-30297	2	1	12	12	NUM
fcis-30297	2	2	,	,	PUNCT
fcis-30297	2	3	no	no	INTJ
fcis-30297	2	4	.	.	NOUN
fcis-30297	2	5	1	1	NUM
fcis-30297	2	6	,	,	PUNCT
fcis-30297	2	7	2025	2025	NUM
fcis-30297	2	8	1	1	NUM
fcis-30297	2	9	rt‐detr‐based	rt‐detr‐base	VERB
fcis-30297	2	10	wideband	wideband	NOUN
fcis-30297	2	11	signal	signal	NOUN
fcis-30297	2	12	detection	detection	NOUN
fcis-30297	2	13	and	and	CCONJ
fcis-30297	2	14	modulation	modulation	NOUN
fcis-30297	2	15	classification	classification	NOUN
fcis-30297	2	16	minghao	minghao	PROPN
fcis-30297	2	17	cao	cao	PROPN
fcis-30297	2	18	,	,	PUNCT
fcis-30297	2	19	peng	peng	PROPN
fcis-30297	2	20	chu	chu	PROPN
fcis-30297	2	21	*	*	PROPN
fcis-30297	2	22	,	,	PUNCT
fcis-30297	2	23	pengfei	pengfei	PROPN
fcis-30297	2	24	ma	ma	PROPN
fcis-30297	2	25	,	,	PUNCT
fcis-30297	2	26	bo	bo	PROPN
fcis-30297	2	27	fang	fang	PROPN
fcis-30297	2	28	school	school	NOUN
fcis-30297	2	29	of	of	ADP
fcis-30297	2	30	electronic	electronic	ADJ
fcis-30297	2	31	information	information	NOUN
fcis-30297	2	32	,	,	PUNCT
fcis-30297	2	33	xijing	xijing	PROPN
fcis-30297	2	34	university	university	PROPN
fcis-30297	2	35	,	,	PUNCT
fcis-30297	2	36	xi’an	xi’an	PROPN
fcis-30297	2	37	,	,	PUNCT
fcis-30297	2	38	shaanxi	shaanxi	PROPN
fcis-30297	2	39	,	,	PUNCT
fcis-30297	2	40	china	china	PROPN
fcis-30297	2	41	*	*	PUNCT
fcis-30297	2	42	corresponding	correspond	VERB
fcis-30297	2	43	author	author	NOUN
fcis-30297	2	44	:	:	PUNCT
fcis-30297	2	45	peng	peng	PROPN
fcis-30297	2	46	chu	chu	PROPN
fcis-30297	2	47	abstract	abstract	PROPN
fcis-30297	2	48	:	:	PUNCT
fcis-30297	2	49	in	in	ADP
fcis-30297	2	50	to	to	PART
fcis-30297	2	51	address	address	VERB
fcis-30297	2	52	the	the	DET
fcis-30297	2	53	problems	problem	NOUN
fcis-30297	2	54	of	of	ADP
fcis-30297	2	55	high	high	ADJ
fcis-30297	2	56	computational	computational	ADJ
fcis-30297	2	57	complexity	complexity	NOUN
fcis-30297	2	58	,	,	PUNCT
fcis-30297	2	59	low	low	ADJ
fcis-30297	2	60	accuracy	accuracy	NOUN
fcis-30297	2	61	,	,	PUNCT
fcis-30297	2	62	and	and	CCONJ
fcis-30297	2	63	the	the	DET
fcis-30297	2	64	cumbersome	cumbersome	ADJ
fcis-30297	2	65	manual	manual	ADJ
fcis-30297	2	66	feature	feature	NOUN
fcis-30297	2	67	extraction	extraction	NOUN
fcis-30297	2	68	process	process	NOUN
fcis-30297	2	69	in	in	ADP
fcis-30297	2	70	traditional	traditional	ADJ
fcis-30297	2	71	machine	machine	NOUN
fcis-30297	2	72	learning	learning	NOUN
fcis-30297	2	73	methods	method	NOUN
fcis-30297	2	74	for	for	ADP
fcis-30297	2	75	communication	communication	NOUN
fcis-30297	2	76	signal	signal	NOUN
fcis-30297	2	77	modulation	modulation	NOUN
fcis-30297	2	78	recognition	recognition	NOUN
fcis-30297	2	79	,	,	PUNCT
fcis-30297	2	80	this	this	DET
fcis-30297	2	81	study	study	NOUN
fcis-30297	2	82	proposes	propose	VERB
fcis-30297	2	83	a	a	DET
fcis-30297	2	84	deep	deep	ADJ
fcis-30297	2	85	learning	learning	NOUN
fcis-30297	2	86	-	-	PUNCT
fcis-30297	2	87	based	base	VERB
fcis-30297	2	88	end	end	NOUN
fcis-30297	2	89	-	-	PUNCT
fcis-30297	2	90	to	to	ADP
fcis-30297	2	91	-	-	PUNCT
fcis-30297	2	92	end	end	NOUN
fcis-30297	2	93	recognition	recognition	NOUN
fcis-30297	2	94	model	model	NOUN
fcis-30297	2	95	.	.	PUNCT
fcis-30297	3	1	built	build	VERB
fcis-30297	3	2	upon	upon	SCONJ
fcis-30297	3	3	the	the	DET
fcis-30297	3	4	transformer	transformer	NOUN
fcis-30297	3	5	architecture	architecture	NOUN
fcis-30297	3	6	using	use	VERB
fcis-30297	3	7	the	the	DET
fcis-30297	3	8	rt	rt	PROPN
fcis-30297	3	9	-	-	PUNCT
fcis-30297	3	10	detr	detr	NOUN
fcis-30297	3	11	framework	framework	NOUN
fcis-30297	3	12	,	,	PUNCT
fcis-30297	3	13	the	the	DET
fcis-30297	3	14	model	model	NOUN
fcis-30297	3	15	directly	directly	ADV
fcis-30297	3	16	identifies	identify	VERB
fcis-30297	3	17	modulation	modulation	NOUN
fcis-30297	3	18	types	type	NOUN
fcis-30297	3	19	from	from	ADP
fcis-30297	3	20	sampled	sample	VERB
fcis-30297	3	21	communication	communication	NOUN
fcis-30297	3	22	signals	signal	NOUN
fcis-30297	3	23	.	.	PUNCT
fcis-30297	4	1	it	it	PRON
fcis-30297	4	2	features	feature	VERB
fcis-30297	4	3	high	high	ADJ
fcis-30297	4	4	recognition	recognition	NOUN
fcis-30297	4	5	accuracy	accuracy	NOUN
fcis-30297	4	6	,	,	PUNCT
fcis-30297	4	7	strong	strong	ADJ
fcis-30297	4	8	generalization	generalization	NOUN
fcis-30297	4	9	ability	ability	NOUN
fcis-30297	4	10	,	,	PUNCT
fcis-30297	4	11	robustness	robustness	NOUN
fcis-30297	4	12	to	to	ADP
fcis-30297	4	13	noise	noise	NOUN
fcis-30297	4	14	,	,	PUNCT
fcis-30297	4	15	and	and	CCONJ
fcis-30297	4	16	a	a	DET
fcis-30297	4	17	streamlined	streamlined	ADJ
fcis-30297	4	18	processing	processing	NOUN
fcis-30297	4	19	pipeline	pipeline	NOUN
fcis-30297	4	20	.	.	PUNCT
fcis-30297	5	1	extensive	extensive	ADJ
fcis-30297	5	2	experiments	experiment	NOUN
fcis-30297	5	3	validate	validate	VERB
fcis-30297	5	4	the	the	DET
fcis-30297	5	5	model	model	NOUN
fcis-30297	5	6	’s	’s	PART
fcis-30297	5	7	effectiveness	effectiveness	NOUN
fcis-30297	5	8	,	,	PUNCT
fcis-30297	5	9	demonstrating	demonstrate	VERB
fcis-30297	5	10	its	its	PRON
fcis-30297	5	11	superior	superior	ADJ
fcis-30297	5	12	performance	performance	NOUN
fcis-30297	5	13	in	in	ADP
fcis-30297	5	14	automatic	automatic	ADJ
fcis-30297	5	15	feature	feature	NOUN
fcis-30297	5	16	extraction	extraction	NOUN
fcis-30297	5	17	and	and	CCONJ
fcis-30297	5	18	modulation	modulation	NOUN
fcis-30297	5	19	classification	classification	NOUN
fcis-30297	5	20	compared	compare	VERB
fcis-30297	5	21	to	to	ADP
fcis-30297	5	22	traditional	traditional	ADJ
fcis-30297	5	23	approaches	approach	NOUN
fcis-30297	5	24	.	.	PUNCT
fcis-30297	6	1	keywords	keyword	NOUN
fcis-30297	6	2	:	:	PUNCT
fcis-30297	6	3	feature	feature	NOUN
fcis-30297	6	4	extraction	extraction	NOUN
fcis-30297	6	5	networks	network	NOUN
fcis-30297	6	6	;	;	PUNCT
fcis-30297	6	7	signal	signal	ADJ
fcis-30297	6	8	modulation	modulation	NOUN
fcis-30297	6	9	recognition	recognition	NOUN
fcis-30297	6	10	;	;	PUNCT
fcis-30297	6	11	deep	deep	ADJ
fcis-30297	6	12	learning	learning	NOUN
fcis-30297	6	13	.	.	PUNCT
fcis-30297	7	1	1	1	X
fcis-30297	7	2	.	.	X
fcis-30297	7	3	introduction	introduction	NOUN
fcis-30297	7	4	with	with	ADP
fcis-30297	7	5	the	the	DET
fcis-30297	7	6	rapid	rapid	ADJ
fcis-30297	7	7	advancement	advancement	NOUN
fcis-30297	7	8	of	of	ADP
fcis-30297	7	9	wireless	wireless	ADJ
fcis-30297	7	10	communication	communication	NOUN
fcis-30297	7	11	and	and	CCONJ
fcis-30297	7	12	signal	signal	NOUN
fcis-30297	7	13	processing	processing	NOUN
fcis-30297	7	14	technologies	technology	NOUN
fcis-30297	7	15	,	,	PUNCT
fcis-30297	7	16	signal	signal	ADJ
fcis-30297	7	17	modulation	modulation	NOUN
fcis-30297	7	18	recognition	recognition	NOUN
fcis-30297	7	19	has	have	AUX
fcis-30297	7	20	become	become	VERB
fcis-30297	7	21	a	a	DET
fcis-30297	7	22	critical	critical	ADJ
fcis-30297	7	23	task	task	NOUN
fcis-30297	7	24	in	in	ADP
fcis-30297	7	25	various	various	ADJ
fcis-30297	7	26	fields	field	NOUN
fcis-30297	7	27	,	,	PUNCT
fcis-30297	7	28	including	include	VERB
fcis-30297	7	29	military	military	ADJ
fcis-30297	7	30	communications	communication	NOUN
fcis-30297	7	31	,	,	PUNCT
fcis-30297	7	32	radio	radio	NOUN
fcis-30297	7	33	monitoring	monitoring	NOUN
fcis-30297	7	34	,	,	PUNCT
fcis-30297	7	35	and	and	CCONJ
fcis-30297	7	36	cognitive	cognitive	ADJ
fcis-30297	7	37	radio	radio	NOUN
fcis-30297	7	38	.	.	PUNCT
fcis-30297	8	1	its	its	PRON
fcis-30297	8	2	primary	primary	ADJ
fcis-30297	8	3	goal	goal	NOUN
fcis-30297	8	4	is	be	AUX
fcis-30297	8	5	to	to	PART
fcis-30297	8	6	extract	extract	VERB
fcis-30297	8	7	modulation	modulation	NOUN
fcis-30297	8	8	characteristics	characteristic	NOUN
fcis-30297	8	9	from	from	ADP
fcis-30297	8	10	received	receive	VERB
fcis-30297	8	11	signals	signal	NOUN
fcis-30297	8	12	and	and	CCONJ
fcis-30297	8	13	accurately	accurately	ADV
fcis-30297	8	14	determine	determine	VERB
fcis-30297	8	15	their	their	PRON
fcis-30297	8	16	modulation	modulation	NOUN
fcis-30297	8	17	types	type	NOUN
fcis-30297	8	18	.	.	PUNCT
fcis-30297	9	1	traditional	traditional	ADJ
fcis-30297	9	2	modulation	modulation	NOUN
fcis-30297	9	3	recognition	recognition	NOUN
fcis-30297	9	4	methods	method	NOUN
fcis-30297	9	5	often	often	ADV
fcis-30297	9	6	rely	rely	VERB
fcis-30297	9	7	on	on	ADP
fcis-30297	9	8	manually	manually	ADV
fcis-30297	9	9	crafted	craft	VERB
fcis-30297	9	10	feature	feature	NOUN
fcis-30297	9	11	extraction	extraction	NOUN
fcis-30297	9	12	combined	combine	VERB
fcis-30297	9	13	with	with	ADP
fcis-30297	9	14	classical	classical	ADJ
fcis-30297	9	15	classification	classification	NOUN
fcis-30297	9	16	algorithms	algorithm	NOUN
fcis-30297	9	17	.	.	PUNCT
fcis-30297	10	1	however	however	ADV
fcis-30297	10	2	,	,	PUNCT
fcis-30297	10	3	these	these	DET
fcis-30297	10	4	approaches	approach	NOUN
fcis-30297	10	5	struggle	struggle	VERB
fcis-30297	10	6	to	to	PART
fcis-30297	10	7	maintain	maintain	VERB
fcis-30297	10	8	high	high	ADJ
fcis-30297	10	9	performance	performance	NOUN
fcis-30297	10	10	in	in	ADP
fcis-30297	10	11	complex	complex	ADJ
fcis-30297	10	12	electromagnetic	electromagnetic	ADJ
fcis-30297	10	13	environments	environment	NOUN
fcis-30297	10	14	,	,	PUNCT
fcis-30297	10	15	resulting	result	VERB
fcis-30297	10	16	in	in	ADP
fcis-30297	10	17	limited	limited	ADJ
fcis-30297	10	18	recognition	recognition	NOUN
fcis-30297	10	19	accuracy[1	accuracy[1	PUNCT
fcis-30297	10	20	]	]	PUNCT
fcis-30297	10	21	.	.	PUNCT
fcis-30297	11	1	in	in	ADP
fcis-30297	11	2	recent	recent	ADJ
fcis-30297	11	3	years	year	NOUN
fcis-30297	11	4	,	,	PUNCT
fcis-30297	11	5	the	the	DET
fcis-30297	11	6	emergence	emergence	NOUN
fcis-30297	11	7	of	of	ADP
fcis-30297	11	8	deep	deep	ADJ
fcis-30297	11	9	learning	learning	NOUN
fcis-30297	11	10	has	have	AUX
fcis-30297	11	11	brought	bring	VERB
fcis-30297	11	12	new	new	ADJ
fcis-30297	11	13	opportunities	opportunity	NOUN
fcis-30297	11	14	to	to	ADP
fcis-30297	11	15	this	this	DET
fcis-30297	11	16	field	field	NOUN
fcis-30297	11	17	.	.	PUNCT
fcis-30297	12	1	deep	deep	ADJ
fcis-30297	12	2	neural	neural	ADJ
fcis-30297	12	3	networks	network	NOUN
fcis-30297	12	4	have	have	AUX
fcis-30297	12	5	shown	show	VERB
fcis-30297	12	6	remarkable	remarkable	ADJ
fcis-30297	12	7	capability	capability	NOUN
fcis-30297	12	8	in	in	ADP
fcis-30297	12	9	automatically	automatically	ADV
fcis-30297	12	10	learning	learn	VERB
fcis-30297	12	11	signal	signal	NOUN
fcis-30297	12	12	features	feature	NOUN
fcis-30297	12	13	,	,	PUNCT
fcis-30297	12	14	offering	offer	VERB
fcis-30297	12	15	enhanced	enhance	VERB
fcis-30297	12	16	robustness	robustness	NOUN
fcis-30297	12	17	and	and	CCONJ
fcis-30297	12	18	higher	high	ADJ
fcis-30297	12	19	accuracy	accuracy	NOUN
fcis-30297	12	20	under	under	ADP
fcis-30297	12	21	challenging	challenge	VERB
fcis-30297	12	22	conditions[2	conditions[2	NOUN
fcis-30297	12	23	]	]	X
fcis-30297	12	24	.	.	PUNCT
fcis-30297	13	1	despite	despite	SCONJ
fcis-30297	13	2	these	these	DET
fcis-30297	13	3	advantages	advantage	NOUN
fcis-30297	13	4	,	,	PUNCT
fcis-30297	13	5	many	many	ADJ
fcis-30297	13	6	existing	exist	VERB
fcis-30297	13	7	deep	deep	ADJ
fcis-30297	13	8	learning	learning	NOUN
fcis-30297	13	9	-	-	PUNCT
fcis-30297	13	10	based	base	VERB
fcis-30297	13	11	methods	method	NOUN
fcis-30297	13	12	still	still	ADV
fcis-30297	13	13	face	face	VERB
fcis-30297	13	14	tradeoffs	tradeoff	NOUN
fcis-30297	13	15	between	between	ADP
fcis-30297	13	16	model	model	NOUN
fcis-30297	13	17	complexity	complexity	NOUN
fcis-30297	13	18	,	,	PUNCT
fcis-30297	13	19	computational	computational	ADJ
fcis-30297	13	20	efficiency	efficiency	NOUN
fcis-30297	13	21	,	,	PUNCT
fcis-30297	13	22	and	and	CCONJ
fcis-30297	13	23	recognition	recognition	NOUN
fcis-30297	13	24	accuracy[3	accuracy[3	PROPN
fcis-30297	13	25	]	]	X
fcis-30297	13	26	.	.	PUNCT
fcis-30297	14	1	therefore	therefore	ADV
fcis-30297	14	2	,	,	PUNCT
fcis-30297	14	3	designing	design	VERB
fcis-30297	14	4	a	a	DET
fcis-30297	14	5	lightweight	lightweight	ADJ
fcis-30297	14	6	and	and	CCONJ
fcis-30297	14	7	efficient	efficient	ADJ
fcis-30297	14	8	deep	deep	ADJ
fcis-30297	14	9	learning	learning	NOUN
fcis-30297	14	10	model	model	NOUN
fcis-30297	14	11	that	that	PRON
fcis-30297	14	12	achieves	achieve	VERB
fcis-30297	14	13	high	high	ADJ
fcis-30297	14	14	recognition	recognition	NOUN
fcis-30297	14	15	accuracy	accuracy	NOUN
fcis-30297	14	16	remains	remain	VERB
fcis-30297	14	17	a	a	DET
fcis-30297	14	18	pressing	press	VERB
fcis-30297	14	19	research	research	NOUN
fcis-30297	14	20	challenge	challenge	NOUN
fcis-30297	14	21	.	.	PUNCT
fcis-30297	15	1	in	in	ADP
fcis-30297	15	2	this	this	DET
fcis-30297	15	3	study	study	NOUN
fcis-30297	15	4	,	,	PUNCT
fcis-30297	15	5	we	we	PRON
fcis-30297	15	6	propose	propose	VERB
fcis-30297	15	7	a	a	DET
fcis-30297	15	8	novel	novel	ADJ
fcis-30297	15	9	signal	signal	NOUN
fcis-30297	15	10	modulation	modulation	NOUN
fcis-30297	15	11	recognition	recognition	NOUN
fcis-30297	15	12	method	method	NOUN
fcis-30297	15	13	based	base	VERB
fcis-30297	15	14	on	on	ADP
fcis-30297	15	15	the	the	DET
fcis-30297	15	16	rt	rt	PROPN
fcis-30297	15	17	-	-	PUNCT
fcis-30297	15	18	detr	detr	NOUN
fcis-30297	15	19	architecture	architecture	NOUN
fcis-30297	15	20	.	.	PUNCT
fcis-30297	16	1	rtdetr	rtdetr	PROPN
fcis-30297	16	2	,	,	PUNCT
fcis-30297	16	3	originally	originally	ADV
fcis-30297	16	4	designed	design	VERB
fcis-30297	16	5	for	for	ADP
fcis-30297	16	6	real	real	ADJ
fcis-30297	16	7	-	-	PUNCT
fcis-30297	16	8	time	time	NOUN
fcis-30297	16	9	object	object	NOUN
fcis-30297	16	10	detection	detection	NOUN
fcis-30297	16	11	,	,	PUNCT
fcis-30297	16	12	provides	provide	VERB
fcis-30297	16	13	a	a	DET
fcis-30297	16	14	strong	strong	ADJ
fcis-30297	16	15	balance	balance	NOUN
fcis-30297	16	16	between	between	ADP
fcis-30297	16	17	computational	computational	ADJ
fcis-30297	16	18	efficiency	efficiency	NOUN
fcis-30297	16	19	and	and	CCONJ
fcis-30297	16	20	generalization	generalization	NOUN
fcis-30297	16	21	ability[4	ability[4	ADV
fcis-30297	16	22	]	]	PUNCT
fcis-30297	16	23	.	.	PUNCT
fcis-30297	17	1	to	to	PART
fcis-30297	17	2	further	far	ADV
fcis-30297	17	3	improve	improve	VERB
fcis-30297	17	4	its	its	PRON
fcis-30297	17	5	performance	performance	NOUN
fcis-30297	17	6	for	for	ADP
fcis-30297	17	7	modulation	modulation	NOUN
fcis-30297	17	8	recognition	recognition	NOUN
fcis-30297	17	9	tasks	task	NOUN
fcis-30297	17	10	,	,	PUNCT
fcis-30297	17	11	we	we	PRON
fcis-30297	17	12	introduce	introduce	VERB
fcis-30297	17	13	two	two	NUM
fcis-30297	17	14	key	key	ADJ
fcis-30297	17	15	modifications	modification	NOUN
fcis-30297	17	16	:	:	PUNCT
fcis-30297	17	17	1.efficientformerv2	1.efficientformerv2	NUM
fcis-30297	17	18	is	be	AUX
fcis-30297	17	19	employed	employ	VERB
fcis-30297	17	20	as	as	ADP
fcis-30297	17	21	the	the	DET
fcis-30297	17	22	backbone	backbone	NOUN
fcis-30297	17	23	network	network	NOUN
fcis-30297	17	24	in	in	ADP
fcis-30297	17	25	place	place	NOUN
fcis-30297	17	26	of	of	ADP
fcis-30297	17	27	resnet-18	resnet-18	PROPN
fcis-30297	17	28	,	,	PUNCT
fcis-30297	17	29	enhancing	enhance	VERB
fcis-30297	17	30	feature	feature	NOUN
fcis-30297	17	31	extraction	extraction	NOUN
fcis-30297	17	32	while	while	SCONJ
fcis-30297	17	33	maintaining	maintain	VERB
fcis-30297	17	34	a	a	DET
fcis-30297	17	35	lightweight	lightweight	ADJ
fcis-30297	17	36	structure	structure	NOUN
fcis-30297	17	37	[	[	X
fcis-30297	17	38	5	5	NUM
fcis-30297	17	39	]	]	PUNCT
fcis-30297	17	40	.	.	PUNCT
fcis-30297	18	1	2.the	2.the	PRON
fcis-30297	18	2	dynamic	dynamic	ADJ
fcis-30297	18	3	-	-	PUNCT
fcis-30297	18	4	range	range	NOUN
fcis-30297	18	5	histogram	histogram	NOUN
fcis-30297	18	6	self	self	NOUN
fcis-30297	18	7	-	-	PUNCT
fcis-30297	18	8	attention	attention	NOUN
fcis-30297	18	9	(	(	PUNCT
fcis-30297	18	10	dhsattention	dhsattention	NOUN
fcis-30297	18	11	)	)	PUNCT
fcis-30297	18	12	module	module	NOUN
fcis-30297	18	13	,	,	PUNCT
fcis-30297	18	14	inspired	inspire	VERB
fcis-30297	18	15	by	by	ADP
fcis-30297	18	16	histoformer	histoformer	NOUN
fcis-30297	18	17	,	,	PUNCT
fcis-30297	18	18	is	be	AUX
fcis-30297	18	19	integrated	integrate	VERB
fcis-30297	18	20	to	to	PART
fcis-30297	18	21	replace	replace	VERB
fcis-30297	18	22	the	the	DET
fcis-30297	18	23	original	original	ADJ
fcis-30297	18	24	aifi	aifi	NOUN
fcis-30297	19	1	[	[	X
fcis-30297	19	2	6	6	NUM
fcis-30297	19	3	]	]	PUNCT
fcis-30297	19	4	.	.	PUNCT
fcis-30297	20	1	this	this	DET
fcis-30297	20	2	module	module	NOUN
fcis-30297	20	3	improves	improve	VERB
fcis-30297	20	4	attention	attention	NOUN
fcis-30297	20	5	distribution	distribution	NOUN
fcis-30297	20	6	and	and	CCONJ
fcis-30297	20	7	feature	feature	NOUN
fcis-30297	20	8	fusion	fusion	NOUN
fcis-30297	20	9	through	through	ADP
fcis-30297	20	10	dynamic	dynamic	ADJ
fcis-30297	20	11	histogrambased	histogrambased	ADJ
fcis-30297	20	12	modeling	modeling	NOUN
fcis-30297	20	13	.	.	PUNCT
fcis-30297	21	1	these	these	DET
fcis-30297	21	2	improvements	improvement	NOUN
fcis-30297	21	3	enable	enable	VERB
fcis-30297	21	4	the	the	DET
fcis-30297	21	5	model	model	NOUN
fcis-30297	21	6	to	to	PART
fcis-30297	21	7	better	well	ADJ
fcis-30297	21	8	capture	capture	VERB
fcis-30297	21	9	modulation	modulation	NOUN
fcis-30297	21	10	features	feature	NOUN
fcis-30297	21	11	while	while	SCONJ
fcis-30297	21	12	maintaining	maintain	VERB
fcis-30297	21	13	low	low	ADJ
fcis-30297	21	14	computational	computational	ADJ
fcis-30297	21	15	cost	cost	NOUN
fcis-30297	21	16	.	.	PUNCT
fcis-30297	22	1	the	the	DET
fcis-30297	22	2	main	main	ADJ
fcis-30297	22	3	contributions	contribution	NOUN
fcis-30297	22	4	of	of	ADP
fcis-30297	22	5	this	this	DET
fcis-30297	22	6	study	study	NOUN
fcis-30297	22	7	are	be	AUX
fcis-30297	22	8	summarized	summarize	VERB
fcis-30297	22	9	as	as	SCONJ
fcis-30297	22	10	follows	follow	VERB
fcis-30297	22	11	:	:	PUNCT
fcis-30297	22	12	we	we	PRON
fcis-30297	22	13	propose	propose	VERB
fcis-30297	22	14	a	a	DET
fcis-30297	22	15	lightweight	lightweight	ADJ
fcis-30297	22	16	and	and	CCONJ
fcis-30297	22	17	efficient	efficient	ADJ
fcis-30297	22	18	signal	signal	NOUN
fcis-30297	22	19	modulation	modulation	NOUN
fcis-30297	22	20	recognition	recognition	NOUN
fcis-30297	22	21	method	method	NOUN
fcis-30297	22	22	based	base	VERB
fcis-30297	22	23	on	on	ADP
fcis-30297	22	24	rt	rt	PROPN
fcis-30297	22	25	-	-	PUNCT
fcis-30297	22	26	detr	detr	NOUN
fcis-30297	22	27	,	,	PUNCT
fcis-30297	22	28	integrating	integrate	VERB
fcis-30297	22	29	efficientformerv2	efficientformerv2	PROPN
fcis-30297	22	30	and	and	CCONJ
fcis-30297	22	31	dhs	dhs	PROPN
fcis-30297	22	32	-	-	PUNCT
fcis-30297	22	33	attention	attention	NOUN
fcis-30297	22	34	to	to	PART
fcis-30297	22	35	enhance	enhance	VERB
fcis-30297	22	36	feature	feature	NOUN
fcis-30297	22	37	extraction	extraction	NOUN
fcis-30297	22	38	and	and	CCONJ
fcis-30297	22	39	fusion	fusion	NOUN
fcis-30297	22	40	capabilities	capability	NOUN
fcis-30297	22	41	.	.	PUNCT
fcis-30297	23	1	we	we	PRON
fcis-30297	23	2	design	design	VERB
fcis-30297	23	3	a	a	DET
fcis-30297	23	4	comprehensive	comprehensive	ADJ
fcis-30297	23	5	set	set	NOUN
fcis-30297	23	6	of	of	ADP
fcis-30297	23	7	experiments	experiment	NOUN
fcis-30297	23	8	to	to	PART
fcis-30297	23	9	evaluate	evaluate	VERB
fcis-30297	23	10	the	the	DET
fcis-30297	23	11	proposed	propose	VERB
fcis-30297	23	12	model	model	NOUN
fcis-30297	23	13	across	across	ADP
fcis-30297	23	14	various	various	ADJ
fcis-30297	23	15	modulation	modulation	NOUN
fcis-30297	23	16	types	type	NOUN
fcis-30297	23	17	and	and	CCONJ
fcis-30297	23	18	snr	snr	NOUN
fcis-30297	23	19	conditions	condition	NOUN
fcis-30297	23	20	,	,	PUNCT
fcis-30297	23	21	and	and	CCONJ
fcis-30297	23	22	perform	perform	VERB
fcis-30297	23	23	comparative	comparative	ADJ
fcis-30297	23	24	analyses	analysis	NOUN
fcis-30297	23	25	with	with	ADP
fcis-30297	23	26	traditional	traditional	ADJ
fcis-30297	23	27	methods	method	NOUN
fcis-30297	23	28	and	and	CCONJ
fcis-30297	23	29	baseline	baseline	NOUN
fcis-30297	23	30	models	model	NOUN
fcis-30297	23	31	.	.	PUNCT
fcis-30297	24	1	the	the	DET
fcis-30297	24	2	remainder	remainder	NOUN
fcis-30297	24	3	of	of	ADP
fcis-30297	24	4	this	this	DET
fcis-30297	24	5	paper	paper	NOUN
fcis-30297	24	6	is	be	AUX
fcis-30297	24	7	organized	organize	VERB
fcis-30297	24	8	as	as	SCONJ
fcis-30297	24	9	follows	follow	VERB
fcis-30297	24	10	:	:	PUNCT
fcis-30297	24	11	section	section	NOUN
fcis-30297	24	12	2	2	NUM
fcis-30297	24	13	reviews	review	NOUN
fcis-30297	24	14	related	relate	VERB
fcis-30297	24	15	work	work	NOUN
fcis-30297	24	16	,	,	PUNCT
fcis-30297	24	17	section	section	NOUN
fcis-30297	24	18	3	3	NUM
fcis-30297	24	19	introduces	introduce	NOUN
fcis-30297	24	20	the	the	DET
fcis-30297	24	21	proposed	propose	VERB
fcis-30297	24	22	method	method	NOUN
fcis-30297	24	23	in	in	ADP
fcis-30297	24	24	detail	detail	NOUN
fcis-30297	24	25	,	,	PUNCT
fcis-30297	24	26	and	and	CCONJ
fcis-30297	24	27	section	section	NOUN
fcis-30297	24	28	4	4	NUM
fcis-30297	24	29	presents	present	VERB
fcis-30297	24	30	the	the	DET
fcis-30297	24	31	experimental	experimental	ADJ
fcis-30297	24	32	setup	setup	NOUN
fcis-30297	24	33	,	,	PUNCT
fcis-30297	24	34	results	result	NOUN
fcis-30297	24	35	,	,	PUNCT
fcis-30297	24	36	and	and	CCONJ
fcis-30297	24	37	analysis	analysis	NOUN
fcis-30297	24	38	.	.	PUNCT
fcis-30297	25	1	2	2	X
fcis-30297	25	2	.	.	X
fcis-30297	25	3	related	relate	VERB
fcis-30297	25	4	research	research	NOUN
fcis-30297	25	5	2.1	2.1	NUM
fcis-30297	25	6	.	.	PUNCT
fcis-30297	26	1	traditional	traditional	ADJ
fcis-30297	26	2	signal	signal	NOUN
fcis-30297	26	3	modulation	modulation	NOUN
fcis-30297	26	4	recognition	recognition	NOUN
fcis-30297	26	5	methods	method	NOUN
fcis-30297	26	6	signal	signal	VERB
fcis-30297	26	7	modulation	modulation	NOUN
fcis-30297	26	8	recognition	recognition	NOUN
fcis-30297	26	9	plays	play	VERB
fcis-30297	26	10	a	a	DET
fcis-30297	26	11	vital	vital	ADJ
fcis-30297	26	12	role	role	NOUN
fcis-30297	26	13	in	in	ADP
fcis-30297	26	14	wireless	wireless	ADJ
fcis-30297	26	15	communications	communication	NOUN
fcis-30297	26	16	.	.	PUNCT
fcis-30297	27	1	early	early	ADJ
fcis-30297	27	2	research	research	NOUN
fcis-30297	27	3	primarily	primarily	ADV
fcis-30297	27	4	relied	rely	VERB
fcis-30297	27	5	on	on	ADP
fcis-30297	27	6	traditional	traditional	ADJ
fcis-30297	27	7	handcrafted	handcraft	VERB
fcis-30297	27	8	feature	feature	NOUN
fcis-30297	27	9	extraction	extraction	NOUN
fcis-30297	27	10	combined	combine	VERB
fcis-30297	27	11	with	with	ADP
fcis-30297	27	12	classical	classical	ADJ
fcis-30297	27	13	classification	classification	NOUN
fcis-30297	27	14	algorithms	algorithm	NOUN
fcis-30297	27	15	.	.	PUNCT
fcis-30297	28	1	common	common	ADJ
fcis-30297	28	2	approaches	approach	NOUN
fcis-30297	28	3	include	include	VERB
fcis-30297	28	4	recognition	recognition	NOUN
fcis-30297	28	5	based	base	VERB
fcis-30297	28	6	on	on	ADP
fcis-30297	28	7	statistical	statistical	ADJ
fcis-30297	28	8	features	feature	NOUN
fcis-30297	28	9	,	,	PUNCT
fcis-30297	28	10	instantaneous	instantaneous	ADJ
fcis-30297	28	11	parameters	parameter	NOUN
fcis-30297	28	12	,	,	PUNCT
fcis-30297	28	13	and	and	CCONJ
fcis-30297	28	14	cyclostationary	cyclostationary	ADJ
fcis-30297	28	15	feature	feature	NOUN
fcis-30297	28	16	extraction	extraction	NOUN
fcis-30297	28	17	.	.	PUNCT
fcis-30297	29	1	these	these	DET
fcis-30297	29	2	methods	method	NOUN
fcis-30297	29	3	typically	typically	ADV
fcis-30297	29	4	derive	derive	VERB
fcis-30297	29	5	frequency	frequency	NOUN
fcis-30297	29	6	-	-	PUNCT
fcis-30297	29	7	domain	domain	NOUN
fcis-30297	29	8	,	,	PUNCT
fcis-30297	29	9	time	time	NOUN
fcis-30297	29	10	-	-	PUNCT
fcis-30297	29	11	domain	domain	NOUN
fcis-30297	29	12	,	,	PUNCT
fcis-30297	29	13	or	or	CCONJ
fcis-30297	29	14	instantaneous	instantaneous	ADJ
fcis-30297	29	15	characteristics	characteristic	NOUN
fcis-30297	29	16	of	of	ADP
fcis-30297	29	17	signals	signal	NOUN
fcis-30297	29	18	—	—	PUNCT
fcis-30297	29	19	such	such	ADJ
fcis-30297	29	20	as	as	ADP
fcis-30297	29	21	amplitude	amplitude	NOUN
fcis-30297	29	22	,	,	PUNCT
fcis-30297	29	23	phase	phase	NOUN
fcis-30297	29	24	,	,	PUNCT
fcis-30297	29	25	frequency	frequency	NOUN
fcis-30297	29	26	,	,	PUNCT
fcis-30297	29	27	and	and	CCONJ
fcis-30297	29	28	instantaneous	instantaneous	ADJ
fcis-30297	29	29	power	power	NOUN
fcis-30297	29	30	—	—	PUNCT
fcis-30297	29	31	to	to	PART
fcis-30297	29	32	distinguish	distinguish	VERB
fcis-30297	29	33	between	between	ADP
fcis-30297	29	34	different	different	ADJ
fcis-30297	29	35	modulation	modulation	NOUN
fcis-30297	29	36	types	type	NOUN
fcis-30297	29	37	.	.	PUNCT
fcis-30297	30	1	for	for	ADP
fcis-30297	30	2	example	example	NOUN
fcis-30297	30	3	,	,	PUNCT
fcis-30297	30	4	techniques	technique	NOUN
fcis-30297	30	5	based	base	VERB
fcis-30297	30	6	on	on	ADP
fcis-30297	30	7	higher	high	ADJ
fcis-30297	30	8	-	-	PUNCT
fcis-30297	30	9	order	order	NOUN
fcis-30297	30	10	cumulants	cumulant	NOUN
fcis-30297	30	11	(	(	PUNCT
fcis-30297	30	12	hoc	hoc	X
fcis-30297	30	13	)	)	PUNCT
fcis-30297	30	14	have	have	AUX
fcis-30297	30	15	shown	show	VERB
fcis-30297	30	16	strong	strong	ADJ
fcis-30297	30	17	performance	performance	NOUN
fcis-30297	30	18	in	in	ADP
fcis-30297	30	19	modulation	modulation	NOUN
fcis-30297	30	20	recognition	recognition	NOUN
fcis-30297	30	21	,	,	PUNCT
fcis-30297	30	22	particularly	particularly	ADV
fcis-30297	30	23	in	in	ADP
fcis-30297	30	24	high	high	ADJ
fcis-30297	30	25	signal	signal	NOUN
fcis-30297	30	26	-	-	PUNCT
fcis-30297	30	27	to	to	ADP
fcis-30297	30	28	-	-	PUNCT
fcis-30297	30	29	noise	noise	NOUN
fcis-30297	30	30	ratio	ratio	NOUN
fcis-30297	30	31	(	(	PUNCT
fcis-30297	30	32	snr	snr	NOUN
fcis-30297	30	33	)	)	PUNCT
fcis-30297	30	34	environments	environment	NOUN
fcis-30297	30	35	.	.	PUNCT
fcis-30297	31	1	however	however	ADV
fcis-30297	31	2	,	,	PUNCT
fcis-30297	31	3	these	these	DET
fcis-30297	31	4	traditional	traditional	ADJ
fcis-30297	31	5	methods	method	NOUN
fcis-30297	31	6	heavily	heavily	ADV
fcis-30297	31	7	depend	depend	VERB
fcis-30297	31	8	on	on	ADP
fcis-30297	31	9	the	the	DET
fcis-30297	31	10	accuracy	accuracy	NOUN
fcis-30297	31	11	and	and	CCONJ
fcis-30297	31	12	robustness	robustness	NOUN
fcis-30297	31	13	of	of	ADP
fcis-30297	31	14	manual	manual	ADJ
fcis-30297	31	15	feature	feature	NOUN
fcis-30297	31	16	extraction	extraction	NOUN
fcis-30297	31	17	.	.	PUNCT
fcis-30297	32	1	their	their	PRON
fcis-30297	32	2	performance	performance	NOUN
fcis-30297	32	3	often	often	ADV
fcis-30297	32	4	degrades	degrade	VERB
fcis-30297	32	5	significantly	significantly	ADV
fcis-30297	32	6	in	in	ADP
fcis-30297	32	7	complex	complex	ADJ
fcis-30297	32	8	and	and	CCONJ
fcis-30297	32	9	dynamic	dynamic	ADJ
fcis-30297	32	10	electromagnetic	electromagnetic	ADJ
fcis-30297	32	11	environments	environment	NOUN
fcis-30297	32	12	.	.	PUNCT
fcis-30297	33	1	to	to	PART
fcis-30297	33	2	address	address	VERB
fcis-30297	33	3	these	these	DET
fcis-30297	33	4	limitations	limitation	NOUN
fcis-30297	33	5	,	,	PUNCT
fcis-30297	33	6	machine	machine	NOUN
fcis-30297	33	7	learning	learning	NOUN
fcis-30297	33	8	approaches	approach	NOUN
fcis-30297	33	9	—	—	PUNCT
fcis-30297	33	10	such	such	ADJ
fcis-30297	33	11	as	as	ADP
fcis-30297	33	12	support	support	NOUN
fcis-30297	33	13	vector	vector	NOUN
fcis-30297	33	14	machines	machine	NOUN
fcis-30297	33	15	(	(	PUNCT
fcis-30297	33	16	svms	svms	NOUN
fcis-30297	33	17	)	)	PUNCT
fcis-30297	33	18	and	and	CCONJ
fcis-30297	33	19	decision	decision	NOUN
fcis-30297	33	20	trees	tree	NOUN
fcis-30297	33	21	—	—	PUNCT
fcis-30297	33	22	have	have	AUX
fcis-30297	33	23	been	be	AUX
fcis-30297	33	24	2	2	NUM
fcis-30297	33	25	widely	widely	ADV
fcis-30297	33	26	applied	apply	VERB
fcis-30297	33	27	.	.	PUNCT
fcis-30297	34	1	these	these	DET
fcis-30297	34	2	models	model	NOUN
fcis-30297	34	3	classify	classify	VERB
fcis-30297	34	4	signals	signal	NOUN
fcis-30297	34	5	by	by	ADP
fcis-30297	34	6	feeding	feed	VERB
fcis-30297	34	7	manually	manually	ADV
fcis-30297	34	8	extracted	extract	VERB
fcis-30297	34	9	features	feature	NOUN
fcis-30297	34	10	into	into	ADP
fcis-30297	34	11	the	the	DET
fcis-30297	34	12	classifiers	classifier	NOUN
fcis-30297	34	13	.	.	PUNCT
fcis-30297	35	1	nevertheless	nevertheless	ADV
fcis-30297	35	2	,	,	PUNCT
fcis-30297	35	3	handcrafted	handcrafted	ADJ
fcis-30297	35	4	feature	feature	NOUN
fcis-30297	35	5	extraction	extraction	NOUN
fcis-30297	35	6	remains	remain	VERB
fcis-30297	35	7	a	a	DET
fcis-30297	35	8	critical	critical	ADJ
fcis-30297	35	9	bottleneck	bottleneck	NOUN
fcis-30297	35	10	,	,	PUNCT
fcis-30297	35	11	especially	especially	ADV
fcis-30297	35	12	when	when	SCONJ
fcis-30297	35	13	signals	signal	NOUN
fcis-30297	35	14	are	be	AUX
fcis-30297	35	15	distorted	distort	VERB
fcis-30297	35	16	by	by	ADP
fcis-30297	35	17	noise	noise	NOUN
fcis-30297	35	18	,	,	PUNCT
fcis-30297	35	19	interference	interference	NOUN
fcis-30297	35	20	,	,	PUNCT
fcis-30297	35	21	or	or	CCONJ
fcis-30297	35	22	multipath	multipath	NOUN
fcis-30297	35	23	effects	effect	NOUN
fcis-30297	35	24	,	,	PUNCT
fcis-30297	35	25	leading	lead	VERB
fcis-30297	35	26	to	to	ADP
fcis-30297	35	27	substantial	substantial	ADJ
fcis-30297	35	28	drops	drop	NOUN
fcis-30297	35	29	in	in	ADP
fcis-30297	35	30	recognition	recognition	NOUN
fcis-30297	35	31	performance	performance	NOUN
fcis-30297	35	32	under	under	ADP
fcis-30297	35	33	challenging	challenging	ADJ
fcis-30297	35	34	conditions	condition	NOUN
fcis-30297	35	35	.	.	PUNCT
fcis-30297	36	1	2.2	2.2	NUM
fcis-30297	36	2	.	.	PUNCT
fcis-30297	37	1	deep	deep	ADJ
fcis-30297	37	2	learning	learning	NOUN
fcis-30297	37	3	-	-	PUNCT
fcis-30297	37	4	based	base	VERB
fcis-30297	37	5	signal	signal	NOUN
fcis-30297	37	6	modulation	modulation	NOUN
fcis-30297	37	7	recognition	recognition	NOUN
fcis-30297	37	8	with	with	ADP
fcis-30297	37	9	the	the	DET
fcis-30297	37	10	rapid	rapid	ADJ
fcis-30297	37	11	development	development	NOUN
fcis-30297	37	12	of	of	ADP
fcis-30297	37	13	deep	deep	ADJ
fcis-30297	37	14	learning	learning	NOUN
fcis-30297	37	15	,	,	PUNCT
fcis-30297	37	16	researchers	researcher	NOUN
fcis-30297	37	17	have	have	AUX
fcis-30297	37	18	begun	begin	VERB
fcis-30297	37	19	exploring	explore	VERB
fcis-30297	37	20	the	the	DET
fcis-30297	37	21	use	use	NOUN
fcis-30297	37	22	of	of	ADP
fcis-30297	37	23	deep	deep	ADJ
fcis-30297	37	24	neural	neural	ADJ
fcis-30297	37	25	networks	network	NOUN
fcis-30297	37	26	to	to	PART
fcis-30297	37	27	automatically	automatically	ADV
fcis-30297	37	28	extract	extract	VERB
fcis-30297	37	29	features	feature	NOUN
fcis-30297	37	30	from	from	ADP
fcis-30297	37	31	signals	signal	NOUN
fcis-30297	37	32	,	,	PUNCT
fcis-30297	37	33	addressing	address	VERB
fcis-30297	37	34	the	the	DET
fcis-30297	37	35	shortcomings	shortcoming	NOUN
fcis-30297	37	36	of	of	ADP
fcis-30297	37	37	traditional	traditional	ADJ
fcis-30297	37	38	handcrafted	handcraft	VERB
fcis-30297	37	39	feature	feature	NOUN
fcis-30297	37	40	extraction	extraction	NOUN
fcis-30297	37	41	.	.	PUNCT
fcis-30297	38	1	convolutional	convolutional	ADJ
fcis-30297	38	2	neural	neural	ADJ
fcis-30297	38	3	networks	network	NOUN
fcis-30297	38	4	(	(	PUNCT
fcis-30297	38	5	cnns	cnns	PROPN
fcis-30297	38	6	)	)	PUNCT
fcis-30297	38	7	,	,	PUNCT
fcis-30297	38	8	known	know	VERB
fcis-30297	38	9	for	for	ADP
fcis-30297	38	10	their	their	PRON
fcis-30297	38	11	excellent	excellent	ADJ
fcis-30297	38	12	performance	performance	NOUN
fcis-30297	38	13	in	in	ADP
fcis-30297	38	14	image	image	NOUN
fcis-30297	38	15	processing	processing	NOUN
fcis-30297	38	16	,	,	PUNCT
fcis-30297	38	17	were	be	AUX
fcis-30297	38	18	introduced	introduce	VERB
fcis-30297	38	19	into	into	ADP
fcis-30297	38	20	signal	signal	ADJ
fcis-30297	38	21	modulation	modulation	NOUN
fcis-30297	38	22	recognition	recognition	NOUN
fcis-30297	38	23	.	.	PUNCT
fcis-30297	39	1	o'shea	o'shea	PROPN
fcis-30297	39	2	et	et	PROPN
fcis-30297	39	3	al	al	PROPN
fcis-30297	39	4	.	.	PROPN
fcis-30297	40	1	first	first	PROPN
fcis-30297	40	2	proposed	propose	VERB
fcis-30297	40	3	using	use	VERB
fcis-30297	40	4	cnns	cnn	NOUN
fcis-30297	40	5	to	to	PART
fcis-30297	40	6	directly	directly	ADV
fcis-30297	40	7	learn	learn	VERB
fcis-30297	40	8	features	feature	NOUN
fcis-30297	40	9	from	from	ADP
fcis-30297	40	10	the	the	DET
fcis-30297	40	11	i	i	PROPN
fcis-30297	40	12	/	/	SYM
fcis-30297	40	13	q	q	NOUN
fcis-30297	40	14	data	datum	NOUN
fcis-30297	40	15	of	of	ADP
fcis-30297	40	16	signals	signal	NOUN
fcis-30297	40	17	and	and	CCONJ
fcis-30297	40	18	perform	perform	VERB
fcis-30297	40	19	modulation	modulation	NOUN
fcis-30297	40	20	classification	classification	NOUN
fcis-30297	40	21	.	.	PUNCT
fcis-30297	41	1	this	this	DET
fcis-30297	41	2	method	method	NOUN
fcis-30297	41	3	achieved	achieve	VERB
fcis-30297	41	4	remarkable	remarkable	ADJ
fcis-30297	41	5	results	result	NOUN
fcis-30297	41	6	on	on	ADP
fcis-30297	41	7	multiple	multiple	ADJ
fcis-30297	41	8	public	public	ADJ
fcis-30297	41	9	datasets	dataset	NOUN
fcis-30297	41	10	.	.	PUNCT
fcis-30297	42	1	subsequently	subsequently	ADV
fcis-30297	42	2	,	,	PUNCT
fcis-30297	42	3	recurrent	recurrent	ADJ
fcis-30297	42	4	neural	neural	ADJ
fcis-30297	42	5	networks	network	NOUN
fcis-30297	42	6	(	(	PUNCT
fcis-30297	42	7	rnns	rnns	PROPN
fcis-30297	42	8	)	)	PUNCT
fcis-30297	42	9	and	and	CCONJ
fcis-30297	42	10	their	their	PRON
fcis-30297	42	11	variants	variant	NOUN
fcis-30297	42	12	,	,	PUNCT
fcis-30297	42	13	such	such	ADJ
fcis-30297	42	14	as	as	ADP
fcis-30297	42	15	long	long	ADJ
fcis-30297	42	16	short	short	ADJ
fcis-30297	42	17	-	-	PUNCT
fcis-30297	42	18	term	term	NOUN
fcis-30297	42	19	memory	memory	NOUN
fcis-30297	42	20	networks	network	NOUN
fcis-30297	42	21	(	(	PUNCT
fcis-30297	42	22	lstms	lstms	ADJ
fcis-30297	42	23	)	)	PUNCT
fcis-30297	42	24	,	,	PUNCT
fcis-30297	42	25	were	be	AUX
fcis-30297	42	26	introduced	introduce	VERB
fcis-30297	42	27	to	to	PART
fcis-30297	42	28	capture	capture	VERB
fcis-30297	42	29	temporal	temporal	ADJ
fcis-30297	42	30	information	information	NOUN
fcis-30297	42	31	in	in	ADP
fcis-30297	42	32	signals	signal	NOUN
fcis-30297	42	33	[	[	X
fcis-30297	42	34	9	9	NUM
fcis-30297	42	35	]	]	PUNCT
fcis-30297	42	36	.	.	PUNCT
fcis-30297	43	1	by	by	ADP
fcis-30297	43	2	modeling	model	VERB
fcis-30297	43	3	the	the	DET
fcis-30297	43	4	temporal	temporal	ADJ
fcis-30297	43	5	dependencies	dependency	NOUN
fcis-30297	43	6	of	of	ADP
fcis-30297	43	7	the	the	DET
fcis-30297	43	8	signals	signal	NOUN
fcis-30297	43	9	,	,	PUNCT
fcis-30297	43	10	these	these	DET
fcis-30297	43	11	methods	method	NOUN
fcis-30297	43	12	have	have	AUX
fcis-30297	43	13	demonstrated	demonstrate	VERB
fcis-30297	43	14	strong	strong	ADJ
fcis-30297	43	15	robustness	robustness	NOUN
fcis-30297	43	16	in	in	ADP
fcis-30297	43	17	low	low	ADJ
fcis-30297	43	18	snr	snr	NOUN
fcis-30297	43	19	environments	environment	NOUN
fcis-30297	43	20	.	.	PUNCT
fcis-30297	44	1	however	however	ADV
fcis-30297	44	2	,	,	PUNCT
fcis-30297	44	3	rnn	rnn	NOUN
fcis-30297	44	4	-	-	PUNCT
fcis-30297	44	5	based	base	VERB
fcis-30297	44	6	models	model	NOUN
fcis-30297	44	7	are	be	AUX
fcis-30297	44	8	time	time	NOUN
fcis-30297	44	9	-	-	PUNCT
fcis-30297	44	10	consuming	consume	VERB
fcis-30297	44	11	to	to	PART
fcis-30297	44	12	train	train	VERB
fcis-30297	44	13	and	and	CCONJ
fcis-30297	44	14	are	be	AUX
fcis-30297	44	15	prone	prone	ADJ
fcis-30297	44	16	to	to	ADP
fcis-30297	44	17	gradient	gradient	VERB
fcis-30297	44	18	vanishing	vanishing	NOUN
fcis-30297	44	19	or	or	CCONJ
fcis-30297	44	20	exploding	explode	VERB
fcis-30297	44	21	issues	issue	NOUN
fcis-30297	44	22	.	.	PUNCT
fcis-30297	45	1	in	in	ADP
fcis-30297	45	2	recent	recent	ADJ
fcis-30297	45	3	years	year	NOUN
fcis-30297	45	4	,	,	PUNCT
fcis-30297	45	5	hybrid	hybrid	ADJ
fcis-30297	45	6	architectures	architecture	NOUN
fcis-30297	45	7	combining	combine	VERB
fcis-30297	45	8	cnns	cnn	NOUN
fcis-30297	45	9	and	and	CCONJ
fcis-30297	45	10	rnns	rnn	NOUN
fcis-30297	45	11	,	,	PUNCT
fcis-30297	45	12	such	such	ADJ
fcis-30297	45	13	as	as	ADP
fcis-30297	45	14	cnn	cnn	PROPN
fcis-30297	45	15	-	-	PUNCT
fcis-30297	45	16	rnn	rnn	PROPN
fcis-30297	45	17	and	and	CCONJ
fcis-30297	45	18	cnn	cnn	PROPN
fcis-30297	45	19	-	-	PUNCT
fcis-30297	45	20	lstm	lstm	ADJ
fcis-30297	45	21	models	model	NOUN
fcis-30297	45	22	,	,	PUNCT
fcis-30297	45	23	have	have	AUX
fcis-30297	45	24	gained	gain	VERB
fcis-30297	45	25	attention	attention	NOUN
fcis-30297	45	26	.	.	PUNCT
fcis-30297	46	1	these	these	DET
fcis-30297	46	2	models	model	NOUN
fcis-30297	46	3	improve	improve	VERB
fcis-30297	46	4	modulation	modulation	NOUN
fcis-30297	46	5	recognition	recognition	NOUN
fcis-30297	46	6	accuracy	accuracy	NOUN
fcis-30297	46	7	by	by	ADP
fcis-30297	46	8	fusing	fuse	VERB
fcis-30297	46	9	spatial	spatial	ADJ
fcis-30297	46	10	and	and	CCONJ
fcis-30297	46	11	temporal	temporal	ADJ
fcis-30297	46	12	features	feature	NOUN
fcis-30297	46	13	.	.	PUNCT
fcis-30297	47	1	although	although	SCONJ
fcis-30297	47	2	these	these	DET
fcis-30297	47	3	deep	deep	ADJ
fcis-30297	47	4	learning	learning	NOUN
fcis-30297	47	5	methods	method	NOUN
fcis-30297	47	6	outperform	outperform	VERB
fcis-30297	47	7	traditional	traditional	ADJ
fcis-30297	47	8	methods	method	NOUN
fcis-30297	47	9	,	,	PUNCT
fcis-30297	47	10	they	they	PRON
fcis-30297	47	11	often	often	ADV
fcis-30297	47	12	require	require	VERB
fcis-30297	47	13	substantial	substantial	ADJ
fcis-30297	47	14	computational	computational	ADJ
fcis-30297	47	15	resources	resource	NOUN
fcis-30297	47	16	and	and	CCONJ
fcis-30297	47	17	training	training	NOUN
fcis-30297	47	18	time	time	NOUN
fcis-30297	47	19	,	,	PUNCT
fcis-30297	47	20	posing	pose	VERB
fcis-30297	47	21	challenges	challenge	NOUN
fcis-30297	47	22	for	for	ADP
fcis-30297	47	23	real	real	ADJ
fcis-30297	47	24	-	-	PUNCT
fcis-30297	47	25	time	time	NOUN
fcis-30297	47	26	applications	application	NOUN
fcis-30297	47	27	.	.	PUNCT
fcis-30297	48	1	2.3	2.3	NUM
fcis-30297	48	2	.	.	PUNCT
fcis-30297	49	1	lightweight	lightweight	ADJ
fcis-30297	49	2	networks	network	NOUN
fcis-30297	49	3	and	and	CCONJ
fcis-30297	49	4	transformer	transformer	NOUN
fcis-30297	49	5	applications	application	NOUN
fcis-30297	49	6	to	to	PART
fcis-30297	49	7	address	address	VERB
fcis-30297	49	8	the	the	DET
fcis-30297	49	9	high	high	ADJ
fcis-30297	49	10	computational	computational	ADJ
fcis-30297	49	11	complexity	complexity	NOUN
fcis-30297	49	12	of	of	ADP
fcis-30297	49	13	deep	deep	ADJ
fcis-30297	49	14	learning	learning	NOUN
fcis-30297	49	15	models	model	NOUN
fcis-30297	49	16	,	,	PUNCT
fcis-30297	49	17	lightweight	lightweight	ADJ
fcis-30297	49	18	network	network	NOUN
fcis-30297	49	19	structures	structure	NOUN
fcis-30297	49	20	have	have	AUX
fcis-30297	49	21	been	be	AUX
fcis-30297	49	22	widely	widely	ADV
fcis-30297	49	23	researched	research	VERB
fcis-30297	49	24	.	.	PUNCT
fcis-30297	50	1	lightweight	lightweight	PROPN
fcis-30297	50	2	cnn	cnn	PROPN
fcis-30297	50	3	models	model	NOUN
fcis-30297	50	4	,	,	PUNCT
fcis-30297	50	5	such	such	ADJ
fcis-30297	50	6	as	as	ADP
fcis-30297	50	7	mobilenet	mobilenet	NOUN
fcis-30297	50	8	and	and	CCONJ
fcis-30297	50	9	squeezenet	squeezenet	NOUN
fcis-30297	50	10	,	,	PUNCT
fcis-30297	50	11	reduce	reduce	VERB
fcis-30297	50	12	the	the	DET
fcis-30297	50	13	number	number	NOUN
fcis-30297	50	14	of	of	ADP
fcis-30297	50	15	parameters	parameter	NOUN
fcis-30297	50	16	and	and	CCONJ
fcis-30297	50	17	computation	computation	NOUN
fcis-30297	50	18	,	,	PUNCT
fcis-30297	50	19	enabling	enable	VERB
fcis-30297	50	20	efficient	efficient	ADJ
fcis-30297	50	21	performance	performance	NOUN
fcis-30297	50	22	on	on	ADP
fcis-30297	50	23	resourceconstrained	resourceconstraine	VERB
fcis-30297	50	24	platforms	platform	NOUN
fcis-30297	50	25	like	like	ADP
fcis-30297	50	26	mobile	mobile	ADJ
fcis-30297	50	27	devices	device	NOUN
fcis-30297	50	28	.	.	PUNCT
fcis-30297	51	1	while	while	SCONJ
fcis-30297	51	2	these	these	DET
fcis-30297	51	3	models	model	NOUN
fcis-30297	51	4	perform	perform	VERB
fcis-30297	51	5	well	well	ADV
fcis-30297	51	6	in	in	ADP
fcis-30297	51	7	specific	specific	ADJ
fcis-30297	51	8	application	application	NOUN
fcis-30297	51	9	scenarios	scenario	NOUN
fcis-30297	51	10	,	,	PUNCT
fcis-30297	51	11	they	they	PRON
fcis-30297	51	12	still	still	ADV
fcis-30297	51	13	struggle	struggle	VERB
fcis-30297	51	14	with	with	ADP
fcis-30297	51	15	accuracy	accuracy	NOUN
fcis-30297	51	16	when	when	SCONJ
fcis-30297	51	17	handling	handle	VERB
fcis-30297	51	18	complex	complex	ADJ
fcis-30297	51	19	signal	signal	ADJ
fcis-30297	51	20	modulation	modulation	NOUN
fcis-30297	51	21	recognition	recognition	NOUN
fcis-30297	51	22	tasks	task	NOUN
fcis-30297	51	23	.	.	PUNCT
fcis-30297	52	1	the	the	DET
fcis-30297	52	2	transformer	transformer	NOUN
fcis-30297	52	3	structure	structure	NOUN
fcis-30297	52	4	,	,	PUNCT
fcis-30297	52	5	known	know	VERB
fcis-30297	52	6	for	for	ADP
fcis-30297	52	7	its	its	PRON
fcis-30297	52	8	success	success	NOUN
fcis-30297	52	9	in	in	ADP
fcis-30297	52	10	natural	natural	ADJ
fcis-30297	52	11	language	language	NOUN
fcis-30297	52	12	processing	processing	NOUN
fcis-30297	52	13	,	,	PUNCT
fcis-30297	52	14	has	have	AUX
fcis-30297	52	15	garnered	garner	VERB
fcis-30297	52	16	attention	attention	NOUN
fcis-30297	52	17	and	and	CCONJ
fcis-30297	52	18	has	have	AUX
fcis-30297	52	19	been	be	AUX
fcis-30297	52	20	gradually	gradually	ADV
fcis-30297	52	21	introduced	introduce	VERB
fcis-30297	52	22	into	into	ADP
fcis-30297	52	23	visual	visual	ADJ
fcis-30297	52	24	tasks	task	NOUN
fcis-30297	52	25	.	.	PUNCT
fcis-30297	53	1	its	its	PRON
fcis-30297	53	2	self	self	NOUN
fcis-30297	53	3	-	-	PUNCT
fcis-30297	53	4	attention	attention	NOUN
fcis-30297	53	5	mechanism	mechanism	NOUN
fcis-30297	53	6	can	can	AUX
fcis-30297	53	7	capture	capture	VERB
fcis-30297	53	8	global	global	ADJ
fcis-30297	53	9	features	feature	NOUN
fcis-30297	53	10	without	without	ADP
fcis-30297	53	11	relying	rely	VERB
fcis-30297	53	12	on	on	ADP
fcis-30297	53	13	temporal	temporal	ADJ
fcis-30297	53	14	sequences	sequence	NOUN
fcis-30297	53	15	,	,	PUNCT
fcis-30297	53	16	providing	provide	VERB
fcis-30297	53	17	new	new	ADJ
fcis-30297	53	18	approaches	approach	NOUN
fcis-30297	53	19	to	to	PART
fcis-30297	53	20	feature	feature	NOUN
fcis-30297	53	21	extraction	extraction	NOUN
fcis-30297	53	22	.	.	PUNCT
fcis-30297	54	1	rt	rt	PROPN
fcis-30297	54	2	-	-	PUNCT
fcis-30297	54	3	detr	detr	NOUN
fcis-30297	54	4	,	,	PUNCT
fcis-30297	54	5	a	a	DET
fcis-30297	54	6	lightweight	lightweight	ADJ
fcis-30297	54	7	object	object	NOUN
fcis-30297	54	8	detection	detection	NOUN
fcis-30297	54	9	model	model	NOUN
fcis-30297	54	10	that	that	PRON
fcis-30297	54	11	integrates	integrate	VERB
fcis-30297	54	12	the	the	DET
fcis-30297	54	13	advantages	advantage	NOUN
fcis-30297	54	14	of	of	ADP
fcis-30297	54	15	the	the	DET
fcis-30297	54	16	transformer	transformer	NOUN
fcis-30297	54	17	,	,	PUNCT
fcis-30297	54	18	can	can	AUX
fcis-30297	54	19	extract	extract	VERB
fcis-30297	54	20	rich	rich	ADJ
fcis-30297	54	21	feature	feature	NOUN
fcis-30297	54	22	information	information	NOUN
fcis-30297	54	23	while	while	SCONJ
fcis-30297	54	24	maintaining	maintain	VERB
fcis-30297	54	25	computational	computational	ADJ
fcis-30297	54	26	efficiency	efficiency	NOUN
fcis-30297	54	27	,	,	PUNCT
fcis-30297	54	28	making	make	VERB
fcis-30297	54	29	it	it	PRON
fcis-30297	54	30	highly	highly	ADV
fcis-30297	54	31	promising	promising	ADJ
fcis-30297	54	32	for	for	ADP
fcis-30297	54	33	signal	signal	ADJ
fcis-30297	54	34	modulation	modulation	NOUN
fcis-30297	54	35	recognition	recognition	NOUN
fcis-30297	54	36	tasks	task	NOUN
fcis-30297	54	37	.	.	PUNCT
fcis-30297	55	1	2.4	2.4	NUM
fcis-30297	55	2	.	.	PUNCT
fcis-30297	55	3	baseline	baseline	PROPN
fcis-30297	55	4	model	model	PROPN
fcis-30297	55	5	–	–	PUNCT
fcis-30297	55	6	rt	rt	PROPN
fcis-30297	55	7	-	-	PUNCT
fcis-30297	55	8	detr	detr	PROPN
fcis-30297	55	9	rt	rt	PROPN
fcis-30297	55	10	-	-	PUNCT
fcis-30297	55	11	detr	detr	NOUN
fcis-30297	55	12	(	(	PUNCT
fcis-30297	55	13	real	real	ADJ
fcis-30297	55	14	-	-	PUNCT
fcis-30297	55	15	time	time	NOUN
fcis-30297	55	16	detection	detection	NOUN
fcis-30297	55	17	transformer	transformer	NOUN
fcis-30297	55	18	)	)	PUNCT
fcis-30297	55	19	is	be	AUX
fcis-30297	55	20	a	a	DET
fcis-30297	55	21	recent	recent	ADJ
fcis-30297	55	22	end	end	NOUN
fcis-30297	55	23	-	-	PUNCT
fcis-30297	55	24	to	to	ADP
fcis-30297	55	25	-	-	PUNCT
fcis-30297	55	26	end	end	NOUN
fcis-30297	55	27	object	object	NOUN
fcis-30297	55	28	detection	detection	NOUN
fcis-30297	55	29	model	model	NOUN
fcis-30297	55	30	that	that	PRON
fcis-30297	55	31	eliminates	eliminate	VERB
fcis-30297	55	32	the	the	DET
fcis-30297	55	33	need	need	NOUN
fcis-30297	55	34	for	for	ADP
fcis-30297	55	35	post	post	ADJ
fcis-30297	55	36	-	-	ADJ
fcis-30297	55	37	processing	processing	ADJ
fcis-30297	55	38	steps	step	NOUN
fcis-30297	55	39	such	such	ADJ
fcis-30297	55	40	as	as	ADP
fcis-30297	55	41	non	non	ADJ
fcis-30297	55	42	-	-	ADJ
fcis-30297	55	43	maximum	maximum	ADJ
fcis-30297	55	44	suppression	suppression	NOUN
fcis-30297	55	45	(	(	PUNCT
fcis-30297	55	46	nms	nms	NOUN
fcis-30297	55	47	)	)	PUNCT
fcis-30297	55	48	.	.	PUNCT
fcis-30297	56	1	built	build	VERB
fcis-30297	56	2	upon	upon	SCONJ
fcis-30297	56	3	the	the	DET
fcis-30297	56	4	detr	detr	NOUN
fcis-30297	56	5	framework	framework	NOUN
fcis-30297	56	6	,	,	PUNCT
fcis-30297	56	7	rt	rt	PROPN
fcis-30297	56	8	-	-	PUNCT
fcis-30297	56	9	detr	detr	NOUN
fcis-30297	56	10	introduces	introduce	NOUN
fcis-30297	56	11	several	several	ADJ
fcis-30297	56	12	architectural	architectural	ADJ
fcis-30297	56	13	optimizations	optimization	NOUN
fcis-30297	56	14	to	to	PART
fcis-30297	56	15	achieve	achieve	VERB
fcis-30297	56	16	a	a	DET
fcis-30297	56	17	better	well	ADJ
fcis-30297	56	18	balance	balance	NOUN
fcis-30297	56	19	between	between	ADP
fcis-30297	56	20	accuracy	accuracy	NOUN
fcis-30297	56	21	and	and	CCONJ
fcis-30297	56	22	inference	inference	NOUN
fcis-30297	56	23	speed	speed	NOUN
fcis-30297	56	24	,	,	PUNCT
fcis-30297	56	25	making	make	VERB
fcis-30297	56	26	it	it	PRON
fcis-30297	56	27	the	the	DET
fcis-30297	56	28	first	first	ADJ
fcis-30297	56	29	transformer	transformer	NOUN
fcis-30297	56	30	-	-	PUNCT
fcis-30297	56	31	based	base	VERB
fcis-30297	56	32	object	object	NOUN
fcis-30297	56	33	detector	detector	NOUN
fcis-30297	56	34	to	to	PART
fcis-30297	56	35	operate	operate	VERB
fcis-30297	56	36	effectively	effectively	ADV
fcis-30297	56	37	in	in	ADP
fcis-30297	56	38	real	real	ADJ
fcis-30297	56	39	-	-	PUNCT
fcis-30297	56	40	time	time	NOUN
fcis-30297	56	41	.	.	PUNCT
fcis-30297	57	1	the	the	DET
fcis-30297	57	2	model	model	NOUN
fcis-30297	57	3	consists	consist	VERB
fcis-30297	57	4	of	of	ADP
fcis-30297	57	5	four	four	NUM
fcis-30297	57	6	main	main	ADJ
fcis-30297	57	7	components	component	NOUN
fcis-30297	57	8	:	:	PUNCT
fcis-30297	57	9	the	the	DET
fcis-30297	57	10	input	input	NOUN
fcis-30297	57	11	module	module	NOUN
fcis-30297	57	12	,	,	PUNCT
fcis-30297	57	13	backbone	backbone	NOUN
fcis-30297	57	14	,	,	PUNCT
fcis-30297	57	15	neck	neck	NOUN
fcis-30297	57	16	encoder	encoder	NOUN
fcis-30297	57	17	,	,	PUNCT
fcis-30297	57	18	and	and	CCONJ
fcis-30297	57	19	head	head	NOUN
fcis-30297	57	20	decoder	decoder	NOUN
fcis-30297	57	21	,	,	PUNCT
fcis-30297	57	22	as	as	SCONJ
fcis-30297	57	23	illustrated	illustrate	VERB
fcis-30297	57	24	in	in	ADP
fcis-30297	57	25	figure	figure	NOUN
fcis-30297	57	26	1	1	NUM
fcis-30297	57	27	.	.	PUNCT
fcis-30297	57	28	figure	figure	NOUN
fcis-30297	57	29	1	1	NUM
fcis-30297	57	30	.	.	PUNCT
fcis-30297	57	31	overview	overview	NOUN
fcis-30297	57	32	of	of	ADP
fcis-30297	57	33	rt	rt	NOUN
fcis-30297	57	34	-	-	PUNCT
fcis-30297	57	35	detr	detr	NOUN
fcis-30297	57	36	2.4.1	2.4.1	NUM
fcis-30297	57	37	.	.	PUNCT
fcis-30297	58	1	input	input	NOUN
fcis-30297	58	2	module	module	NOUN
fcis-30297	58	3	the	the	DET
fcis-30297	58	4	input	input	NOUN
fcis-30297	58	5	module	module	NOUN
fcis-30297	58	6	is	be	AUX
fcis-30297	58	7	responsible	responsible	ADJ
fcis-30297	58	8	for	for	ADP
fcis-30297	58	9	preprocessing	preprocesse	VERB
fcis-30297	58	10	raw	raw	ADJ
fcis-30297	58	11	images	image	NOUN
fcis-30297	58	12	into	into	ADP
fcis-30297	58	13	a	a	DET
fcis-30297	58	14	unified	unified	ADJ
fcis-30297	58	15	format	format	NOUN
fcis-30297	58	16	suitable	suitable	ADJ
fcis-30297	58	17	for	for	ADP
fcis-30297	58	18	the	the	DET
fcis-30297	58	19	model	model	NOUN
fcis-30297	58	20	.	.	PUNCT
fcis-30297	59	1	typically	typically	ADV
fcis-30297	59	2	,	,	PUNCT
fcis-30297	59	3	images	image	NOUN
fcis-30297	59	4	are	be	AUX
fcis-30297	59	5	resized	resize	VERB
fcis-30297	59	6	to	to	ADP
fcis-30297	59	7	a	a	DET
fcis-30297	59	8	fixed	fix	VERB
fcis-30297	59	9	resolution	resolution	NOUN
fcis-30297	59	10	(	(	PUNCT
fcis-30297	59	11	e.g.	e.g.	ADV
fcis-30297	59	12	,	,	PUNCT
fcis-30297	59	13	640	640	NUM
fcis-30297	59	14	×	×	NOUN
fcis-30297	59	15	640	640	NUM
fcis-30297	59	16	)	)	PUNCT
fcis-30297	59	17	and	and	CCONJ
fcis-30297	59	18	normalized	normalize	VERB
fcis-30297	59	19	.	.	PUNCT
fcis-30297	60	1	to	to	PART
fcis-30297	60	2	enhance	enhance	VERB
fcis-30297	60	3	generalization	generalization	NOUN
fcis-30297	60	4	performance	performance	NOUN
fcis-30297	60	5	,	,	PUNCT
fcis-30297	60	6	a	a	DET
fcis-30297	60	7	variety	variety	NOUN
fcis-30297	60	8	of	of	ADP
fcis-30297	60	9	data	datum	NOUN
fcis-30297	60	10	augmentation	augmentation	NOUN
fcis-30297	60	11	techniques	technique	NOUN
fcis-30297	60	12	are	be	AUX
fcis-30297	60	13	applied	apply	VERB
fcis-30297	60	14	during	during	ADP
fcis-30297	60	15	training	training	NOUN
fcis-30297	60	16	2.4.2	2.4.2	NUM
fcis-30297	60	17	.	.	PUNCT
fcis-30297	61	1	backbone	backbone	VERB
fcis-30297	61	2	the	the	DET
fcis-30297	61	3	backbone	backbone	NOUN
fcis-30297	61	4	of	of	ADP
fcis-30297	61	5	rt	rt	PROPN
fcis-30297	61	6	-	-	PUNCT
fcis-30297	61	7	detr	detr	NOUN
fcis-30297	61	8	is	be	AUX
fcis-30297	61	9	a	a	DET
fcis-30297	61	10	convolutional	convolutional	ADJ
fcis-30297	61	11	neural	neural	ADJ
fcis-30297	61	12	network	network	NOUN
fcis-30297	61	13	used	use	VERB
fcis-30297	61	14	to	to	PART
fcis-30297	61	15	extract	extract	VERB
fcis-30297	61	16	multi	multi	ADJ
fcis-30297	61	17	-	-	ADJ
fcis-30297	61	18	scale	scale	ADJ
fcis-30297	61	19	semantic	semantic	ADJ
fcis-30297	61	20	features	feature	NOUN
fcis-30297	61	21	from	from	ADP
fcis-30297	61	22	the	the	DET
fcis-30297	61	23	input	input	NOUN
fcis-30297	61	24	image	image	NOUN
fcis-30297	61	25	.	.	PUNCT
fcis-30297	62	1	common	common	ADJ
fcis-30297	62	2	backbone	backbone	NOUN
fcis-30297	62	3	options	option	NOUN
fcis-30297	62	4	include	include	VERB
fcis-30297	62	5	resnet-50	resnet-50	PROPN
fcis-30297	62	6	and	and	CCONJ
fcis-30297	62	7	resnet-101	resnet-101	NOUN
fcis-30297	62	8	.	.	PUNCT
fcis-30297	63	1	the	the	DET
fcis-30297	63	2	model	model	NOUN
fcis-30297	63	3	utilizes	utilize	VERB
fcis-30297	63	4	three	three	NUM
fcis-30297	63	5	feature	feature	NOUN
fcis-30297	63	6	levels	level	NOUN
fcis-30297	63	7	(	(	PUNCT
fcis-30297	63	8	denoted	denote	VERB
fcis-30297	63	9	as	as	ADP
fcis-30297	63	10	s3	s3	PROPN
fcis-30297	63	11	,	,	PUNCT
fcis-30297	63	12	s4	s4	PROPN
fcis-30297	63	13	,	,	PUNCT
fcis-30297	63	14	and	and	CCONJ
fcis-30297	63	15	s5	s5	NOUN
fcis-30297	63	16	)	)	PUNCT
fcis-30297	63	17	,	,	PUNCT
fcis-30297	63	18	each	each	PRON
fcis-30297	63	19	corresponding	correspond	VERB
fcis-30297	63	20	to	to	ADP
fcis-30297	63	21	a	a	DET
fcis-30297	63	22	different	different	ADJ
fcis-30297	63	23	receptive	receptive	ADJ
fcis-30297	63	24	field	field	NOUN
fcis-30297	63	25	and	and	CCONJ
fcis-30297	63	26	semantic	semantic	ADJ
fcis-30297	63	27	level	level	NOUN
fcis-30297	63	28	.	.	PUNCT
fcis-30297	64	1	2.4.3	2.4.3	X
fcis-30297	64	2	.	.	PUNCT
fcis-30297	64	3	neck	neck	NOUN
fcis-30297	64	4	encoder	encoder	NOUN
fcis-30297	64	5	to	to	PART
fcis-30297	64	6	address	address	VERB
fcis-30297	64	7	the	the	DET
fcis-30297	64	8	inefficiency	inefficiency	NOUN
fcis-30297	64	9	of	of	ADP
fcis-30297	64	10	the	the	DET
fcis-30297	64	11	original	original	ADJ
fcis-30297	64	12	detr	detr	NOUN
fcis-30297	64	13	encoder	encoder	NOUN
fcis-30297	64	14	,	,	PUNCT
fcis-30297	64	15	rt	rt	PROPN
fcis-30297	64	16	-	-	PUNCT
fcis-30297	64	17	detr	detr	NOUN
fcis-30297	64	18	proposes	propose	VERB
fcis-30297	64	19	an	an	DET
fcis-30297	64	20	efficient	efficient	ADJ
fcis-30297	64	21	hybrid	hybrid	NOUN
fcis-30297	64	22	encoder	encoder	NOUN
fcis-30297	64	23	consisting	consist	VERB
fcis-30297	64	24	of	of	ADP
fcis-30297	64	25	two	two	NUM
fcis-30297	64	26	components	component	NOUN
fcis-30297	64	27	:	:	PUNCT
fcis-30297	64	28	aifi	aifi	NOUN
fcis-30297	64	29	(	(	PUNCT
fcis-30297	64	30	attention	attention	NOUN
fcis-30297	64	31	-	-	PUNCT
fcis-30297	64	32	based	base	VERB
fcis-30297	64	33	intra	intra	ADJ
fcis-30297	64	34	-	-	ADJ
fcis-30297	64	35	scale	scale	ADJ
fcis-30297	64	36	feature	feature	NOUN
fcis-30297	64	37	interaction	interaction	NOUN
fcis-30297	64	38	):	):	PUNCT
fcis-30297	64	39	applies	apply	VERB
fcis-30297	64	40	self	self	NOUN
fcis-30297	64	41	-	-	PUNCT
fcis-30297	64	42	attention	attention	NOUN
fcis-30297	64	43	only	only	ADV
fcis-30297	64	44	to	to	ADP
fcis-30297	64	45	high	high	ADJ
fcis-30297	64	46	-	-	PUNCT
fcis-30297	64	47	level	level	NOUN
fcis-30297	64	48	features	feature	NOUN
fcis-30297	64	49	(	(	PUNCT
fcis-30297	64	50	s5	s5	PROPN
fcis-30297	64	51	)	)	PUNCT
fcis-30297	64	52	,	,	PUNCT
fcis-30297	64	53	modeling	model	VERB
fcis-30297	64	54	long	long	ADJ
fcis-30297	64	55	-	-	PUNCT
fcis-30297	64	56	range	range	NOUN
fcis-30297	64	57	dependencies	dependency	NOUN
fcis-30297	64	58	while	while	SCONJ
fcis-30297	64	59	reducing	reduce	VERB
fcis-30297	64	60	sequence	sequence	NOUN
fcis-30297	64	61	length	length	NOUN
fcis-30297	64	62	and	and	CCONJ
fcis-30297	64	63	computation	computation	NOUN
fcis-30297	64	64	cost	cost	NOUN
fcis-30297	64	65	;	;	PUNCT
fcis-30297	64	66	ccff	ccff	PROPN
fcis-30297	64	67	(	(	PUNCT
fcis-30297	64	68	cnn	cnn	PROPN
fcis-30297	64	69	-	-	PUNCT
fcis-30297	64	70	based	base	VERB
fcis-30297	64	71	cross	cross	ADJ
fcis-30297	64	72	-	-	ADJ
fcis-30297	64	73	scale	scale	ADJ
fcis-30297	64	74	feature	feature	NOUN
fcis-30297	64	75	fusion	fusion	NOUN
fcis-30297	64	76	):	):	PUNCT
fcis-30297	64	77	a	a	DET
fcis-30297	64	78	lightweight	lightweight	ADJ
fcis-30297	64	79	convolutional	convolutional	ADJ
fcis-30297	64	80	module	module	NOUN
fcis-30297	64	81	that	that	PRON
fcis-30297	64	82	fuses	fuse	VERB
fcis-30297	64	83	s3	s3	PROPN
fcis-30297	64	84	,	,	PUNCT
fcis-30297	64	85	s4	s4	PROPN
fcis-30297	64	86	,	,	PUNCT
fcis-30297	64	87	and	and	CCONJ
fcis-30297	64	88	the	the	DET
fcis-30297	64	89	transformed	transform	VERB
fcis-30297	64	90	s5	s5	PROPN
fcis-30297	64	91	features	feature	VERB
fcis-30297	64	92	to	to	PART
fcis-30297	64	93	integrate	integrate	VERB
fcis-30297	64	94	fine	fine	ADV
fcis-30297	64	95	-	-	PUNCT
fcis-30297	64	96	grained	grain	VERB
fcis-30297	64	97	and	and	CCONJ
fcis-30297	64	98	semantic	semantic	ADJ
fcis-30297	64	99	information	information	NOUN
fcis-30297	64	100	across	across	ADP
fcis-30297	64	101	scales	scale	NOUN
fcis-30297	64	102	.	.	PUNCT
fcis-30297	65	1	this	this	DET
fcis-30297	65	2	decoupled	decouple	VERB
fcis-30297	65	3	design	design	NOUN
fcis-30297	65	4	of	of	ADP
fcis-30297	65	5	intra	intra	ADJ
fcis-30297	65	6	-	-	ADJ
fcis-30297	65	7	scale	scale	ADJ
fcis-30297	65	8	and	and	CCONJ
fcis-30297	65	9	cross	cross	ADJ
fcis-30297	65	10	-	-	ADJ
fcis-30297	65	11	scale	scale	ADJ
fcis-30297	65	12	operations	operation	NOUN
fcis-30297	65	13	ensures	ensure	VERB
fcis-30297	65	14	efficient	efficient	ADJ
fcis-30297	65	15	computation	computation	NOUN
fcis-30297	65	16	without	without	ADP
fcis-30297	65	17	sacrificing	sacrifice	VERB
fcis-30297	65	18	detection	detection	NOUN
fcis-30297	65	19	performance	performance	NOUN
fcis-30297	65	20	.	.	PUNCT
fcis-30297	66	1	3	3	NUM
fcis-30297	66	2	2.4.4	2.4.4	NUM
fcis-30297	66	3	.	.	PUNCT
fcis-30297	67	1	head	head	NOUN
fcis-30297	67	2	decoder	decoder	NOUN
fcis-30297	67	3	the	the	DET
fcis-30297	67	4	head	head	NOUN
fcis-30297	67	5	decoder	decoder	NOUN
fcis-30297	67	6	predicts	predict	NOUN
fcis-30297	67	7	object	object	VERB
fcis-30297	67	8	categories	category	NOUN
fcis-30297	67	9	and	and	CCONJ
fcis-30297	67	10	bounding	bounding	NOUN
fcis-30297	67	11	boxes	box	NOUN
fcis-30297	67	12	based	base	VERB
fcis-30297	67	13	on	on	ADP
fcis-30297	67	14	the	the	DET
fcis-30297	67	15	features	feature	NOUN
fcis-30297	67	16	processed	process	VERB
fcis-30297	67	17	by	by	ADP
fcis-30297	67	18	the	the	DET
fcis-30297	67	19	encoder	encoder	NOUN
fcis-30297	67	20	.	.	PUNCT
fcis-30297	68	1	it	it	PRON
fcis-30297	68	2	retains	retain	VERB
fcis-30297	68	3	the	the	DET
fcis-30297	68	4	standard	standard	ADJ
fcis-30297	68	5	transformer	transformer	NOUN
fcis-30297	68	6	decoder	decoder	NOUN
fcis-30297	68	7	structure	structure	NOUN
fcis-30297	68	8	and	and	CCONJ
fcis-30297	68	9	introduces	introduce	VERB
fcis-30297	68	10	a	a	DET
fcis-30297	68	11	novel	novel	ADJ
fcis-30297	68	12	uncertainty	uncertainty	NOUN
fcis-30297	68	13	-	-	PUNCT
fcis-30297	68	14	minimal	minimal	ADJ
fcis-30297	68	15	query	query	NOUN
fcis-30297	68	16	selection	selection	NOUN
fcis-30297	68	17	mechanism	mechanism	NOUN
fcis-30297	68	18	,	,	PUNCT
fcis-30297	68	19	characterized	characterize	VERB
fcis-30297	68	20	by	by	ADP
fcis-30297	68	21	:	:	PUNCT
fcis-30297	68	22	evaluating	evaluate	VERB
fcis-30297	68	23	both	both	CCONJ
fcis-30297	68	24	classification	classification	NOUN
fcis-30297	68	25	and	and	CCONJ
fcis-30297	68	26	localization	localization	NOUN
fcis-30297	68	27	confidence	confidence	NOUN
fcis-30297	68	28	to	to	PART
fcis-30297	68	29	quantify	quantify	VERB
fcis-30297	68	30	the	the	DET
fcis-30297	68	31	uncertainty	uncertainty	NOUN
fcis-30297	68	32	of	of	ADP
fcis-30297	68	33	each	each	DET
fcis-30297	68	34	encoder	encoder	NOUN
fcis-30297	68	35	feature	feature	NOUN
fcis-30297	68	36	;	;	PUNCT
fcis-30297	68	37	selecting	select	VERB
fcis-30297	68	38	the	the	DET
fcis-30297	68	39	most	most	ADV
fcis-30297	68	40	reliable	reliable	ADJ
fcis-30297	68	41	features	feature	NOUN
fcis-30297	68	42	as	as	ADP
fcis-30297	68	43	object	object	NOUN
fcis-30297	68	44	queries	query	NOUN
fcis-30297	68	45	;	;	PUNCT
fcis-30297	68	46	iteratively	iteratively	ADV
fcis-30297	68	47	refining	refine	VERB
fcis-30297	68	48	these	these	DET
fcis-30297	68	49	queries	query	NOUN
fcis-30297	68	50	through	through	ADP
fcis-30297	68	51	multiple	multiple	ADJ
fcis-30297	68	52	decoder	decoder	NOUN
fcis-30297	68	53	layers	layer	NOUN
fcis-30297	68	54	to	to	PART
fcis-30297	68	55	produce	produce	VERB
fcis-30297	68	56	final	final	ADJ
fcis-30297	68	57	predictions	prediction	NOUN
fcis-30297	68	58	.	.	PUNCT
fcis-30297	69	1	unlike	unlike	ADP
fcis-30297	69	2	traditional	traditional	ADJ
fcis-30297	69	3	detectors	detector	NOUN
fcis-30297	69	4	,	,	PUNCT
fcis-30297	69	5	rt	rt	PROPN
fcis-30297	69	6	-	-	PUNCT
fcis-30297	69	7	detr	detr	NOUN
fcis-30297	69	8	eliminates	eliminate	VERB
fcis-30297	69	9	the	the	DET
fcis-30297	69	10	need	need	NOUN
fcis-30297	69	11	for	for	ADP
fcis-30297	69	12	nms	nms	NOUN
fcis-30297	69	13	by	by	ADP
fcis-30297	69	14	employing	employ	VERB
fcis-30297	69	15	hungarian	hungarian	ADJ
fcis-30297	69	16	matching	matching	NOUN
fcis-30297	69	17	to	to	PART
fcis-30297	69	18	establish	establish	VERB
fcis-30297	69	19	a	a	DET
fcis-30297	69	20	one	one	NUM
fcis-30297	69	21	-	-	PUNCT
fcis-30297	69	22	to	to	ADP
fcis-30297	69	23	-	-	PUNCT
fcis-30297	69	24	one	one	NUM
fcis-30297	69	25	correspondence	correspondence	NOUN
fcis-30297	69	26	between	between	ADP
fcis-30297	69	27	predictions	prediction	NOUN
fcis-30297	69	28	and	and	CCONJ
fcis-30297	69	29	groundtruth	groundtruth	NOUN
fcis-30297	69	30	objects	object	NOUN
fcis-30297	69	31	.	.	PUNCT
fcis-30297	70	1	this	this	PRON
fcis-30297	70	2	enables	enable	VERB
fcis-30297	70	3	fully	fully	ADV
fcis-30297	70	4	end	end	NOUN
fcis-30297	70	5	-	-	PUNCT
fcis-30297	70	6	to	to	ADP
fcis-30297	70	7	-	-	PUNCT
fcis-30297	70	8	end	end	NOUN
fcis-30297	70	9	training	training	NOUN
fcis-30297	70	10	and	and	CCONJ
fcis-30297	70	11	simplifies	simplify	VERB
fcis-30297	70	12	the	the	DET
fcis-30297	70	13	detection	detection	NOUN
fcis-30297	70	14	pipeline	pipeline	NOUN
fcis-30297	70	15	.	.	PUNCT
fcis-30297	71	1	2.5	2.5	NUM
fcis-30297	71	2	.	.	PUNCT
fcis-30297	72	1	model	model	NOUN
fcis-30297	72	2	enhancement	enhancement	NOUN
fcis-30297	72	3	and	and	CCONJ
fcis-30297	72	4	feature	feature	NOUN
fcis-30297	72	5	fusion	fusion	NOUN
fcis-30297	72	6	techniques	technique	NOUN
fcis-30297	72	7	to	to	PART
fcis-30297	72	8	further	far	ADV
fcis-30297	72	9	enhance	enhance	VERB
fcis-30297	72	10	the	the	DET
fcis-30297	72	11	performance	performance	NOUN
fcis-30297	72	12	of	of	ADP
fcis-30297	72	13	deep	deep	ADJ
fcis-30297	72	14	learning	learning	NOUN
fcis-30297	72	15	models	model	NOUN
fcis-30297	72	16	in	in	ADP
fcis-30297	72	17	signal	signal	ADJ
fcis-30297	72	18	modulation	modulation	NOUN
fcis-30297	72	19	recognition	recognition	NOUN
fcis-30297	72	20	,	,	PUNCT
fcis-30297	72	21	researchers	researcher	NOUN
fcis-30297	72	22	have	have	AUX
fcis-30297	72	23	proposed	propose	VERB
fcis-30297	72	24	various	various	ADJ
fcis-30297	72	25	enhancement	enhancement	NOUN
fcis-30297	72	26	and	and	CCONJ
fcis-30297	72	27	optimization	optimization	NOUN
fcis-30297	72	28	strategies	strategy	NOUN
fcis-30297	72	29	.	.	PUNCT
fcis-30297	73	1	attention	attention	NOUN
fcis-30297	73	2	mechanisms	mechanism	NOUN
fcis-30297	73	3	have	have	AUX
fcis-30297	73	4	been	be	AUX
fcis-30297	73	5	widely	widely	ADV
fcis-30297	73	6	applied	apply	VERB
fcis-30297	73	7	in	in	ADP
fcis-30297	73	8	deep	deep	ADJ
fcis-30297	73	9	learning	learning	NOUN
fcis-30297	73	10	,	,	PUNCT
fcis-30297	73	11	particularly	particularly	ADV
fcis-30297	73	12	when	when	SCONJ
fcis-30297	73	13	dealing	deal	VERB
fcis-30297	73	14	with	with	ADP
fcis-30297	73	15	signals	signal	NOUN
fcis-30297	73	16	with	with	ADP
fcis-30297	73	17	complex	complex	ADJ
fcis-30297	73	18	patterns	pattern	NOUN
fcis-30297	73	19	,	,	PUNCT
fcis-30297	73	20	as	as	SCONJ
fcis-30297	73	21	they	they	PRON
fcis-30297	73	22	effectively	effectively	ADV
fcis-30297	73	23	enhance	enhance	VERB
fcis-30297	73	24	the	the	DET
fcis-30297	73	25	expression	expression	NOUN
fcis-30297	73	26	of	of	ADP
fcis-30297	73	27	important	important	ADJ
fcis-30297	73	28	features	feature	NOUN
fcis-30297	73	29	.	.	PUNCT
fcis-30297	74	1	aifi	aifi	NOUN
fcis-30297	74	2	,	,	PUNCT
fcis-30297	74	3	a	a	DET
fcis-30297	74	4	dynamic	dynamic	ADJ
fcis-30297	74	5	attention	attention	NOUN
fcis-30297	74	6	mechanism	mechanism	NOUN
fcis-30297	74	7	,	,	PUNCT
fcis-30297	74	8	can	can	AUX
fcis-30297	74	9	adaptively	adaptively	ADV
fcis-30297	74	10	adjust	adjust	VERB
fcis-30297	74	11	attention	attention	NOUN
fcis-30297	74	12	weights	weight	NOUN
fcis-30297	74	13	according	accord	VERB
fcis-30297	74	14	to	to	ADP
fcis-30297	74	15	the	the	DET
fcis-30297	74	16	characteristics	characteristic	NOUN
fcis-30297	74	17	of	of	ADP
fcis-30297	74	18	the	the	DET
fcis-30297	74	19	input	input	NOUN
fcis-30297	74	20	signal	signal	NOUN
fcis-30297	74	21	,	,	PUNCT
fcis-30297	74	22	improving	improve	VERB
fcis-30297	74	23	the	the	DET
fcis-30297	74	24	model	model	NOUN
fcis-30297	74	25	's	's	PART
fcis-30297	74	26	feature	feature	NOUN
fcis-30297	74	27	extraction	extraction	NOUN
fcis-30297	74	28	capabilities	capability	NOUN
fcis-30297	74	29	.	.	PUNCT
fcis-30297	75	1	moreover	moreover	ADV
fcis-30297	75	2	,	,	PUNCT
fcis-30297	75	3	feature	feature	NOUN
fcis-30297	75	4	fusion	fusion	NOUN
fcis-30297	75	5	techniques	technique	NOUN
fcis-30297	75	6	have	have	AUX
fcis-30297	75	7	been	be	AUX
fcis-30297	75	8	applied	apply	VERB
fcis-30297	75	9	in	in	ADP
fcis-30297	75	10	multimodal	multimodal	NOUN
fcis-30297	75	11	signal	signal	NOUN
fcis-30297	75	12	processing	processing	NOUN
fcis-30297	75	13	and	and	CCONJ
fcis-30297	75	14	complex	complex	ADJ
fcis-30297	75	15	scene	scene	NOUN
fcis-30297	75	16	recognition	recognition	NOUN
fcis-30297	75	17	.	.	PUNCT
fcis-30297	76	1	the	the	DET
fcis-30297	76	2	aifi	aifi	NOUN
fcis-30297	76	3	-	-	PUNCT
fcis-30297	76	4	efficientadditive	efficientadditive	ADJ
fcis-30297	76	5	module	module	NOUN
fcis-30297	76	6	improves	improve	VERB
fcis-30297	76	7	traditional	traditional	ADJ
fcis-30297	76	8	additive	additive	ADJ
fcis-30297	76	9	feature	feature	NOUN
fcis-30297	76	10	fusion	fusion	NOUN
fcis-30297	76	11	methods	method	NOUN
fcis-30297	76	12	,	,	PUNCT
fcis-30297	76	13	enhancing	enhance	VERB
fcis-30297	76	14	the	the	DET
fcis-30297	76	15	synergy	synergy	NOUN
fcis-30297	76	16	between	between	ADP
fcis-30297	76	17	features	feature	NOUN
fcis-30297	76	18	while	while	SCONJ
fcis-30297	76	19	maintaining	maintain	VERB
fcis-30297	76	20	computational	computational	ADJ
fcis-30297	76	21	efficiency	efficiency	NOUN
fcis-30297	76	22	,	,	PUNCT
fcis-30297	76	23	thereby	thereby	ADV
fcis-30297	76	24	improving	improve	VERB
fcis-30297	76	25	modulation	modulation	NOUN
fcis-30297	76	26	recognition	recognition	NOUN
fcis-30297	76	27	accuracy	accuracy	NOUN
fcis-30297	76	28	.	.	PUNCT
fcis-30297	77	1	2.6	2.6	NUM
fcis-30297	77	2	.	.	PUNCT
fcis-30297	78	1	contributions	contribution	NOUN
fcis-30297	78	2	of	of	ADP
fcis-30297	78	3	this	this	DET
fcis-30297	78	4	research	research	NOUN
fcis-30297	78	5	building	building	NOUN
fcis-30297	78	6	upon	upon	SCONJ
fcis-30297	78	7	previous	previous	ADJ
fcis-30297	78	8	research	research	NOUN
fcis-30297	78	9	,	,	PUNCT
fcis-30297	78	10	this	this	DET
fcis-30297	78	11	study	study	NOUN
fcis-30297	78	12	presents	present	VERB
fcis-30297	78	13	a	a	DET
fcis-30297	78	14	novel	novel	ADJ
fcis-30297	78	15	signal	signal	NOUN
fcis-30297	78	16	modulation	modulation	NOUN
fcis-30297	78	17	recognition	recognition	NOUN
fcis-30297	78	18	method	method	NOUN
fcis-30297	78	19	based	base	VERB
fcis-30297	78	20	on	on	ADP
fcis-30297	78	21	the	the	DET
fcis-30297	78	22	rtdetr	rtdetr	NOUN
fcis-30297	78	23	architecture	architecture	NOUN
fcis-30297	78	24	.	.	PUNCT
fcis-30297	79	1	by	by	ADP
fcis-30297	79	2	incorporating	incorporate	VERB
fcis-30297	79	3	efficientformerv2	efficientformerv2	NOUN
fcis-30297	79	4	as	as	ADP
fcis-30297	79	5	the	the	DET
fcis-30297	79	6	backbone	backbone	NOUN
fcis-30297	79	7	network	network	NOUN
fcis-30297	79	8	and	and	CCONJ
fcis-30297	79	9	integrating	integrate	VERB
fcis-30297	79	10	the	the	DET
fcis-30297	79	11	dynamic	dynamic	ADJ
fcis-30297	79	12	-	-	PUNCT
fcis-30297	79	13	range	range	NOUN
fcis-30297	79	14	histogram	histogram	NOUN
fcis-30297	79	15	self	self	NOUN
fcis-30297	79	16	-	-	PUNCT
fcis-30297	79	17	attention	attention	NOUN
fcis-30297	79	18	(	(	PUNCT
fcis-30297	79	19	dhs	dhs	PROPN
fcis-30297	79	20	-	-	PUNCT
fcis-30297	79	21	attention	attention	NOUN
fcis-30297	79	22	)	)	PUNCT
fcis-30297	79	23	mechanism	mechanism	NOUN
fcis-30297	79	24	,	,	PUNCT
fcis-30297	79	25	the	the	DET
fcis-30297	79	26	proposed	propose	VERB
fcis-30297	79	27	model	model	NOUN
fcis-30297	79	28	achieves	achieve	VERB
fcis-30297	79	29	significant	significant	ADJ
fcis-30297	79	30	improvements	improvement	NOUN
fcis-30297	79	31	in	in	ADP
fcis-30297	79	32	both	both	DET
fcis-30297	79	33	recognition	recognition	NOUN
fcis-30297	79	34	accuracy	accuracy	NOUN
fcis-30297	79	35	and	and	CCONJ
fcis-30297	79	36	computational	computational	ADJ
fcis-30297	79	37	efficiency	efficiency	NOUN
fcis-30297	79	38	,	,	PUNCT
fcis-30297	79	39	while	while	SCONJ
fcis-30297	79	40	maintaining	maintain	VERB
fcis-30297	79	41	a	a	DET
fcis-30297	79	42	lightweight	lightweight	ADJ
fcis-30297	79	43	structure	structure	NOUN
fcis-30297	79	44	.	.	PUNCT
fcis-30297	80	1	extensive	extensive	ADJ
fcis-30297	80	2	experiments	experiment	NOUN
fcis-30297	80	3	conducted	conduct	VERB
fcis-30297	80	4	under	under	ADP
fcis-30297	80	5	various	various	ADJ
fcis-30297	80	6	complex	complex	ADJ
fcis-30297	80	7	channel	channel	NOUN
fcis-30297	80	8	conditions	condition	NOUN
fcis-30297	80	9	demonstrate	demonstrate	VERB
fcis-30297	80	10	that	that	SCONJ
fcis-30297	80	11	the	the	DET
fcis-30297	80	12	proposed	propose	VERB
fcis-30297	80	13	approach	approach	NOUN
fcis-30297	80	14	consistently	consistently	ADV
fcis-30297	80	15	outperforms	outperform	VERB
fcis-30297	80	16	traditional	traditional	ADJ
fcis-30297	80	17	methods	method	NOUN
fcis-30297	80	18	in	in	ADP
fcis-30297	80	19	terms	term	NOUN
fcis-30297	80	20	of	of	ADP
fcis-30297	80	21	both	both	PRON
fcis-30297	80	22	classification	classification	NOUN
fcis-30297	80	23	performance	performance	NOUN
fcis-30297	80	24	and	and	CCONJ
fcis-30297	80	25	processing	processing	NOUN
fcis-30297	80	26	speed	speed	NOUN
fcis-30297	80	27	.	.	PUNCT
fcis-30297	81	1	3	3	X
fcis-30297	81	2	.	.	X
fcis-30297	81	3	wideband	wideband	NOUN
fcis-30297	81	4	modulated	modulate	VERB
fcis-30297	81	5	signal	signal	NOUN
fcis-30297	81	6	dataset	dataset	VERB
fcis-30297	81	7	3.1	3.1	NUM
fcis-30297	81	8	.	.	PUNCT
fcis-30297	82	1	data	datum	NOUN
fcis-30297	82	2	preprocessing	preprocesse	VERB
fcis-30297	82	3	before	before	ADP
fcis-30297	82	4	performing	perform	VERB
fcis-30297	82	5	signal	signal	NOUN
fcis-30297	82	6	modulation	modulation	NOUN
fcis-30297	82	7	recognition	recognition	NOUN
fcis-30297	82	8	,	,	PUNCT
fcis-30297	82	9	we	we	PRON
fcis-30297	82	10	preprocess	preprocess	VERB
fcis-30297	82	11	the	the	DET
fcis-30297	82	12	input	input	NOUN
fcis-30297	82	13	signal	signal	NOUN
fcis-30297	82	14	data	datum	NOUN
fcis-30297	82	15	using	use	VERB
fcis-30297	82	16	the	the	DET
fcis-30297	82	17	following	following	ADJ
fcis-30297	82	18	methods	method	NOUN
fcis-30297	82	19	:	:	PUNCT
fcis-30297	82	20	normalization	normalization	NOUN
fcis-30297	82	21	:	:	PUNCT
fcis-30297	82	22	all	all	DET
fcis-30297	82	23	input	input	NOUN
fcis-30297	82	24	signal	signal	NOUN
fcis-30297	82	25	data	datum	NOUN
fcis-30297	82	26	is	be	AUX
fcis-30297	82	27	normalized	normalize	VERB
fcis-30297	82	28	to	to	ADP
fcis-30297	82	29	the	the	DET
fcis-30297	82	30	[	[	X
fcis-30297	82	31	0,1	0,1	NUM
fcis-30297	82	32	]	]	PUNCT
fcis-30297	82	33	range	range	NOUN
fcis-30297	82	34	to	to	PART
fcis-30297	82	35	eliminate	eliminate	VERB
fcis-30297	82	36	amplitude	amplitude	NOUN
fcis-30297	82	37	differences	difference	NOUN
fcis-30297	82	38	between	between	ADP
fcis-30297	82	39	signals	signal	NOUN
fcis-30297	82	40	,	,	PUNCT
fcis-30297	82	41	reducing	reduce	VERB
fcis-30297	82	42	their	their	PRON
fcis-30297	82	43	impact	impact	NOUN
fcis-30297	82	44	on	on	ADP
fcis-30297	82	45	the	the	DET
fcis-30297	82	46	model[10	model[10	NOUN
fcis-30297	82	47	]	]	PUNCT
fcis-30297	82	48	.	.	PUNCT
fcis-30297	83	1	signal	signal	ADJ
fcis-30297	83	2	framing	framing	NOUN
fcis-30297	83	3	and	and	CCONJ
fcis-30297	83	4	overlapping	overlap	VERB
fcis-30297	83	5	:	:	PUNCT
fcis-30297	83	6	to	to	PART
fcis-30297	83	7	capture	capture	VERB
fcis-30297	83	8	richer	rich	ADJ
fcis-30297	83	9	temporal	temporal	ADJ
fcis-30297	83	10	information	information	NOUN
fcis-30297	83	11	,	,	PUNCT
fcis-30297	83	12	the	the	DET
fcis-30297	83	13	input	input	NOUN
fcis-30297	83	14	signal	signal	NOUN
fcis-30297	83	15	is	be	AUX
fcis-30297	83	16	divided	divide	VERB
fcis-30297	83	17	into	into	ADP
fcis-30297	83	18	multiple	multiple	ADJ
fcis-30297	83	19	frames	frame	NOUN
fcis-30297	83	20	,	,	PUNCT
fcis-30297	83	21	with	with	ADP
fcis-30297	83	22	a	a	DET
fcis-30297	83	23	certain	certain	ADJ
fcis-30297	83	24	overlap	overlap	NOUN
fcis-30297	83	25	between	between	ADP
fcis-30297	83	26	frames	frame	NOUN
fcis-30297	83	27	.	.	PUNCT
fcis-30297	84	1	each	each	DET
fcis-30297	84	2	frame	frame	NOUN
fcis-30297	84	3	of	of	ADP
fcis-30297	84	4	the	the	DET
fcis-30297	84	5	signal	signal	NOUN
fcis-30297	84	6	is	be	AUX
fcis-30297	84	7	treated	treat	VERB
fcis-30297	84	8	as	as	ADP
fcis-30297	84	9	an	an	DET
fcis-30297	84	10	independent	independent	ADJ
fcis-30297	84	11	input	input	NOUN
fcis-30297	84	12	sample	sample	NOUN
fcis-30297	84	13	.	.	PUNCT
fcis-30297	85	1	data	datum	NOUN
fcis-30297	85	2	augmentation	augmentation	NOUN
fcis-30297	85	3	:	:	PUNCT
fcis-30297	85	4	to	to	PART
fcis-30297	85	5	increase	increase	VERB
fcis-30297	85	6	the	the	DET
fcis-30297	85	7	diversity	diversity	NOUN
fcis-30297	85	8	of	of	ADP
fcis-30297	85	9	the	the	DET
fcis-30297	85	10	training	training	NOUN
fcis-30297	85	11	data	datum	NOUN
fcis-30297	85	12	,	,	PUNCT
fcis-30297	85	13	we	we	PRON
fcis-30297	85	14	applied	apply	VERB
fcis-30297	85	15	data	datum	NOUN
fcis-30297	85	16	augmentation	augmentation	NOUN
fcis-30297	85	17	techniques	technique	NOUN
fcis-30297	85	18	such	such	ADJ
fcis-30297	85	19	as	as	ADP
fcis-30297	85	20	random	random	ADJ
fcis-30297	85	21	noise	noise	NOUN
fcis-30297	85	22	addition	addition	NOUN
fcis-30297	85	23	and	and	CCONJ
fcis-30297	85	24	frequency	frequency	NOUN
fcis-30297	85	25	shifting[11	shifting[11	NOUN
fcis-30297	85	26	]	]	PUNCT
fcis-30297	85	27	.	.	PUNCT
fcis-30297	86	1	these	these	DET
fcis-30297	86	2	augmentations	augmentation	NOUN
fcis-30297	86	3	help	help	VERB
fcis-30297	86	4	the	the	DET
fcis-30297	86	5	model	model	NOUN
fcis-30297	86	6	generalize	generalize	VERB
fcis-30297	86	7	better	well	ADV
fcis-30297	86	8	to	to	ADP
fcis-30297	86	9	different	different	ADJ
fcis-30297	86	10	channel	channel	NOUN
fcis-30297	86	11	environments	environment	NOUN
fcis-30297	86	12	.	.	PUNCT
fcis-30297	87	1	3.2	3.2	NUM
fcis-30297	87	2	.	.	PUNCT
fcis-30297	87	3	network	network	NOUN
fcis-30297	87	4	architecture	architecture	NOUN
fcis-30297	87	5	design	design	VERB
fcis-30297	87	6	the	the	DET
fcis-30297	87	7	overall	overall	ADJ
fcis-30297	87	8	structure	structure	NOUN
fcis-30297	87	9	of	of	ADP
fcis-30297	87	10	the	the	DET
fcis-30297	87	11	rt	rt	PROPN
fcis-30297	87	12	-	-	PUNCT
fcis-30297	87	13	detr	detr	NOUN
fcis-30297	87	14	model	model	NOUN
fcis-30297	87	15	is	be	AUX
fcis-30297	87	16	composed	compose	VERB
fcis-30297	87	17	of	of	ADP
fcis-30297	87	18	the	the	DET
fcis-30297	87	19	following	follow	VERB
fcis-30297	87	20	parts	part	NOUN
fcis-30297	87	21	:	:	PUNCT
fcis-30297	87	22	backbone	backbone	NOUN
fcis-30297	87	23	network	network	NOUN
fcis-30297	87	24	:	:	PUNCT
fcis-30297	87	25	the	the	DET
fcis-30297	87	26	original	original	ADJ
fcis-30297	87	27	resnet-18	resnet-18	PROPN
fcis-30297	87	28	backbone	backbone	NOUN
fcis-30297	87	29	in	in	ADP
fcis-30297	87	30	rt	rt	PROPN
fcis-30297	87	31	-	-	PUNCT
fcis-30297	87	32	detr	detr	NOUN
fcis-30297	87	33	was	be	AUX
fcis-30297	87	34	replaced	replace	VERB
fcis-30297	87	35	with	with	ADP
fcis-30297	87	36	.	.	PUNCT
fcis-30297	88	1	efficientformerv2	efficientformerv2	PROPN
fcis-30297	88	2	provides	provide	VERB
fcis-30297	88	3	a	a	DET
fcis-30297	88	4	better	well	ADJ
fcis-30297	88	5	trade	trade	NOUN
fcis-30297	88	6	-	-	PUNCT
fcis-30297	88	7	off	off	NOUN
fcis-30297	88	8	between	between	ADP
fcis-30297	88	9	performance	performance	NOUN
fcis-30297	88	10	and	and	CCONJ
fcis-30297	88	11	efficiency	efficiency	NOUN
fcis-30297	88	12	,	,	PUNCT
fcis-30297	88	13	offering	offer	VERB
fcis-30297	88	14	stronger	strong	ADJ
fcis-30297	88	15	feature	feature	NOUN
fcis-30297	88	16	representation	representation	NOUN
fcis-30297	88	17	while	while	SCONJ
fcis-30297	88	18	maintaining	maintain	VERB
fcis-30297	88	19	a	a	DET
fcis-30297	88	20	lightweight	lightweight	ADJ
fcis-30297	88	21	structure	structure	NOUN
fcis-30297	88	22	.	.	PUNCT
fcis-30297	89	1	this	this	PRON
fcis-30297	89	2	allows	allow	VERB
fcis-30297	89	3	the	the	DET
fcis-30297	89	4	model	model	NOUN
fcis-30297	89	5	to	to	PART
fcis-30297	89	6	extract	extract	VERB
fcis-30297	89	7	more	more	ADJ
fcis-30297	89	8	discriminative	discriminative	NOUN
fcis-30297	89	9	time	time	NOUN
fcis-30297	89	10	-	-	PUNCT
fcis-30297	89	11	frequency	frequency	NOUN
fcis-30297	89	12	features	feature	NOUN
fcis-30297	89	13	,	,	PUNCT
fcis-30297	89	14	improving	improve	VERB
fcis-30297	89	15	recognition	recognition	NOUN
fcis-30297	89	16	accuracy	accuracy	NOUN
fcis-30297	89	17	for	for	ADP
fcis-30297	89	18	complex	complex	ADJ
fcis-30297	89	19	modulation	modulation	NOUN
fcis-30297	89	20	types	type	NOUN
fcis-30297	89	21	.	.	PUNCT
fcis-30297	90	1	transformer	transformer	ADJ
fcis-30297	90	2	encoder	encoder	NOUN
fcis-30297	90	3	:	:	PUNCT
fcis-30297	90	4	after	after	ADP
fcis-30297	90	5	extracting	extract	VERB
fcis-30297	90	6	primary	primary	ADJ
fcis-30297	90	7	features	feature	NOUN
fcis-30297	90	8	,	,	PUNCT
fcis-30297	90	9	the	the	DET
fcis-30297	90	10	feature	feature	NOUN
fcis-30297	90	11	maps	map	NOUN
fcis-30297	90	12	are	be	AUX
fcis-30297	90	13	input	input	VERB
fcis-30297	90	14	into	into	ADP
fcis-30297	90	15	the	the	DET
fcis-30297	90	16	transformer	transformer	NOUN
fcis-30297	90	17	encoder	encoder	NOUN
fcis-30297	90	18	.	.	PUNCT
fcis-30297	91	1	the	the	DET
fcis-30297	91	2	self	self	NOUN
fcis-30297	91	3	-	-	PUNCT
fcis-30297	91	4	attention	attention	NOUN
fcis-30297	91	5	mechanism	mechanism	NOUN
fcis-30297	91	6	of	of	ADP
fcis-30297	91	7	the	the	DET
fcis-30297	91	8	transformer	transformer	NOUN
fcis-30297	91	9	allows	allow	VERB
fcis-30297	91	10	it	it	PRON
fcis-30297	91	11	to	to	PART
fcis-30297	91	12	effectively	effectively	ADV
fcis-30297	91	13	capture	capture	VERB
fcis-30297	91	14	global	global	ADJ
fcis-30297	91	15	information	information	NOUN
fcis-30297	91	16	in	in	ADP
fcis-30297	91	17	the	the	DET
fcis-30297	91	18	signal	signal	NOUN
fcis-30297	91	19	,	,	PUNCT
fcis-30297	91	20	enhancing	enhance	VERB
fcis-30297	91	21	recognition	recognition	NOUN
fcis-30297	91	22	performance	performance	NOUN
fcis-30297	91	23	[	[	X
fcis-30297	91	24	13	13	NUM
fcis-30297	91	25	]	]	PUNCT
fcis-30297	91	26	.	.	PUNCT
fcis-30297	92	1	aifi	aifi	PROPN
fcis-30297	92	2	module	module	NOUN
fcis-30297	92	3	:	:	PUNCT
fcis-30297	92	4	to	to	PART
fcis-30297	92	5	further	far	ADV
fcis-30297	92	6	refine	refine	VERB
fcis-30297	92	7	the	the	DET
fcis-30297	92	8	attention	attention	NOUN
fcis-30297	92	9	mechanism	mechanism	NOUN
fcis-30297	92	10	following	follow	VERB
fcis-30297	92	11	the	the	DET
fcis-30297	92	12	transformer	transformer	NOUN
fcis-30297	92	13	encoder	encoder	NOUN
fcis-30297	92	14	,	,	PUNCT
fcis-30297	92	15	we	we	PRON
fcis-30297	92	16	replaced	replace	VERB
fcis-30297	92	17	the	the	DET
fcis-30297	92	18	aifi	aifi	NOUN
fcis-30297	92	19	module	module	NOUN
fcis-30297	92	20	with	with	ADP
fcis-30297	92	21	the	the	DET
fcis-30297	92	22	dynamic	dynamic	ADJ
fcis-30297	92	23	-	-	PUNCT
fcis-30297	92	24	range	range	NOUN
fcis-30297	92	25	histogram	histogram	NOUN
fcis-30297	92	26	self	self	NOUN
fcis-30297	92	27	-	-	PUNCT
fcis-30297	92	28	attention	attention	NOUN
fcis-30297	92	29	(	(	PUNCT
fcis-30297	92	30	dhs	dhs	PROPN
fcis-30297	92	31	-	-	PUNCT
fcis-30297	92	32	attention	attention	NOUN
fcis-30297	92	33	)	)	PUNCT
fcis-30297	92	34	module	module	NOUN
fcis-30297	92	35	,	,	PUNCT
fcis-30297	92	36	inspired	inspire	VERB
fcis-30297	92	37	by	by	ADP
fcis-30297	92	38	histoformer	histoformer	PROPN
fcis-30297	92	39	.	.	PUNCT
fcis-30297	93	1	dhsattention	dhsattention	NOUN
fcis-30297	93	2	introduces	introduce	VERB
fcis-30297	93	3	histogram	histogram	NOUN
fcis-30297	93	4	-	-	PUNCT
fcis-30297	93	5	based	base	VERB
fcis-30297	93	6	dynamic	dynamic	ADJ
fcis-30297	93	7	range	range	NOUN
fcis-30297	93	8	modeling	modeling	NOUN
fcis-30297	93	9	,	,	PUNCT
fcis-30297	93	10	which	which	PRON
fcis-30297	93	11	enables	enable	VERB
fcis-30297	93	12	the	the	DET
fcis-30297	93	13	network	network	NOUN
fcis-30297	93	14	to	to	PART
fcis-30297	93	15	capture	capture	VERB
fcis-30297	93	16	long	long	ADJ
fcis-30297	93	17	-	-	PUNCT
fcis-30297	93	18	range	range	NOUN
fcis-30297	93	19	dependencies	dependency	NOUN
fcis-30297	93	20	more	more	ADV
fcis-30297	93	21	effectively	effectively	ADV
fcis-30297	93	22	,	,	PUNCT
fcis-30297	93	23	enhancing	enhance	VERB
fcis-30297	93	24	its	its	PRON
fcis-30297	93	25	ability	ability	NOUN
fcis-30297	93	26	to	to	PART
fcis-30297	93	27	distinguish	distinguish	VERB
fcis-30297	93	28	subtle	subtle	ADJ
fcis-30297	93	29	differences	difference	NOUN
fcis-30297	93	30	among	among	ADP
fcis-30297	93	31	modulation	modulation	NOUN
fcis-30297	93	32	patterns	pattern	NOUN
fcis-30297	93	33	.	.	PUNCT
fcis-30297	94	1	classifier	classifier	NOUN
fcis-30297	94	2	:	:	PUNCT
fcis-30297	94	3	the	the	DET
fcis-30297	94	4	final	final	ADJ
fcis-30297	94	5	feature	feature	NOUN
fcis-30297	94	6	map	map	NOUN
fcis-30297	94	7	is	be	AUX
fcis-30297	94	8	input	input	VERB
fcis-30297	94	9	into	into	ADP
fcis-30297	94	10	a	a	DET
fcis-30297	94	11	fully	fully	ADV
fcis-30297	94	12	connected	connect	VERB
fcis-30297	94	13	layer	layer	NOUN
fcis-30297	94	14	for	for	ADP
fcis-30297	94	15	classification	classification	NOUN
fcis-30297	94	16	.	.	PUNCT
fcis-30297	95	1	this	this	DET
fcis-30297	95	2	layer	layer	NOUN
fcis-30297	95	3	outputs	output	VERB
fcis-30297	95	4	the	the	DET
fcis-30297	95	5	probability	probability	NOUN
fcis-30297	95	6	distribution	distribution	NOUN
fcis-30297	95	7	corresponding	correspond	VERB
fcis-30297	95	8	to	to	ADP
fcis-30297	95	9	various	various	ADJ
fcis-30297	95	10	modulation	modulation	NOUN
fcis-30297	95	11	signals	signal	NOUN
fcis-30297	95	12	,	,	PUNCT
fcis-30297	95	13	with	with	ADP
fcis-30297	95	14	the	the	DET
fcis-30297	95	15	predicted	predict	VERB
fcis-30297	95	16	modulation	modulation	NOUN
fcis-30297	95	17	type	type	NOUN
fcis-30297	95	18	being	be	AUX
fcis-30297	95	19	the	the	DET
fcis-30297	95	20	one	one	NOUN
fcis-30297	95	21	with	with	ADP
fcis-30297	95	22	the	the	DET
fcis-30297	95	23	highest	high	ADJ
fcis-30297	95	24	probability	probability	NOUN
fcis-30297	95	25	.	.	PUNCT
fcis-30297	96	1	3.3	3.3	NUM
fcis-30297	96	2	.	.	PUNCT
fcis-30297	96	3	model	model	NOUN
fcis-30297	96	4	training	training	NOUN
fcis-30297	96	5	and	and	CCONJ
fcis-30297	96	6	optimization	optimization	NOUN
fcis-30297	96	7	we	we	PRON
fcis-30297	96	8	use	use	VERB
fcis-30297	96	9	the	the	DET
fcis-30297	96	10	cross	cross	ADJ
fcis-30297	96	11	-	-	ADJ
fcis-30297	96	12	entropy	entropy	ADJ
fcis-30297	96	13	loss	loss	NOUN
fcis-30297	96	14	function	function	NOUN
fcis-30297	96	15	as	as	ADP
fcis-30297	96	16	the	the	DET
fcis-30297	96	17	optimization	optimization	NOUN
fcis-30297	96	18	objective	objective	NOUN
fcis-30297	96	19	of	of	ADP
fcis-30297	96	20	the	the	DET
fcis-30297	96	21	model	model	NOUN
fcis-30297	96	22	.	.	PUNCT
fcis-30297	97	1	the	the	DET
fcis-30297	97	2	model	model	NOUN
fcis-30297	97	3	training	training	NOUN
fcis-30297	97	4	process	process	NOUN
fcis-30297	97	5	uses	use	VERB
fcis-30297	97	6	the	the	DET
fcis-30297	97	7	adam	adam	PROPN
fcis-30297	97	8	optimizer	optimizer	NOUN
fcis-30297	97	9	with	with	ADP
fcis-30297	97	10	the	the	DET
fcis-30297	97	11	following	follow	VERB
fcis-30297	97	12	hyperparameters	hyperparameter	NOUN
fcis-30297	97	13	:	:	PUNCT
fcis-30297	97	14	learning	learn	VERB
fcis-30297	97	15	rate	rate	NOUN
fcis-30297	97	16	:	:	PUNCT
fcis-30297	97	17	the	the	DET
fcis-30297	97	18	initial	initial	ADJ
fcis-30297	97	19	learning	learning	NOUN
fcis-30297	97	20	rate	rate	NOUN
fcis-30297	97	21	is	be	AUX
fcis-30297	97	22	set	set	VERB
fcis-30297	97	23	to	to	ADP
fcis-30297	97	24	0.001	0.001	NUM
fcis-30297	97	25	,	,	PUNCT
fcis-30297	97	26	and	and	CCONJ
fcis-30297	97	27	a	a	DET
fcis-30297	97	28	cosine	cosine	NOUN
fcis-30297	97	29	annealing	anneal	VERB
fcis-30297	97	30	strategy	strategy	NOUN
fcis-30297	97	31	is	be	AUX
fcis-30297	97	32	used	use	VERB
fcis-30297	97	33	to	to	PART
fcis-30297	97	34	dynamically	dynamically	ADV
fcis-30297	97	35	adjust	adjust	VERB
fcis-30297	97	36	the	the	DET
fcis-30297	97	37	learning	learning	NOUN
fcis-30297	97	38	rate	rate	NOUN
fcis-30297	97	39	to	to	PART
fcis-30297	97	40	ensure	ensure	VERB
fcis-30297	97	41	stability	stability	NOUN
fcis-30297	97	42	and	and	CCONJ
fcis-30297	97	43	fast	fast	ADJ
fcis-30297	97	44	convergence	convergence	NOUN
fcis-30297	97	45	during	during	ADP
fcis-30297	97	46	training	training	NOUN
fcis-30297	97	47	.	.	PUNCT
fcis-30297	98	1	batch	batch	NOUN
fcis-30297	98	2	size	size	NOUN
fcis-30297	98	3	:	:	PUNCT
fcis-30297	98	4	the	the	DET
fcis-30297	98	5	batch	batch	NOUN
fcis-30297	98	6	size	size	NOUN
fcis-30297	98	7	is	be	AUX
fcis-30297	98	8	set	set	VERB
fcis-30297	98	9	to	to	ADP
fcis-30297	98	10	4	4	NUM
fcis-30297	98	11	,	,	PUNCT
fcis-30297	98	12	striking	strike	VERB
fcis-30297	98	13	a	a	DET
fcis-30297	98	14	good	good	ADJ
fcis-30297	98	15	balance	balance	NOUN
fcis-30297	98	16	between	between	ADP
fcis-30297	98	17	training	training	NOUN
fcis-30297	98	18	speed	speed	NOUN
fcis-30297	98	19	and	and	CCONJ
fcis-30297	98	20	model	model	NOUN
fcis-30297	98	21	performance	performance	NOUN
fcis-30297	98	22	.	.	PUNCT
fcis-30297	99	1	epochs	epoch	NOUN
fcis-30297	99	2	:	:	PUNCT
fcis-30297	99	3	the	the	DET
fcis-30297	99	4	model	model	NOUN
fcis-30297	99	5	was	be	AUX
fcis-30297	99	6	trained	train	VERB
fcis-30297	99	7	for	for	ADP
fcis-30297	99	8	250	250	NUM
fcis-30297	99	9	epochs	epoch	NOUN
fcis-30297	99	10	.	.	PUNCT
fcis-30297	100	1	after	after	ADP
fcis-30297	100	2	each	each	DET
fcis-30297	100	3	epoch	epoch	NOUN
fcis-30297	100	4	,	,	PUNCT
fcis-30297	100	5	the	the	DET
fcis-30297	100	6	model	model	NOUN
fcis-30297	100	7	's	's	PART
fcis-30297	100	8	performance	performance	NOUN
fcis-30297	100	9	was	be	AUX
fcis-30297	100	10	evaluated	evaluate	VERB
fcis-30297	100	11	on	on	ADP
fcis-30297	100	12	the	the	DET
fcis-30297	100	13	validation	validation	NOUN
fcis-30297	100	14	set	set	NOUN
fcis-30297	100	15	,	,	PUNCT
fcis-30297	100	16	and	and	CCONJ
fcis-30297	100	17	the	the	DET
fcis-30297	100	18	model	model	NOUN
fcis-30297	100	19	with	with	ADP
fcis-30297	100	20	the	the	DET
fcis-30297	100	21	highest	high	ADJ
fcis-30297	100	22	validation	validation	NOUN
fcis-30297	100	23	accuracy	accuracy	NOUN
fcis-30297	100	24	was	be	AUX
fcis-30297	100	25	saved	save	VERB
fcis-30297	100	26	as	as	ADP
fcis-30297	100	27	the	the	DET
fcis-30297	100	28	final	final	ADJ
fcis-30297	100	29	model	model	NOUN
fcis-30297	100	30	.	.	PUNCT
fcis-30297	101	1	weight	weight	NOUN
fcis-30297	101	2	decay	decay	NOUN
fcis-30297	101	3	:	:	PUNCT
fcis-30297	101	4	to	to	PART
fcis-30297	101	5	prevent	prevent	VERB
fcis-30297	101	6	overfitting	overfitting	NOUN
fcis-30297	101	7	,	,	PUNCT
fcis-30297	101	8	a	a	DET
fcis-30297	101	9	weight	weight	NOUN
fcis-30297	101	10	decay	decay	NOUN
fcis-30297	101	11	strategy	strategy	NOUN
fcis-30297	101	12	with	with	ADP
fcis-30297	101	13	a	a	DET
fcis-30297	101	14	decay	decay	NOUN
fcis-30297	101	15	rate	rate	NOUN
fcis-30297	101	16	of	of	ADP
fcis-30297	101	17	0.0001	0.0001	NUM
fcis-30297	101	18	was	be	AUX
fcis-30297	101	19	employed	employ	VERB
fcis-30297	101	20	.	.	PUNCT
fcis-30297	102	1	gradient	gradient	ADJ
fcis-30297	102	2	clipping	clipping	NOUN
fcis-30297	102	3	:	:	PUNCT
fcis-30297	102	4	to	to	PART
fcis-30297	102	5	avoid	avoid	VERB
fcis-30297	102	6	the	the	DET
fcis-30297	102	7	gradient	gradient	ADJ
fcis-30297	102	8	explosion	explosion	NOUN
fcis-30297	102	9	problem	problem	NOUN
fcis-30297	102	10	,	,	PUNCT
fcis-30297	102	11	a	a	DET
fcis-30297	102	12	gradient	gradient	NOUN
fcis-30297	102	13	clipping	clip	VERB
fcis-30297	102	14	strategy	strategy	NOUN
fcis-30297	102	15	was	be	AUX
fcis-30297	102	16	used	use	VERB
fcis-30297	102	17	with	with	ADP
fcis-30297	102	18	a	a	DET
fcis-30297	102	19	threshold	threshold	NOUN
fcis-30297	102	20	set	set	VERB
fcis-30297	102	21	at	at	ADP
fcis-30297	102	22	1.0	1.0	NUM
fcis-30297	102	23	.	.	PUNCT
fcis-30297	103	1	additionally	additionally	ADV
fcis-30297	103	2	,	,	PUNCT
fcis-30297	103	3	we	we	PRON
fcis-30297	103	4	applied	apply	VERB
fcis-30297	103	5	an	an	DET
fcis-30297	103	6	early	early	ADJ
fcis-30297	103	7	stopping	stopping	NOUN
fcis-30297	103	8	strategy	strategy	NOUN
fcis-30297	103	9	during	during	ADP
fcis-30297	103	10	training	training	NOUN
fcis-30297	103	11	.	.	PUNCT
fcis-30297	104	1	if	if	SCONJ
fcis-30297	104	2	the	the	DET
fcis-30297	104	3	validation	validation	NOUN
fcis-30297	104	4	loss	loss	NOUN
fcis-30297	104	5	did	do	AUX
fcis-30297	104	6	not	not	PART
fcis-30297	104	7	decrease	decrease	VERB
fcis-30297	104	8	for	for	ADP
fcis-30297	104	9	10	10	NUM
fcis-30297	104	10	consecutive	consecutive	ADJ
fcis-30297	104	11	epochs	epoch	NOUN
fcis-30297	104	12	,	,	PUNCT
fcis-30297	104	13	the	the	DET
fcis-30297	104	14	training	training	NOUN
fcis-30297	104	15	process	process	NOUN
fcis-30297	104	16	was	be	AUX
fcis-30297	104	17	terminated	terminate	VERB
fcis-30297	104	18	early[12	early[12	NOUN
fcis-30297	104	19	]	]	X
fcis-30297	104	20	.	.	PUNCT
fcis-30297	105	1	3.4	3.4	NUM
fcis-30297	105	2	.	.	PUNCT
fcis-30297	105	3	experimental	experimental	ADJ
fcis-30297	105	4	setup	setup	NOUN
fcis-30297	105	5	to	to	PART
fcis-30297	105	6	validate	validate	VERB
fcis-30297	105	7	the	the	DET
fcis-30297	105	8	effectiveness	effectiveness	NOUN
fcis-30297	105	9	of	of	ADP
fcis-30297	105	10	the	the	DET
fcis-30297	105	11	proposed	propose	VERB
fcis-30297	105	12	improved	improve	VERB
fcis-30297	105	13	rtdetr	rtdetr	PROPN
fcis-30297	105	14	model	model	NOUN
fcis-30297	105	15	,	,	PUNCT
fcis-30297	105	16	a	a	DET
fcis-30297	105	17	series	series	NOUN
fcis-30297	105	18	of	of	ADP
fcis-30297	105	19	experiments	experiment	NOUN
fcis-30297	105	20	were	be	AUX
fcis-30297	105	21	conducted	conduct	VERB
fcis-30297	105	22	on	on	ADP
fcis-30297	105	23	multiple	multiple	ADJ
fcis-30297	105	24	public	public	ADJ
fcis-30297	105	25	wideband	wideband	NOUN
fcis-30297	105	26	signal	signal	NOUN
fcis-30297	105	27	datasets	dataset	NOUN
fcis-30297	105	28	.	.	PUNCT
fcis-30297	106	1	these	these	DET
fcis-30297	106	2	datasets	dataset	NOUN
fcis-30297	106	3	cover	cover	VERB
fcis-30297	106	4	a	a	DET
fcis-30297	106	5	variety	variety	NOUN
fcis-30297	106	6	of	of	ADP
fcis-30297	106	7	modulation	modulation	NOUN
fcis-30297	106	8	types	type	NOUN
fcis-30297	106	9	and	and	CCONJ
fcis-30297	106	10	signal	signal	NOUN
fcis-30297	106	11	-	-	PUNCT
fcis-30297	106	12	to	to	ADP
fcis-30297	106	13	-	-	PUNCT
fcis-30297	106	14	noise	noise	NOUN
fcis-30297	106	15	ratio	ratio	NOUN
fcis-30297	106	16	(	(	PUNCT
fcis-30297	106	17	snr	snr	NOUN
fcis-30297	106	18	)	)	PUNCT
fcis-30297	106	19	conditions	condition	NOUN
fcis-30297	106	20	to	to	PART
fcis-30297	106	21	ensure	ensure	VERB
fcis-30297	106	22	the	the	DET
fcis-30297	106	23	robustness	robustness	NOUN
fcis-30297	106	24	and	and	CCONJ
fcis-30297	106	25	generalization	generalization	NOUN
fcis-30297	106	26	of	of	ADP
fcis-30297	106	27	the	the	DET
fcis-30297	106	28	model	model	NOUN
fcis-30297	106	29	.	.	PUNCT
fcis-30297	107	1	4	4	NUM
fcis-30297	107	2	we	we	PRON
fcis-30297	107	3	compared	compare	VERB
fcis-30297	107	4	the	the	DET
fcis-30297	107	5	performance	performance	NOUN
fcis-30297	107	6	of	of	ADP
fcis-30297	107	7	models	model	NOUN
fcis-30297	107	8	equipped	equip	VERB
fcis-30297	107	9	with	with	ADP
fcis-30297	107	10	different	different	ADJ
fcis-30297	107	11	backbone	backbone	NOUN
fcis-30297	107	12	networks	network	NOUN
fcis-30297	107	13	,	,	PUNCT
fcis-30297	107	14	including	include	VERB
fcis-30297	107	15	resnet-18	resnet-18	PROPN
fcis-30297	107	16	and	and	CCONJ
fcis-30297	107	17	the	the	DET
fcis-30297	107	18	proposed	propose	VERB
fcis-30297	107	19	efficientformerv2	efficientformerv2	NOUN
fcis-30297	107	20	.	.	PUNCT
fcis-30297	108	1	additionally	additionally	ADV
fcis-30297	108	2	,	,	PUNCT
fcis-30297	108	3	we	we	PRON
fcis-30297	108	4	evaluated	evaluate	VERB
fcis-30297	108	5	the	the	DET
fcis-30297	108	6	impact	impact	NOUN
fcis-30297	108	7	of	of	ADP
fcis-30297	108	8	replacing	replace	VERB
fcis-30297	108	9	the	the	DET
fcis-30297	108	10	original	original	ADJ
fcis-30297	108	11	aifi	aifi	NOUN
fcis-30297	108	12	-	-	PUNCT
fcis-30297	108	13	dattention	dattention	NOUN
fcis-30297	108	14	module	module	NOUN
fcis-30297	108	15	with	with	ADP
fcis-30297	108	16	the	the	DET
fcis-30297	108	17	dynamic	dynamic	ADJ
fcis-30297	108	18	-	-	PUNCT
fcis-30297	108	19	range	range	NOUN
fcis-30297	108	20	histogram	histogram	NOUN
fcis-30297	108	21	self	self	NOUN
fcis-30297	108	22	-	-	PUNCT
fcis-30297	108	23	attention	attention	NOUN
fcis-30297	108	24	(	(	PUNCT
fcis-30297	108	25	dhsattention	dhsattention	NOUN
fcis-30297	108	26	)	)	PUNCT
fcis-30297	108	27	from	from	ADP
fcis-30297	108	28	histoformer	histoformer	PROPN
fcis-30297	108	29	.	.	PUNCT
fcis-30297	108	30	model	model	NOUN
fcis-30297	108	31	performance	performance	NOUN
fcis-30297	108	32	was	be	AUX
fcis-30297	108	33	evaluated	evaluate	VERB
fcis-30297	108	34	using	use	VERB
fcis-30297	108	35	standard	standard	ADJ
fcis-30297	108	36	metrics	metric	NOUN
fcis-30297	108	37	such	such	ADJ
fcis-30297	108	38	as	as	ADP
fcis-30297	108	39	accuracy	accuracy	NOUN
fcis-30297	108	40	,	,	PUNCT
fcis-30297	108	41	precision	precision	NOUN
fcis-30297	108	42	,	,	PUNCT
fcis-30297	108	43	and	and	CCONJ
fcis-30297	108	44	recall	recall	NOUN
fcis-30297	108	45	.	.	PUNCT
fcis-30297	109	1	these	these	DET
fcis-30297	109	2	experiments	experiment	NOUN
fcis-30297	109	3	aim	aim	VERB
fcis-30297	109	4	to	to	PART
fcis-30297	109	5	demonstrate	demonstrate	VERB
fcis-30297	109	6	the	the	DET
fcis-30297	109	7	benefits	benefit	NOUN
fcis-30297	109	8	of	of	ADP
fcis-30297	109	9	the	the	DET
fcis-30297	109	10	proposed	propose	VERB
fcis-30297	109	11	architectural	architectural	ADJ
fcis-30297	109	12	improvements	improvement	NOUN
fcis-30297	109	13	in	in	ADP
fcis-30297	109	14	both	both	DET
fcis-30297	109	15	recognition	recognition	NOUN
fcis-30297	109	16	accuracy	accuracy	NOUN
fcis-30297	109	17	and	and	CCONJ
fcis-30297	109	18	efficiency	efficiency	NOUN
fcis-30297	109	19	under	under	ADP
fcis-30297	109	20	various	various	ADJ
fcis-30297	109	21	snr	snr	NOUN
fcis-30297	109	22	conditions	condition	NOUN
fcis-30297	109	23	.	.	PUNCT
fcis-30297	110	1	3.5	3.5	NUM
fcis-30297	110	2	.	.	PUNCT
fcis-30297	110	3	dataset	dataset	NOUN
fcis-30297	110	4	generation	generation	NOUN
fcis-30297	110	5	the	the	DET
fcis-30297	110	6	dataset	dataset	NOUN
fcis-30297	110	7	for	for	ADP
fcis-30297	110	8	this	this	DET
fcis-30297	110	9	study	study	NOUN
fcis-30297	110	10	was	be	AUX
fcis-30297	110	11	generated	generate	VERB
fcis-30297	110	12	through	through	ADP
fcis-30297	110	13	matlab	matlab	PROPN
fcis-30297	110	14	simulations	simulation	NOUN
fcis-30297	110	15	,	,	PUNCT
fcis-30297	110	16	with	with	ADP
fcis-30297	110	17	the	the	DET
fcis-30297	110	18	signal	signal	NOUN
fcis-30297	110	19	-	-	PUNCT
fcis-30297	110	20	to	to	ADP
fcis-30297	110	21	-	-	PUNCT
fcis-30297	110	22	noise	noise	NOUN
fcis-30297	110	23	ratio	ratio	NOUN
fcis-30297	110	24	(	(	PUNCT
fcis-30297	110	25	snr	snr	PROPN
fcis-30297	110	26	)	)	PUNCT
fcis-30297	110	27	for	for	ADP
fcis-30297	110	28	all	all	DET
fcis-30297	110	29	signals	signal	NOUN
fcis-30297	110	30	set	set	VERB
fcis-30297	110	31	at	at	ADP
fcis-30297	110	32	30	30	NUM
fcis-30297	110	33	db	db	NOUN
fcis-30297	110	34	.	.	PUNCT
fcis-30297	111	1	this	this	DET
fcis-30297	111	2	snr	snr	NOUN
fcis-30297	111	3	level	level	NOUN
fcis-30297	111	4	represents	represent	VERB
fcis-30297	111	5	a	a	DET
fcis-30297	111	6	low	low	ADJ
fcis-30297	111	7	-	-	PUNCT
fcis-30297	111	8	noise	noise	NOUN
fcis-30297	111	9	environment	environment	NOUN
fcis-30297	111	10	,	,	PUNCT
fcis-30297	111	11	suitable	suitable	ADJ
fcis-30297	111	12	for	for	ADP
fcis-30297	111	13	testing	test	VERB
fcis-30297	111	14	the	the	DET
fcis-30297	111	15	model	model	NOUN
fcis-30297	111	16	's	's	PART
fcis-30297	111	17	recognition	recognition	NOUN
fcis-30297	111	18	ability	ability	NOUN
fcis-30297	111	19	under	under	ADP
fcis-30297	111	20	ideal	ideal	ADJ
fcis-30297	111	21	conditions	condition	NOUN
fcis-30297	111	22	.	.	PUNCT
fcis-30297	112	1	the	the	DET
fcis-30297	112	2	dataset	dataset	NOUN
fcis-30297	112	3	generation	generation	NOUN
fcis-30297	112	4	process	process	NOUN
fcis-30297	112	5	is	be	AUX
fcis-30297	112	6	as	as	SCONJ
fcis-30297	112	7	follows	follow	VERB
fcis-30297	112	8	:	:	PUNCT
fcis-30297	112	9	modulation	modulation	NOUN
fcis-30297	112	10	types	type	NOUN
fcis-30297	112	11	:	:	PUNCT
fcis-30297	112	12	the	the	DET
fcis-30297	112	13	dataset	dataset	NOUN
fcis-30297	112	14	includes	include	VERB
fcis-30297	112	15	multiple	multiple	ADJ
fcis-30297	112	16	common	common	ADJ
fcis-30297	112	17	digital	digital	ADJ
fcis-30297	112	18	modulation	modulation	NOUN
fcis-30297	112	19	types	type	NOUN
fcis-30297	112	20	,	,	PUNCT
fcis-30297	112	21	including	include	VERB
fcis-30297	112	22	bpsk	bpsk	NOUN
fcis-30297	112	23	,	,	PUNCT
fcis-30297	112	24	qpsk	qpsk	NOUN
fcis-30297	112	25	,	,	PUNCT
fcis-30297	112	26	8psk	8psk	NUM
fcis-30297	112	27	,	,	PUNCT
fcis-30297	112	28	16qam	16qam	NOUN
fcis-30297	112	29	,	,	PUNCT
fcis-30297	112	30	and	and	CCONJ
fcis-30297	112	31	64qam	64qam	NOUN
fcis-30297	112	32	.	.	PUNCT
fcis-30297	113	1	signal	signal	PROPN
fcis-30297	113	2	length	length	NOUN
fcis-30297	113	3	and	and	CCONJ
fcis-30297	113	4	sampling	sample	VERB
fcis-30297	113	5	rate	rate	NOUN
fcis-30297	113	6	:	:	PUNCT
fcis-30297	113	7	each	each	DET
fcis-30297	113	8	signal	signal	NOUN
fcis-30297	113	9	length	length	NOUN
fcis-30297	113	10	is	be	AUX
fcis-30297	113	11	set	set	VERB
fcis-30297	113	12	to	to	ADP
fcis-30297	113	13	1024	1024	NUM
fcis-30297	113	14	sample	sample	NOUN
fcis-30297	113	15	points	point	NOUN
fcis-30297	113	16	,	,	PUNCT
fcis-30297	113	17	with	with	ADP
fcis-30297	113	18	a	a	DET
fcis-30297	113	19	sampling	sample	VERB
fcis-30297	113	20	rate	rate	NOUN
fcis-30297	113	21	of	of	ADP
fcis-30297	113	22	1	1	NUM
fcis-30297	113	23	mhz	mhz	NOUN
fcis-30297	113	24	.	.	PUNCT
fcis-30297	114	1	the	the	DET
fcis-30297	114	2	generated	generate	VERB
fcis-30297	114	3	signals	signal	NOUN
fcis-30297	114	4	are	be	AUX
fcis-30297	114	5	stored	store	VERB
fcis-30297	114	6	in	in	ADP
fcis-30297	114	7	complex	complex	ADJ
fcis-30297	114	8	i	i	PROPN
fcis-30297	114	9	/	/	SYM
fcis-30297	114	10	q	q	NOUN
fcis-30297	114	11	data	data	NOUN
fcis-30297	114	12	format	format	NOUN
fcis-30297	114	13	for	for	ADP
fcis-30297	114	14	subsequent	subsequent	ADJ
fcis-30297	114	15	processing	processing	NOUN
fcis-30297	114	16	and	and	CCONJ
fcis-30297	114	17	feature	feature	NOUN
fcis-30297	114	18	extraction	extraction	NOUN
fcis-30297	114	19	.	.	PUNCT
fcis-30297	115	1	data	datum	NOUN
fcis-30297	115	2	augmentation	augmentation	NOUN
fcis-30297	115	3	:	:	PUNCT
fcis-30297	115	4	although	although	SCONJ
fcis-30297	115	5	all	all	DET
fcis-30297	115	6	signals	signal	NOUN
fcis-30297	115	7	have	have	VERB
fcis-30297	115	8	an	an	DET
fcis-30297	115	9	snr	snr	NOUN
fcis-30297	115	10	set	set	VERB
fcis-30297	115	11	at	at	ADP
fcis-30297	115	12	30	30	NUM
fcis-30297	115	13	db	db	NOUN
fcis-30297	115	14	,	,	PUNCT
fcis-30297	115	15	other	other	ADJ
fcis-30297	115	16	simulation	simulation	NOUN
fcis-30297	115	17	conditions	condition	NOUN
fcis-30297	115	18	such	such	ADJ
fcis-30297	115	19	as	as	ADP
fcis-30297	115	20	frequency	frequency	NOUN
fcis-30297	115	21	offset	offset	NOUN
fcis-30297	115	22	and	and	CCONJ
fcis-30297	115	23	phase	phase	NOUN
fcis-30297	115	24	shift	shift	NOUN
fcis-30297	115	25	were	be	AUX
fcis-30297	115	26	introduced	introduce	VERB
fcis-30297	115	27	through	through	ADP
fcis-30297	115	28	data	datum	NOUN
fcis-30297	115	29	augmentation	augmentation	NOUN
fcis-30297	115	30	,	,	PUNCT
fcis-30297	115	31	increasing	increase	VERB
fcis-30297	115	32	the	the	DET
fcis-30297	115	33	diversity	diversity	NOUN
fcis-30297	115	34	and	and	CCONJ
fcis-30297	115	35	complexity	complexity	NOUN
fcis-30297	115	36	of	of	ADP
fcis-30297	115	37	the	the	DET
fcis-30297	115	38	data	datum	NOUN
fcis-30297	115	39	.	.	PUNCT
fcis-30297	116	1	dataset	dataset	ADJ
fcis-30297	116	2	splitting	splitting	NOUN
fcis-30297	116	3	:	:	PUNCT
fcis-30297	116	4	the	the	DET
fcis-30297	116	5	generated	generate	VERB
fcis-30297	116	6	signal	signal	NOUN
fcis-30297	116	7	data	datum	NOUN
fcis-30297	116	8	is	be	AUX
fcis-30297	116	9	split	split	VERB
fcis-30297	116	10	into	into	ADP
fcis-30297	116	11	training	training	NOUN
fcis-30297	116	12	,	,	PUNCT
fcis-30297	116	13	validation	validation	NOUN
fcis-30297	116	14	,	,	PUNCT
fcis-30297	116	15	and	and	CCONJ
fcis-30297	116	16	test	test	NOUN
fcis-30297	116	17	sets	set	NOUN
fcis-30297	116	18	in	in	ADP
fcis-30297	116	19	a	a	DET
fcis-30297	116	20	70	70	NUM
fcis-30297	116	21	%	%	NOUN
fcis-30297	116	22	,	,	PUNCT
fcis-30297	116	23	15	15	NUM
fcis-30297	116	24	%	%	NOUN
fcis-30297	116	25	,	,	PUNCT
fcis-30297	116	26	and	and	CCONJ
fcis-30297	116	27	15	15	NUM
fcis-30297	116	28	%	%	NOUN
fcis-30297	116	29	ratio	ratio	NOUN
fcis-30297	116	30	,	,	PUNCT
fcis-30297	116	31	ensuring	ensure	VERB
fcis-30297	116	32	an	an	DET
fcis-30297	116	33	even	even	ADJ
fcis-30297	116	34	distribution	distribution	NOUN
fcis-30297	116	35	of	of	ADP
fcis-30297	116	36	modulation	modulation	NOUN
fcis-30297	116	37	types	type	NOUN
fcis-30297	116	38	across	across	ADP
fcis-30297	116	39	different	different	ADJ
fcis-30297	116	40	datasets	dataset	NOUN
fcis-30297	116	41	.	.	PUNCT
fcis-30297	117	1	figure	figure	NOUN
fcis-30297	117	2	2	2	NUM
fcis-30297	117	3	.	.	NOUN
fcis-30297	117	4	example	example	NOUN
fcis-30297	117	5	of	of	ADP
fcis-30297	117	6	time	time	NOUN
fcis-30297	117	7	-	-	PUNCT
fcis-30297	117	8	frequency	frequency	NOUN
fcis-30297	117	9	diagram	diagram	NOUN
fcis-30297	117	10	of	of	ADP
fcis-30297	117	11	a	a	DET
fcis-30297	117	12	broadband	broadband	NOUN
fcis-30297	117	13	signal	signal	NOUN
fcis-30297	117	14	4	4	NUM
fcis-30297	117	15	.	.	PUNCT
fcis-30297	117	16	experimental	experimental	ADJ
fcis-30297	117	17	process	process	NOUN
fcis-30297	117	18	and	and	CCONJ
fcis-30297	117	19	results	result	VERB
fcis-30297	117	20	analysis	analysis	NOUN
fcis-30297	117	21	4.1	4.1	NUM
fcis-30297	117	22	.	.	PUNCT
fcis-30297	118	1	training	training	NOUN
fcis-30297	118	2	process	process	NOUN
fcis-30297	118	3	in	in	ADP
fcis-30297	118	4	the	the	DET
fcis-30297	118	5	training	training	NOUN
fcis-30297	118	6	process	process	NOUN
fcis-30297	118	7	,	,	PUNCT
fcis-30297	118	8	the	the	DET
fcis-30297	118	9	data	data	NOUN
fcis-30297	118	10	features	feature	NOUN
fcis-30297	118	11	of	of	ADP
fcis-30297	118	12	communication	communication	NOUN
fcis-30297	118	13	signals	signal	NOUN
fcis-30297	118	14	are	be	AUX
fcis-30297	118	15	first	first	ADV
fcis-30297	118	16	extracted	extract	VERB
fcis-30297	118	17	using	use	VERB
fcis-30297	118	18	convolution	convolution	NOUN
fcis-30297	118	19	windows	window	NOUN
fcis-30297	118	20	in	in	ADP
fcis-30297	118	21	the	the	DET
fcis-30297	118	22	convolutional	convolutional	ADJ
fcis-30297	118	23	neural	neural	ADJ
fcis-30297	118	24	network	network	NOUN
fcis-30297	118	25	(	(	PUNCT
fcis-30297	118	26	cnn	cnn	PROPN
fcis-30297	118	27	)	)	PUNCT
fcis-30297	118	28	.	.	PUNCT
fcis-30297	119	1	the	the	DET
fcis-30297	119	2	categorical	categorical	ADJ
fcis-30297	119	3	crossentropy	crossentropy	NOUN
fcis-30297	119	4	loss	loss	NOUN
fcis-30297	119	5	function	function	NOUN
fcis-30297	119	6	is	be	AUX
fcis-30297	119	7	employed	employ	VERB
fcis-30297	119	8	,	,	PUNCT
fcis-30297	119	9	and	and	CCONJ
fcis-30297	119	10	the	the	DET
fcis-30297	119	11	loss	loss	NOUN
fcis-30297	119	12	function	function	NOUN
fcis-30297	119	13	is	be	AUX
fcis-30297	119	14	defined	define	VERB
fcis-30297	119	15	as	as	SCONJ
fcis-30297	119	16	follows	follow	VERB
fcis-30297	119	17	∑	∑	PUNCT
fcis-30297	119	18	,	,	PUNCT
fcis-30297	119	19	log	log	VERB
fcis-30297	119	20	,	,	PUNCT
fcis-30297	119	21	(	(	PUNCT
fcis-30297	119	22	1	1	X
fcis-30297	119	23	)	)	PUNCT
fcis-30297	119	24	where	where	SCONJ
fcis-30297	119	25	t	t	PROPN
fcis-30297	119	26	represents	represent	VERB
fcis-30297	119	27	the	the	DET
fcis-30297	119	28	true	true	ADJ
fcis-30297	119	29	labels	label	NOUN
fcis-30297	119	30	,	,	PUNCT
fcis-30297	119	31	i	i	PRON
fcis-30297	119	32	represents	represent	VERB
fcis-30297	119	33	the	the	DET
fcis-30297	119	34	input	input	NOUN
fcis-30297	119	35	data	datum	NOUN
fcis-30297	119	36	,	,	PUNCT
fcis-30297	119	37	j	j	PROPN
fcis-30297	119	38	represents	represent	VERB
fcis-30297	119	39	the	the	DET
fcis-30297	119	40	categories	category	NOUN
fcis-30297	119	41	,	,	PUNCT
fcis-30297	119	42	and	and	CCONJ
fcis-30297	119	43	p	p	NOUN
fcis-30297	119	44	represents	represent	VERB
fcis-30297	119	45	the	the	DET
fcis-30297	119	46	predicted	predict	VERB
fcis-30297	119	47	results	result	NOUN
fcis-30297	119	48	.	.	PUNCT
fcis-30297	120	1	this	this	DET
fcis-30297	120	2	loss	loss	NOUN
fcis-30297	120	3	function	function	NOUN
fcis-30297	120	4	is	be	AUX
fcis-30297	120	5	commonly	commonly	ADV
fcis-30297	120	6	used	use	VERB
fcis-30297	120	7	in	in	ADP
fcis-30297	120	8	multi	multi	ADJ
fcis-30297	120	9	-	-	ADJ
fcis-30297	120	10	class	class	ADJ
fcis-30297	120	11	classification	classification	NOUN
fcis-30297	120	12	tasks	task	NOUN
fcis-30297	120	13	,	,	PUNCT
fcis-30297	120	14	such	such	ADJ
fcis-30297	120	15	as	as	ADP
fcis-30297	120	16	when	when	SCONJ
fcis-30297	120	17	using	use	VERB
fcis-30297	120	18	the	the	DET
fcis-30297	120	19	softmax	softmax	NOUN
fcis-30297	120	20	function	function	NOUN
fcis-30297	120	21	as	as	ADP
fcis-30297	120	22	the	the	DET
fcis-30297	120	23	final	final	ADJ
fcis-30297	120	24	output	output	NOUN
fcis-30297	120	25	.	.	PUNCT
fcis-30297	121	1	the	the	DET
fcis-30297	121	2	optimizer	optimizer	NOUN
fcis-30297	121	3	continuously	continuously	ADV
fcis-30297	121	4	reduces	reduce	VERB
fcis-30297	121	5	the	the	DET
fcis-30297	121	6	loss	loss	NOUN
fcis-30297	121	7	function	function	NOUN
fcis-30297	121	8	to	to	PART
fcis-30297	121	9	update	update	VERB
fcis-30297	121	10	the	the	DET
fcis-30297	121	11	parameters	parameter	NOUN
fcis-30297	121	12	of	of	ADP
fcis-30297	121	13	the	the	DET
fcis-30297	121	14	hidden	hidden	ADJ
fcis-30297	121	15	layers	layer	NOUN
fcis-30297	121	16	.	.	PUNCT
fcis-30297	122	1	in	in	ADP
fcis-30297	122	2	this	this	DET
fcis-30297	122	3	case	case	NOUN
fcis-30297	122	4	,	,	PUNCT
fcis-30297	122	5	we	we	PRON
fcis-30297	122	6	selected	select	VERB
fcis-30297	122	7	the	the	DET
fcis-30297	122	8	adam	adam	PROPN
fcis-30297	122	9	optimizer	optimizer	NOUN
fcis-30297	122	10	,	,	PUNCT
fcis-30297	122	11	which	which	PRON
fcis-30297	122	12	is	be	AUX
fcis-30297	122	13	currently	currently	ADV
fcis-30297	122	14	one	one	NUM
fcis-30297	122	15	of	of	ADP
fcis-30297	122	16	the	the	DET
fcis-30297	122	17	most	most	ADV
fcis-30297	122	18	widely	widely	ADV
fcis-30297	122	19	used	use	VERB
fcis-30297	122	20	optimizers[7	optimizers[7	ADP
fcis-30297	122	21	]	]	PUNCT
fcis-30297	122	22	.	.	PUNCT
fcis-30297	123	1	according	accord	VERB
fcis-30297	123	2	to	to	ADP
fcis-30297	123	3	the	the	DET
fcis-30297	123	4	literature[16	literature[16	PROPN
fcis-30297	123	5	]	]	PUNCT
fcis-30297	123	6	,	,	PUNCT
fcis-30297	123	7	this	this	DET
fcis-30297	123	8	optimizer	optimizer	NOUN
fcis-30297	123	9	offers	offer	VERB
fcis-30297	123	10	advantages	advantage	NOUN
fcis-30297	123	11	such	such	ADJ
fcis-30297	123	12	as	as	ADP
fcis-30297	123	13	low	low	ADJ
fcis-30297	123	14	memory	memory	NOUN
fcis-30297	123	15	requirements	requirement	NOUN
fcis-30297	123	16	,	,	PUNCT
fcis-30297	123	17	simple	simple	ADJ
fcis-30297	123	18	implementation	implementation	NOUN
fcis-30297	123	19	,	,	PUNCT
fcis-30297	123	20	and	and	CCONJ
fcis-30297	123	21	high	high	ADJ
fcis-30297	123	22	computational	computational	ADJ
fcis-30297	123	23	efficiency	efficiency	NOUN
fcis-30297	123	24	.	.	PUNCT
fcis-30297	124	1	the	the	DET
fcis-30297	124	2	learning	learning	NOUN
fcis-30297	124	3	rate	rate	NOUN
fcis-30297	124	4	for	for	ADP
fcis-30297	124	5	the	the	DET
fcis-30297	124	6	optimizer	optimizer	NOUN
fcis-30297	124	7	was	be	AUX
fcis-30297	124	8	set	set	VERB
fcis-30297	124	9	to	to	ADP
fcis-30297	124	10	a	a	DET
fcis-30297	124	11	fixed	fix	VERB
fcis-30297	124	12	value	value	NOUN
fcis-30297	124	13	of	of	ADP
fcis-30297	124	14	0.001	0.001	NUM
fcis-30297	124	15	,	,	PUNCT
fcis-30297	124	16	and	and	CCONJ
fcis-30297	124	17	a	a	DET
fcis-30297	124	18	dropout	dropout	NOUN
fcis-30297	124	19	rate	rate	NOUN
fcis-30297	124	20	of	of	ADP
fcis-30297	124	21	0.5	0.5	NUM
fcis-30297	124	22	was	be	AUX
fcis-30297	124	23	used	use	VERB
fcis-30297	124	24	to	to	PART
fcis-30297	124	25	randomly	randomly	VERB
fcis-30297	124	26	deactivate	deactivate	VERB
fcis-30297	124	27	neurons	neuron	NOUN
fcis-30297	124	28	,	,	PUNCT
fcis-30297	124	29	preventing	prevent	VERB
fcis-30297	124	30	overfitting	overfitte	VERB
fcis-30297	124	31	due	due	ADP
fcis-30297	124	32	to	to	ADP
fcis-30297	124	33	the	the	DET
fcis-30297	124	34	large	large	ADJ
fcis-30297	124	35	number	number	NOUN
fcis-30297	124	36	of	of	ADP
fcis-30297	124	37	parameters	parameter	NOUN
fcis-30297	124	38	in	in	ADP
fcis-30297	124	39	the	the	DET
fcis-30297	124	40	fully	fully	ADV
fcis-30297	124	41	connected	connect	VERB
fcis-30297	124	42	layer	layer	NOUN
fcis-30297	124	43	.	.	PUNCT
fcis-30297	125	1	the	the	DET
fcis-30297	125	2	confusion	confusion	NOUN
fcis-30297	125	3	matrix	matrix	NOUN
fcis-30297	125	4	in	in	ADP
fcis-30297	125	5	figure	figure	NOUN
fcis-30297	125	6	3	3	NUM
fcis-30297	125	7	shows	show	VERB
fcis-30297	125	8	the	the	DET
fcis-30297	125	9	classification	classification	NOUN
fcis-30297	125	10	performance	performance	NOUN
fcis-30297	125	11	at	at	ADP
fcis-30297	125	12	a	a	DET
fcis-30297	125	13	signal	signal	NOUN
fcis-30297	125	14	-	-	PUNCT
fcis-30297	125	15	to	to	ADP
fcis-30297	125	16	-	-	PUNCT
fcis-30297	125	17	noise	noise	NOUN
fcis-30297	125	18	ratio	ratio	NOUN
fcis-30297	125	19	(	(	PUNCT
fcis-30297	125	20	snr	snr	PROPN
fcis-30297	125	21	)	)	PUNCT
fcis-30297	125	22	of	of	ADP
fcis-30297	125	23	30db	30db	NOUN
fcis-30297	125	24	.	.	PUNCT
fcis-30297	126	1	the	the	DET
fcis-30297	126	2	horizontal	horizontal	ADJ
fcis-30297	126	3	axis	axis	NOUN
fcis-30297	126	4	represents	represent	VERB
fcis-30297	126	5	the	the	DET
fcis-30297	126	6	nine	nine	NUM
fcis-30297	126	7	predicted	predict	VERB
fcis-30297	126	8	modulation	modulation	NOUN
fcis-30297	126	9	categories	category	NOUN
fcis-30297	126	10	,	,	PUNCT
fcis-30297	126	11	while	while	SCONJ
fcis-30297	126	12	the	the	DET
fcis-30297	126	13	vertical	vertical	ADJ
fcis-30297	126	14	axis	axis	NOUN
fcis-30297	126	15	represents	represent	VERB
fcis-30297	126	16	the	the	DET
fcis-30297	126	17	actual	actual	ADJ
fcis-30297	126	18	modulation	modulation	NOUN
fcis-30297	126	19	categories	category	NOUN
fcis-30297	126	20	.	.	PUNCT
fcis-30297	127	1	this	this	DET
fcis-30297	127	2	confusion	confusion	NOUN
fcis-30297	127	3	matrix	matrix	NOUN
fcis-30297	127	4	provides	provide	VERB
fcis-30297	127	5	a	a	DET
fcis-30297	127	6	visual	visual	ADJ
fcis-30297	127	7	representation	representation	NOUN
fcis-30297	127	8	of	of	ADP
fcis-30297	127	9	how	how	SCONJ
fcis-30297	127	10	well	well	ADV
fcis-30297	127	11	the	the	DET
fcis-30297	127	12	model	model	NOUN
fcis-30297	127	13	correctly	correctly	ADV
fcis-30297	127	14	classifies	classify	VERB
fcis-30297	127	15	each	each	DET
fcis-30297	127	16	modulation	modulation	NOUN
fcis-30297	127	17	type	type	VERB
fcis-30297	127	18	under	under	ADP
fcis-30297	127	19	the	the	DET
fcis-30297	127	20	given	give	VERB
fcis-30297	127	21	conditions[8	conditions[8	NOUN
fcis-30297	127	22	]	]	X
fcis-30297	127	23	.	.	PUNCT
fcis-30297	128	1	this	this	DET
fcis-30297	128	2	process	process	NOUN
fcis-30297	128	3	,	,	PUNCT
fcis-30297	128	4	combining	combine	VERB
fcis-30297	128	5	the	the	DET
fcis-30297	128	6	loss	loss	NOUN
fcis-30297	128	7	function	function	NOUN
fcis-30297	128	8	,	,	PUNCT
fcis-30297	128	9	adam	adam	PROPN
fcis-30297	128	10	optimizer	optimizer	NOUN
fcis-30297	128	11	,	,	PUNCT
fcis-30297	128	12	and	and	CCONJ
fcis-30297	128	13	dropout	dropout	NOUN
fcis-30297	128	14	techniques	technique	NOUN
fcis-30297	128	15	,	,	PUNCT
fcis-30297	128	16	helped	help	VERB
fcis-30297	128	17	achieve	achieve	VERB
fcis-30297	128	18	better	well	ADJ
fcis-30297	128	19	generalization	generalization	NOUN
fcis-30297	128	20	and	and	CCONJ
fcis-30297	128	21	avoided	avoid	VERB
fcis-30297	128	22	overfitting	overfitting	NOUN
fcis-30297	128	23	during	during	ADP
fcis-30297	128	24	training	training	NOUN
fcis-30297	128	25	.	.	PUNCT
fcis-30297	129	1	figure	figure	NOUN
fcis-30297	129	2	3	3	NUM
fcis-30297	129	3	.	.	PUNCT
fcis-30297	130	1	the	the	DET
fcis-30297	130	2	result	result	NOUN
fcis-30297	130	3	of	of	ADP
fcis-30297	130	4	a	a	DET
fcis-30297	130	5	signal	signal	NOUN
fcis-30297	130	6	-	-	PUNCT
fcis-30297	130	7	to	to	ADP
fcis-30297	130	8	-	-	PUNCT
fcis-30297	130	9	noise	noise	NOUN
fcis-30297	130	10	ratio	ratio	NOUN
fcis-30297	130	11	(	(	PUNCT
fcis-30297	130	12	snr	snr	PROPN
fcis-30297	130	13	)	)	PUNCT
fcis-30297	130	14	of	of	ADP
fcis-30297	130	15	30	30	NUM
fcis-30297	130	16	db	db	NOUN
fcis-30297	130	17	.	.	PROPN
fcis-30297	130	18	4.2	4.2	NUM
fcis-30297	130	19	.	.	PUNCT
fcis-30297	131	1	results	result	VERB
fcis-30297	131	2	analysis	analysis	NOUN
fcis-30297	131	3	the	the	DET
fcis-30297	131	4	neural	neural	ADJ
fcis-30297	131	5	network	network	NOUN
fcis-30297	131	6	was	be	AUX
fcis-30297	131	7	trained	train	VERB
fcis-30297	131	8	for	for	ADP
fcis-30297	131	9	250	250	NUM
fcis-30297	131	10	epochs	epoch	NOUN
fcis-30297	131	11	using	use	VERB
fcis-30297	131	12	the	the	DET
fcis-30297	131	13	training	training	NOUN
fcis-30297	131	14	dataset	dataset	NOUN
fcis-30297	131	15	,	,	PUNCT
fcis-30297	131	16	while	while	SCONJ
fcis-30297	131	17	the	the	DET
fcis-30297	131	18	test	test	NOUN
fcis-30297	131	19	dataset	dataset	NOUN
fcis-30297	131	20	was	be	AUX
fcis-30297	131	21	used	use	VERB
fcis-30297	131	22	to	to	PART
fcis-30297	131	23	evaluate	evaluate	VERB
fcis-30297	131	24	the	the	DET
fcis-30297	131	25	model	model	NOUN
fcis-30297	131	26	’s	’s	PART
fcis-30297	131	27	performance	performance	NOUN
fcis-30297	131	28	after	after	ADP
fcis-30297	131	29	each	each	DET
fcis-30297	131	30	epoch	epoch	NOUN
fcis-30297	131	31	.	.	PUNCT
fcis-30297	132	1	as	as	SCONJ
fcis-30297	132	2	shown	show	VERB
fcis-30297	132	3	in	in	ADP
fcis-30297	132	4	figure	figure	NOUN
fcis-30297	132	5	4	4	NUM
fcis-30297	132	6	,	,	PUNCT
fcis-30297	132	7	the	the	DET
fcis-30297	132	8	loss	loss	NOUN
fcis-30297	132	9	steadily	steadily	ADV
fcis-30297	132	10	decreases	decrease	VERB
fcis-30297	132	11	throughout	throughout	ADP
fcis-30297	132	12	the	the	DET
fcis-30297	132	13	training	training	NOUN
fcis-30297	132	14	process	process	NOUN
fcis-30297	132	15	,	,	PUNCT
fcis-30297	132	16	reaching	reach	VERB
fcis-30297	132	17	a	a	DET
fcis-30297	132	18	minimum	minimum	ADJ
fcis-30297	132	19	value	value	NOUN
fcis-30297	132	20	of	of	ADP
fcis-30297	132	21	approximately	approximately	ADV
fcis-30297	132	22	0.2	0.2	NUM
fcis-30297	132	23	.	.	PUNCT
fcis-30297	133	1	figure	figure	VERB
fcis-30297	133	2	5	5	NUM
fcis-30297	133	3	presents	present	VERB
fcis-30297	133	4	the	the	DET
fcis-30297	133	5	training	training	NOUN
fcis-30297	133	6	and	and	CCONJ
fcis-30297	133	7	test	test	NOUN
fcis-30297	133	8	accuracy	accuracy	NOUN
fcis-30297	133	9	curves	curve	NOUN
fcis-30297	133	10	.	.	PUNCT
fcis-30297	134	1	the	the	DET
fcis-30297	134	2	training	training	NOUN
fcis-30297	134	3	accuracy	accuracy	NOUN
fcis-30297	134	4	peaks	peak	NOUN
fcis-30297	134	5	at	at	ADP
fcis-30297	134	6	71.7	71.7	NUM
fcis-30297	134	7	%	%	NOUN
fcis-30297	134	8	,	,	PUNCT
fcis-30297	134	9	while	while	SCONJ
fcis-30297	134	10	the	the	DET
fcis-30297	134	11	test	test	NOUN
fcis-30297	134	12	accuracy	accuracy	NOUN
fcis-30297	134	13	reaches	reach	VERB
fcis-30297	134	14	a	a	DET
fcis-30297	134	15	maximum	maximum	NOUN
fcis-30297	134	16	of	of	ADP
fcis-30297	134	17	72	72	NUM
fcis-30297	134	18	%	%	NOUN
fcis-30297	134	19	.	.	PUNCT
fcis-30297	135	1	both	both	DET
fcis-30297	135	2	metrics	metric	NOUN
fcis-30297	135	3	increase	increase	VERB
fcis-30297	135	4	in	in	ADP
fcis-30297	135	5	tandem	tandem	NOUN
fcis-30297	135	6	during	during	ADP
fcis-30297	135	7	training	training	NOUN
fcis-30297	135	8	and	and	CCONJ
fcis-30297	135	9	gradually	gradually	ADV
fcis-30297	135	10	stabilize	stabilize	VERB
fcis-30297	135	11	,	,	PUNCT
fcis-30297	135	12	with	with	ADP
fcis-30297	135	13	no	no	DET
fcis-30297	135	14	indication	indication	NOUN
fcis-30297	135	15	of	of	ADP
fcis-30297	135	16	severe	severe	ADJ
fcis-30297	135	17	divergence	divergence	NOUN
fcis-30297	135	18	between	between	ADP
fcis-30297	135	19	the	the	DET
fcis-30297	135	20	two	two	NUM
fcis-30297	135	21	curves	curve	NOUN
fcis-30297	135	22	.	.	PUNCT
fcis-30297	136	1	these	these	DET
fcis-30297	136	2	results	result	NOUN
fcis-30297	136	3	suggest	suggest	VERB
fcis-30297	136	4	that	that	SCONJ
fcis-30297	136	5	the	the	DET
fcis-30297	136	6	model	model	NOUN
fcis-30297	136	7	does	do	AUX
fcis-30297	136	8	not	not	PART
fcis-30297	136	9	suffer	suffer	VERB
fcis-30297	136	10	from	from	ADP
fcis-30297	136	11	overfitting	overfitte	VERB
fcis-30297	136	12	or	or	CCONJ
fcis-30297	136	13	underfitting	underfitting	NOUN
fcis-30297	136	14	.	.	PUNCT
fcis-30297	137	1	the	the	DET
fcis-30297	137	2	neural	neural	ADJ
fcis-30297	137	3	network	network	NOUN
fcis-30297	137	4	effectively	effectively	ADV
fcis-30297	137	5	5	5	NUM
fcis-30297	137	6	learned	learn	VERB
fcis-30297	137	7	representative	representative	ADJ
fcis-30297	137	8	features	feature	NOUN
fcis-30297	137	9	from	from	ADP
fcis-30297	137	10	the	the	DET
fcis-30297	137	11	training	training	NOUN
fcis-30297	137	12	data	datum	NOUN
fcis-30297	137	13	,	,	PUNCT
fcis-30297	137	14	and	and	CCONJ
fcis-30297	137	15	the	the	DET
fcis-30297	137	16	consistent	consistent	ADJ
fcis-30297	137	17	performance	performance	NOUN
fcis-30297	137	18	on	on	ADP
fcis-30297	137	19	the	the	DET
fcis-30297	137	20	test	test	NOUN
fcis-30297	137	21	set	set	NOUN
fcis-30297	137	22	demonstrates	demonstrate	VERB
fcis-30297	137	23	strong	strong	ADJ
fcis-30297	137	24	generalization	generalization	NOUN
fcis-30297	137	25	capabilities	capability	NOUN
fcis-30297	137	26	of	of	ADP
fcis-30297	137	27	the	the	DET
fcis-30297	137	28	trained	train	VERB
fcis-30297	137	29	model	model	NOUN
fcis-30297	137	30	.	.	PUNCT
fcis-30297	138	1	figure	figure	VERB
fcis-30297	138	2	4	4	NUM
fcis-30297	138	3	.	.	PUNCT
fcis-30297	139	1	the	the	DET
fcis-30297	139	2	variation	variation	NOUN
fcis-30297	139	3	curve	curve	NOUN
fcis-30297	139	4	of	of	ADP
fcis-30297	139	5	the	the	DET
fcis-30297	139	6	classification	classification	NOUN
fcis-30297	139	7	loss	loss	NOUN
fcis-30297	139	8	figure	figure	NOUN
fcis-30297	139	9	5	5	NUM
fcis-30297	139	10	.	.	PUNCT
fcis-30297	140	1	the	the	DET
fcis-30297	140	2	accuracy	accuracy	NOUN
fcis-30297	140	3	of	of	ADP
fcis-30297	140	4	the	the	DET
fcis-30297	140	5	model	model	NOUN
fcis-30297	140	6	during	during	ADP
fcis-30297	140	7	training	training	NOUN
fcis-30297	140	8	.	.	PUNCT
fcis-30297	141	1	5	5	X
fcis-30297	141	2	.	.	X
fcis-30297	141	3	conclusion	conclusion	NOUN
fcis-30297	141	4	this	this	PRON
fcis-30297	141	5	this	this	DET
fcis-30297	141	6	paper	paper	NOUN
fcis-30297	141	7	conducted	conduct	VERB
fcis-30297	141	8	extensive	extensive	ADJ
fcis-30297	141	9	experiments	experiment	NOUN
fcis-30297	141	10	to	to	PART
fcis-30297	141	11	explore	explore	VERB
fcis-30297	141	12	various	various	ADJ
fcis-30297	141	13	hyperparameters	hyperparameter	NOUN
fcis-30297	141	14	and	and	CCONJ
fcis-30297	141	15	iteratively	iteratively	ADV
fcis-30297	141	16	optimize	optimize	VERB
fcis-30297	141	17	the	the	DET
fcis-30297	141	18	network	network	NOUN
fcis-30297	141	19	architecture	architecture	NOUN
fcis-30297	141	20	,	,	PUNCT
fcis-30297	141	21	ultimately	ultimately	ADV
fcis-30297	141	22	resulting	result	VERB
fcis-30297	141	23	in	in	ADP
fcis-30297	141	24	the	the	DET
fcis-30297	141	25	redesign	redesign	NOUN
fcis-30297	141	26	of	of	ADP
fcis-30297	141	27	a	a	DET
fcis-30297	141	28	new	new	ADJ
fcis-30297	141	29	deep	deep	ADJ
fcis-30297	141	30	neural	neural	ADJ
fcis-30297	141	31	network	network	NOUN
fcis-30297	141	32	model	model	NOUN
fcis-30297	141	33	.	.	PUNCT
fcis-30297	142	1	experimental	experimental	ADJ
fcis-30297	142	2	results	result	NOUN
fcis-30297	142	3	validate	validate	VERB
fcis-30297	142	4	the	the	DET
fcis-30297	142	5	effectiveness	effectiveness	NOUN
fcis-30297	142	6	of	of	ADP
fcis-30297	142	7	the	the	DET
fcis-30297	142	8	proposed	propose	VERB
fcis-30297	142	9	approach	approach	NOUN
fcis-30297	142	10	,	,	PUNCT
fcis-30297	142	11	showing	show	VERB
fcis-30297	142	12	significant	significant	ADJ
fcis-30297	142	13	improvements	improvement	NOUN
fcis-30297	142	14	in	in	ADP
fcis-30297	142	15	the	the	DET
fcis-30297	142	16	recognition	recognition	NOUN
fcis-30297	142	17	of	of	ADP
fcis-30297	142	18	various	various	ADJ
fcis-30297	142	19	modulation	modulation	NOUN
fcis-30297	142	20	types	type	NOUN
fcis-30297	142	21	compared	compare	VERB
fcis-30297	142	22	to	to	ADP
fcis-30297	142	23	traditional	traditional	ADJ
fcis-30297	142	24	methods	method	NOUN
fcis-30297	142	25	.	.	PUNCT
fcis-30297	143	1	most	most	ADV
fcis-30297	143	2	notably	notably	ADV
fcis-30297	143	3	,	,	PUNCT
fcis-30297	143	4	this	this	DET
fcis-30297	143	5	method	method	NOUN
fcis-30297	143	6	achieves	achieve	VERB
fcis-30297	143	7	end	end	NOUN
fcis-30297	143	8	-	-	PUNCT
fcis-30297	143	9	to	to	ADP
fcis-30297	143	10	-	-	PUNCT
fcis-30297	143	11	end	end	NOUN
fcis-30297	143	12	signal	signal	NOUN
fcis-30297	143	13	recognition	recognition	NOUN
fcis-30297	143	14	,	,	PUNCT
fcis-30297	143	15	eliminating	eliminate	VERB
fcis-30297	143	16	the	the	DET
fcis-30297	143	17	need	need	NOUN
fcis-30297	143	18	for	for	ADP
fcis-30297	143	19	manual	manual	ADJ
fcis-30297	143	20	feature	feature	NOUN
fcis-30297	143	21	extraction	extraction	NOUN
fcis-30297	143	22	and	and	CCONJ
fcis-30297	143	23	simplifying	simplify	VERB
fcis-30297	143	24	the	the	DET
fcis-30297	143	25	processing	processing	NOUN
fcis-30297	143	26	pipeline	pipeline	NOUN
fcis-30297	143	27	.	.	PUNCT
fcis-30297	144	1	the	the	DET
fcis-30297	144	2	model	model	NOUN
fcis-30297	144	3	achieves	achieve	VERB
fcis-30297	144	4	high	high	ADJ
fcis-30297	144	5	test	test	NOUN
fcis-30297	144	6	accuracy	accuracy	NOUN
fcis-30297	144	7	and	and	CCONJ
fcis-30297	144	8	demonstrates	demonstrate	VERB
fcis-30297	144	9	strong	strong	ADJ
fcis-30297	144	10	generalization	generalization	NOUN
fcis-30297	144	11	capability	capability	NOUN
fcis-30297	144	12	,	,	PUNCT
fcis-30297	144	13	indicating	indicate	VERB
fcis-30297	144	14	its	its	PRON
fcis-30297	144	15	potential	potential	NOUN
fcis-30297	144	16	for	for	ADP
fcis-30297	144	17	practical	practical	ADJ
fcis-30297	144	18	application	application	NOUN
fcis-30297	144	19	.	.	PUNCT
fcis-30297	145	1	with	with	ADP
fcis-30297	145	2	ongoing	ongoing	ADJ
fcis-30297	145	3	research	research	NOUN
fcis-30297	145	4	efforts	effort	NOUN
fcis-30297	145	5	and	and	CCONJ
fcis-30297	145	6	the	the	DET
fcis-30297	145	7	continuous	continuous	ADJ
fcis-30297	145	8	advancement	advancement	NOUN
fcis-30297	145	9	of	of	ADP
fcis-30297	145	10	deep	deep	ADJ
fcis-30297	145	11	learning	learning	NOUN
fcis-30297	145	12	,	,	PUNCT
fcis-30297	145	13	it	it	PRON
fcis-30297	145	14	is	be	AUX
fcis-30297	145	15	expected	expect	VERB
fcis-30297	145	16	that	that	SCONJ
fcis-30297	145	17	further	further	ADJ
fcis-30297	145	18	breakthroughs	breakthrough	NOUN
fcis-30297	145	19	will	will	AUX
fcis-30297	145	20	emerge	emerge	VERB
fcis-30297	145	21	in	in	ADP
fcis-30297	145	22	modulation	modulation	NOUN
fcis-30297	145	23	recognition	recognition	NOUN
fcis-30297	145	24	for	for	ADP
fcis-30297	145	25	signal	signal	NOUN
fcis-30297	145	26	and	and	CCONJ
fcis-30297	145	27	information	information	NOUN
fcis-30297	145	28	processing	processing	NOUN
fcis-30297	145	29	,	,	PUNCT
fcis-30297	145	30	leading	lead	VERB
fcis-30297	145	31	to	to	ADP
fcis-30297	145	32	even	even	ADV
fcis-30297	145	33	greater	great	ADJ
fcis-30297	145	34	accuracy	accuracy	NOUN
fcis-30297	145	35	.	.	PUNCT
fcis-30297	146	1	nevertheless	nevertheless	ADV
fcis-30297	146	2	,	,	PUNCT
fcis-30297	146	3	there	there	PRON
fcis-30297	146	4	remain	remain	VERB
fcis-30297	146	5	areas	area	NOUN
fcis-30297	146	6	for	for	ADP
fcis-30297	146	7	improvement	improvement	NOUN
fcis-30297	146	8	.	.	PUNCT
fcis-30297	147	1	the	the	DET
fcis-30297	147	2	current	current	ADJ
fcis-30297	147	3	experiments	experiment	NOUN
fcis-30297	147	4	were	be	AUX
fcis-30297	147	5	limited	limit	VERB
fcis-30297	147	6	by	by	ADP
fcis-30297	147	7	the	the	DET
fcis-30297	147	8	scale	scale	NOUN
fcis-30297	147	9	of	of	ADP
fcis-30297	147	10	available	available	ADJ
fcis-30297	147	11	data	datum	NOUN
fcis-30297	147	12	,	,	PUNCT
fcis-30297	147	13	and	and	CCONJ
fcis-30297	147	14	further	further	ADJ
fcis-30297	147	15	optimization	optimization	NOUN
fcis-30297	147	16	of	of	ADP
fcis-30297	147	17	the	the	DET
fcis-30297	147	18	network	network	NOUN
fcis-30297	147	19	architecture	architecture	NOUN
fcis-30297	147	20	and	and	CCONJ
fcis-30297	147	21	hyperparameters	hyperparameter	NOUN
fcis-30297	147	22	is	be	AUX
fcis-30297	147	23	still	still	ADV
fcis-30297	147	24	possible	possible	ADJ
fcis-30297	147	25	.	.	PUNCT
fcis-30297	148	1	future	future	ADJ
fcis-30297	148	2	work	work	NOUN
fcis-30297	148	3	will	will	AUX
fcis-30297	148	4	focus	focus	VERB
fcis-30297	148	5	on	on	ADP
fcis-30297	148	6	expanding	expand	VERB
fcis-30297	148	7	the	the	DET
fcis-30297	148	8	dataset	dataset	NOUN
fcis-30297	148	9	,	,	PUNCT
fcis-30297	148	10	enhancing	enhance	VERB
fcis-30297	148	11	model	model	NOUN
fcis-30297	148	12	robustness	robustness	NOUN
fcis-30297	148	13	,	,	PUNCT
fcis-30297	148	14	and	and	CCONJ
fcis-30297	148	15	refining	refine	VERB
fcis-30297	148	16	key	key	ADJ
fcis-30297	148	17	components	component	NOUN
fcis-30297	148	18	of	of	ADP
fcis-30297	148	19	the	the	DET
fcis-30297	148	20	network	network	NOUN
fcis-30297	148	21	through	through	ADP
fcis-30297	148	22	continued	continue	VERB
fcis-30297	148	23	experimentation	experimentation	NOUN
fcis-30297	148	24	.	.	PUNCT
fcis-30297	149	1	references	reference	NOUN
fcis-30297	149	2	[	[	X
fcis-30297	149	3	1	1	X
fcis-30297	149	4	]	]	X
fcis-30297	149	5	zell	zell	PROPN
fcis-30297	149	6	o	o	PROPN
fcis-30297	149	7	,	,	PUNCT
fcis-30297	149	8	pålsson	pålsson	PROPN
fcis-30297	149	9	j	j	PROPN
fcis-30297	149	10	,	,	PUNCT
fcis-30297	149	11	hernandez	hernandez	PROPN
fcis-30297	149	12	-	-	PUNCT
fcis-30297	149	13	diaz	diaz	PROPN
fcis-30297	150	1	k	k	PROPN
fcis-30297	150	2	,	,	PUNCT
fcis-30297	150	3	et	et	PROPN
fcis-30297	150	4	al	al	PROPN
fcis-30297	150	5	.	.	PUNCT
fcis-30297	150	6	image	image	NOUN
fcis-30297	150	7	-	-	PUNCT
fcis-30297	150	8	based	base	VERB
fcis-30297	150	9	fire	fire	NOUN
fcis-30297	150	10	detection	detection	NOUN
fcis-30297	150	11	in	in	ADP
fcis-30297	150	12	industrial	industrial	ADJ
fcis-30297	150	13	environments	environment	NOUN
fcis-30297	150	14	with	with	ADP
fcis-30297	150	15	yolov4[j	yolov4[j	PROPN
fcis-30297	150	16	]	]	PUNCT
fcis-30297	150	17	.	.	PUNCT
fcis-30297	151	1	arxiv	arxiv	PROPN
fcis-30297	151	2	preprint	preprint	VERB
fcis-30297	151	3	arxiv:2212.04786	arxiv:2212.04786	NOUN
fcis-30297	151	4	,	,	PUNCT
fcis-30297	151	5	2022	2022	NUM
fcis-30297	151	6	.	.	PUNCT
fcis-30297	152	1	[	[	X
fcis-30297	152	2	2	2	NUM
fcis-30297	152	3	]	]	X
fcis-30297	152	4	dobre	dobre	X
fcis-30297	152	5	,	,	PUNCT
fcis-30297	152	6	o.	o.	PROPN
fcis-30297	152	7	a.	a.	PROPN
fcis-30297	152	8	,	,	PUNCT
fcis-30297	152	9	et	et	PROPN
fcis-30297	152	10	al	al	PROPN
fcis-30297	152	11	.	.	PUNCT
fcis-30297	153	1	(	(	PUNCT
fcis-30297	153	2	2007	2007	NUM
fcis-30297	153	3	)	)	PUNCT
fcis-30297	153	4	.	.	PUNCT
fcis-30297	154	1	survey	survey	NOUN
fcis-30297	154	2	of	of	ADP
fcis-30297	154	3	automatic	automatic	ADJ
fcis-30297	154	4	modulation	modulation	NOUN
fcis-30297	154	5	classification	classification	NOUN
fcis-30297	154	6	techniques	technique	NOUN
fcis-30297	154	7	:	:	PUNCT
fcis-30297	154	8	classical	classical	ADJ
fcis-30297	154	9	approaches	approach	NOUN
fcis-30297	154	10	and	and	CCONJ
fcis-30297	154	11	new	new	ADJ
fcis-30297	154	12	trends	trend	NOUN
fcis-30297	154	13	.	.	PUNCT
fcis-30297	155	1	ieee	ieee	NOUN
fcis-30297	155	2	communications	communication	NOUN
fcis-30297	155	3	surveys	survey	NOUN
fcis-30297	155	4	&	&	CCONJ
fcis-30297	155	5	tutorials	tutorial	NOUN
fcis-30297	155	6	.	.	PUNCT
fcis-30297	156	1	[	[	X
fcis-30297	156	2	3	3	NUM
fcis-30297	156	3	]	]	X
fcis-30297	156	4	o'shea	o'shea	PROPN
fcis-30297	156	5	,	,	PUNCT
fcis-30297	156	6	t.	t.	PROPN
fcis-30297	156	7	j.	j.	PROPN
fcis-30297	156	8	,	,	PUNCT
fcis-30297	156	9	&	&	CCONJ
fcis-30297	156	10	hoydis	hoydis	PROPN
fcis-30297	156	11	,	,	PUNCT
fcis-30297	156	12	j.	j.	PROPN
fcis-30297	156	13	(	(	PUNCT
fcis-30297	156	14	2017	2017	NUM
fcis-30297	156	15	)	)	PUNCT
fcis-30297	156	16	.	.	PUNCT
fcis-30297	157	1	an	an	DET
fcis-30297	157	2	introduction	introduction	NOUN
fcis-30297	157	3	to	to	ADP
fcis-30297	157	4	deep	deep	ADJ
fcis-30297	157	5	learning	learning	NOUN
fcis-30297	157	6	for	for	ADP
fcis-30297	157	7	the	the	DET
fcis-30297	157	8	physical	physical	ADJ
fcis-30297	157	9	layer	layer	NOUN
fcis-30297	157	10	.	.	PUNCT
fcis-30297	158	1	ieee	ieee	NOUN
fcis-30297	158	2	transactions	transaction	NOUN
fcis-30297	158	3	on	on	ADP
fcis-30297	158	4	cognitive	cognitive	ADJ
fcis-30297	158	5	communications	communication	NOUN
fcis-30297	158	6	and	and	CCONJ
fcis-30297	158	7	networking	networking	NOUN
fcis-30297	158	8	.	.	PUNCT
fcis-30297	159	1	[	[	X
fcis-30297	159	2	4	4	NUM
fcis-30297	159	3	]	]	X
fcis-30297	159	4	kulin	kulin	NOUN
fcis-30297	159	5	,	,	PUNCT
fcis-30297	159	6	m.	m.	NOUN
fcis-30297	159	7	,	,	PUNCT
fcis-30297	159	8	et	et	PROPN
fcis-30297	159	9	al	al	PROPN
fcis-30297	159	10	.	.	PUNCT
fcis-30297	160	1	(	(	PUNCT
fcis-30297	160	2	2018	2018	NUM
fcis-30297	160	3	)	)	PUNCT
fcis-30297	160	4	.	.	PUNCT
fcis-30297	161	1	end	end	NOUN
fcis-30297	161	2	-	-	PUNCT
fcis-30297	161	3	to	to	ADP
fcis-30297	161	4	-	-	PUNCT
fcis-30297	161	5	end	end	NOUN
fcis-30297	161	6	learning	learning	NOUN
fcis-30297	161	7	from	from	ADP
fcis-30297	161	8	spectrum	spectrum	NOUN
fcis-30297	161	9	data	datum	NOUN
fcis-30297	161	10	ieee	ieee	NOUN
fcis-30297	161	11	access	access	NOUN
fcis-30297	161	12	.	.	PUNCT
fcis-30297	162	1	[	[	X
fcis-30297	162	2	5	5	NUM
fcis-30297	162	3	]	]	X
fcis-30297	162	4	zhang	zhang	PROPN
fcis-30297	162	5	,	,	PUNCT
fcis-30297	162	6	f.	f.	PROPN
fcis-30297	162	7	,	,	PUNCT
fcis-30297	162	8	et	et	PROPN
fcis-30297	162	9	al	al	PROPN
fcis-30297	162	10	.	.	PROPN
fcis-30297	162	11	(	(	PUNCT
fcis-30297	162	12	2022	2022	NUM
fcis-30297	162	13	)	)	PUNCT
fcis-30297	162	14	.	.	PUNCT
fcis-30297	163	1	deep	deep	ADJ
fcis-30297	163	2	learning	learning	NOUN
fcis-30297	163	3	based	base	VERB
fcis-30297	163	4	automatic	automatic	ADJ
fcis-30297	163	5	modulation	modulation	NOUN
fcis-30297	163	6	recognition	recognition	NOUN
fcis-30297	163	7	:	:	PUNCT
fcis-30297	163	8	models	model	NOUN
fcis-30297	163	9	,	,	PUNCT
fcis-30297	163	10	datasets	dataset	NOUN
fcis-30297	163	11	,	,	PUNCT
fcis-30297	163	12	and	and	CCONJ
fcis-30297	163	13	challenges	challenge	NOUN
fcis-30297	163	14	.	.	PUNCT
fcis-30297	164	1	[	[	X
fcis-30297	164	2	6	6	NUM
fcis-30297	164	3	]	]	SYM
fcis-30297	164	4	li	li	PROPN
fcis-30297	164	5	,	,	PUNCT
fcis-30297	164	6	y.	y.	PROPN
fcis-30297	164	7	,	,	PUNCT
fcis-30297	164	8	et	et	PROPN
fcis-30297	164	9	al	al	PROPN
fcis-30297	164	10	.	.	PROPN
fcis-30297	164	11	(	(	PUNCT
fcis-30297	164	12	2022	2022	NUM
fcis-30297	164	13	)	)	PUNCT
fcis-30297	164	14	.	.	PUNCT
fcis-30297	165	1	efficientformerv2	efficientformerv2	PROPN
fcis-30297	165	2	:	:	PUNCT
fcis-30297	165	3	efficient	efficient	ADJ
fcis-30297	165	4	and	and	CCONJ
fcis-30297	165	5	accurate	accurate	ADJ
fcis-30297	165	6	lightweight	lightweight	ADJ
fcis-30297	165	7	vision	vision	NOUN
fcis-30297	165	8	transformers	transformer	NOUN
fcis-30297	165	9	.	.	PUNCT
fcis-30297	166	1	[	[	X
fcis-30297	166	2	7	7	NUM
fcis-30297	166	3	]	]	X
fcis-30297	166	4	zhang	zhang	PROPN
fcis-30297	166	5	,	,	PUNCT
fcis-30297	166	6	y.	y.	PROPN
fcis-30297	166	7	,	,	PUNCT
fcis-30297	166	8	et	et	PROPN
fcis-30297	166	9	al	al	PROPN
fcis-30297	166	10	.	.	PUNCT
fcis-30297	166	11	(	(	PUNCT
fcis-30297	166	12	2024	2024	NUM
fcis-30297	166	13	)	)	PUNCT
fcis-30297	166	14	.	.	PUNCT
fcis-30297	166	15	histoformer	histoformer	NOUN
fcis-30297	166	16	:	:	PUNCT
fcis-30297	166	17	histogram	histogram	NOUN
fcis-30297	166	18	transformer	transformer	NOUN
fcis-30297	166	19	for	for	ADP
fcis-30297	166	20	lightweight	lightweight	ADJ
fcis-30297	166	21	image	image	NOUN
fcis-30297	166	22	recognition	recognition	NOUN
fcis-30297	166	23	.	.	PUNCT
fcis-30297	167	1	eccv	eccv	ADJ
fcis-30297	167	2	2024	2024	NUM
fcis-30297	167	3	.	.	PUNCT
fcis-30297	168	1	[	[	X
fcis-30297	168	2	8	8	NUM
fcis-30297	168	3	]	]	X
fcis-30297	168	4	kingma	kingma	PROPN
fcis-30297	168	5	,	,	PUNCT
fcis-30297	168	6	d.	d.	PROPN
fcis-30297	168	7	p.	p.	PROPN
fcis-30297	168	8	,	,	PUNCT
fcis-30297	168	9	&	&	CCONJ
fcis-30297	168	10	ba	ba	PROPN
fcis-30297	168	11	,	,	PUNCT
fcis-30297	168	12	j.	j.	PROPN
fcis-30297	168	13	(	(	PUNCT
fcis-30297	168	14	2014	2014	NUM
fcis-30297	168	15	)	)	PUNCT
fcis-30297	168	16	.	.	PUNCT
fcis-30297	169	1	adam	adam	PROPN
fcis-30297	169	2	:	:	PUNCT
fcis-30297	169	3	a	a	DET
fcis-30297	169	4	method	method	NOUN
fcis-30297	169	5	for	for	ADP
fcis-30297	169	6	stochastic	stochastic	ADJ
fcis-30297	169	7	optimization	optimization	NOUN
fcis-30297	169	8	.	.	PUNCT
fcis-30297	170	1	[	[	X
fcis-30297	170	2	9	9	NUM
fcis-30297	170	3	]	]	SYM
fcis-30297	170	4	amini	amini	PROPN
fcis-30297	170	5	,	,	PUNCT
fcis-30297	170	6	a.	a.	NOUN
fcis-30297	170	7	,	,	PUNCT
fcis-30297	170	8	shirani	shirani	NOUN
fcis-30297	170	9	-	-	PUNCT
fcis-30297	170	10	mehr	mehr	NOUN
fcis-30297	170	11	,	,	PUNCT
fcis-30297	170	12	h.	h.	PROPN
fcis-30297	170	13	,	,	PUNCT
fcis-30297	170	14	&	&	CCONJ
fcis-30297	170	15	karbasi	karbasi	NOUN
fcis-30297	170	16	,	,	PUNCT
fcis-30297	170	17	a.	a.	PROPN
fcis-30297	170	18	digital	digital	PROPN
fcis-30297	170	19	modulation	modulation	NOUN
fcis-30297	170	20	classification	classification	NOUN
fcis-30297	170	21	:	:	PUNCT
fcis-30297	170	22	a	a	DET
fcis-30297	170	23	deep	deep	ADJ
fcis-30297	170	24	learning	learning	NOUN
fcis-30297	170	25	approach	approach	NOUN
fcis-30297	170	26	.	.	PUNCT
fcis-30297	171	1	ieee	ieee	NOUN
fcis-30297	171	2	transactions	transaction	NOUN
fcis-30297	171	3	on	on	ADP
fcis-30297	171	4	communications	communication	NOUN
fcis-30297	171	5	,	,	PUNCT
fcis-30297	171	6	(	(	PUNCT
fcis-30297	171	7	2017	2017	NUM
fcis-30297	171	8	)	)	PUNCT
fcis-30297	171	9	65(11	65(11	NUM
fcis-30297	171	10	)	)	PUNCT
fcis-30297	171	11	,	,	PUNCT
fcis-30297	171	12	4658	4658	NUM
fcis-30297	171	13	-	-	SYM
fcis-30297	171	14	4668	4668	NUM
fcis-30297	171	15	.	.	PUNCT
fcis-30297	172	1	[	[	X
fcis-30297	172	2	10	10	NUM
fcis-30297	172	3	]	]	X
fcis-30297	172	4	hochreiter	hochreiter	PROPN
fcis-30297	172	5	,	,	PUNCT
fcis-30297	172	6	s.	s.	PROPN
fcis-30297	172	7	,	,	PUNCT
fcis-30297	172	8	&	&	CCONJ
fcis-30297	172	9	schmidhuber	schmidhuber	PROPN
fcis-30297	172	10	,	,	PUNCT
fcis-30297	172	11	j.	j.	PROPN
fcis-30297	172	12	long	long	PROPN
fcis-30297	172	13	short	short	ADJ
fcis-30297	172	14	-	-	PUNCT
fcis-30297	172	15	term	term	NOUN
fcis-30297	172	16	memory	memory	NOUN
fcis-30297	172	17	.	.	PUNCT
fcis-30297	173	1	neural	neural	ADJ
fcis-30297	173	2	computation	computation	NOUN
fcis-30297	173	3	,	,	PUNCT
fcis-30297	173	4	(	(	PUNCT
fcis-30297	173	5	1997	1997	NUM
fcis-30297	173	6	)	)	PUNCT
fcis-30297	173	7	9(8	9(8	NUM
fcis-30297	173	8	)	)	PUNCT
fcis-30297	173	9	,	,	PUNCT
fcis-30297	173	10	1735	1735	NUM
fcis-30297	173	11	-	-	SYM
fcis-30297	173	12	1780	1780	NUM
fcis-30297	173	13	.	.	PUNCT
fcis-30297	174	1	[	[	X
fcis-30297	174	2	11	11	NUM
fcis-30297	174	3	]	]	SYM
fcis-30297	174	4	lecun	lecun	ADJ
fcis-30297	174	5	,	,	PUNCT
fcis-30297	174	6	y.	y.	PROPN
fcis-30297	174	7	,	,	PUNCT
fcis-30297	174	8	bottou	bottou	PROPN
fcis-30297	174	9	,	,	PUNCT
fcis-30297	174	10	l.	l.	PROPN
fcis-30297	174	11	,	,	PUNCT
fcis-30297	174	12	orr	orr	PROPN
fcis-30297	174	13	,	,	PUNCT
fcis-30297	174	14	g.	g.	PROPN
fcis-30297	174	15	b.	b.	PROPN
fcis-30297	174	16	,	,	PUNCT
fcis-30297	174	17	&	&	CCONJ
fcis-30297	174	18	müller	müller	PROPN
fcis-30297	174	19	,	,	PUNCT
fcis-30297	174	20	k.-r	k.-r	PROPN
fcis-30297	174	21	.	.	PUNCT
fcis-30297	175	1	(	(	PUNCT
fcis-30297	175	2	2012	2012	NUM
fcis-30297	175	3	)	)	PUNCT
fcis-30297	175	4	.	.	PUNCT
fcis-30297	176	1	efficient	efficient	ADJ
fcis-30297	176	2	backprop	backprop	NOUN
fcis-30297	176	3	.	.	PUNCT
fcis-30297	177	1	in	in	ADP
fcis-30297	177	2	g.	g.	PROPN
fcis-30297	177	3	montavon	montavon	PROPN
fcis-30297	177	4	,	,	PUNCT
fcis-30297	177	5	g.	g.	PROPN
fcis-30297	177	6	b.	b.	PROPN
fcis-30297	177	7	orr	orr	PROPN
fcis-30297	177	8	,	,	PUNCT
fcis-30297	177	9	&	&	CCONJ
fcis-30297	177	10	k.-r	k.-r	PROPN
fcis-30297	177	11	.	.	PUNCT
fcis-30297	178	1	müller	müller	PROPN
fcis-30297	178	2	(	(	PUNCT
fcis-30297	178	3	eds	eds	PROPN
fcis-30297	178	4	.	.	PUNCT
fcis-30297	178	5	)	)	PUNCT
fcis-30297	178	6	,	,	PUNCT
fcis-30297	178	7	neural	neural	ADJ
fcis-30297	178	8	networks	network	NOUN
fcis-30297	178	9	:	:	PUNCT
fcis-30297	178	10	tricks	trick	NOUN
fcis-30297	178	11	of	of	ADP
fcis-30297	178	12	the	the	DET
fcis-30297	178	13	trade	trade	NOUN
fcis-30297	178	14	(	(	PUNCT
fcis-30297	178	15	pp	pp	ADJ
fcis-30297	178	16	.	.	PUNCT
fcis-30297	179	1	9	9	NUM
fcis-30297	179	2	-	-	SYM
fcis-30297	179	3	48	48	NUM
fcis-30297	179	4	)	)	PUNCT
fcis-30297	179	5	.	.	PUNCT
fcis-30297	180	1	springer	springer	NOUN
fcis-30297	180	2	.	.	PUNCT
fcis-30297	181	1	[	[	X
fcis-30297	181	2	12	12	NUM
fcis-30297	181	3	]	]	PUNCT
fcis-30297	181	4	shorten	shorten	NOUN
fcis-30297	181	5	,	,	PUNCT
fcis-30297	181	6	c.	c.	PROPN
fcis-30297	181	7	,	,	PUNCT
fcis-30297	181	8	&	&	CCONJ
fcis-30297	181	9	khoshgoftaar	khoshgoftaar	PROPN
fcis-30297	181	10	,	,	PUNCT
fcis-30297	181	11	t.	t.	PROPN
fcis-30297	181	12	m.	m.	NOUN
fcis-30297	181	13	a	a	DET
fcis-30297	181	14	survey	survey	NOUN
fcis-30297	181	15	on	on	ADP
fcis-30297	181	16	image	image	NOUN
fcis-30297	181	17	data	datum	NOUN
fcis-30297	181	18	augmentation	augmentation	NOUN
fcis-30297	181	19	for	for	ADP
fcis-30297	181	20	deep	deep	ADJ
fcis-30297	181	21	learning	learning	NOUN
fcis-30297	181	22	.	.	PUNCT
fcis-30297	182	1	journal	journal	NOUN
fcis-30297	182	2	of	of	ADP
fcis-30297	182	3	big	big	ADJ
fcis-30297	182	4	data	datum	NOUN
fcis-30297	182	5	,	,	PUNCT
fcis-30297	182	6	6(60	6(60	NUM
fcis-30297	182	7	)	)	PUNCT
fcis-30297	182	8	,	,	PUNCT
fcis-30297	182	9	(	(	PUNCT
fcis-30297	182	10	2019	2019	NUM
fcis-30297	182	11	)	)	PUNCT
fcis-30297	182	12	.	.	PUNCT
fcis-30297	183	1	1	1	NUM
fcis-30297	183	2	-	-	SYM
fcis-30297	183	3	48	48	NUM
fcis-30297	183	4	.	.	PUNCT
fcis-30297	184	1	[	[	X
fcis-30297	184	2	13	13	NUM
fcis-30297	184	3	]	]	X
fcis-30297	184	4	prechelt	prechelt	NOUN
fcis-30297	184	5	,	,	PUNCT
fcis-30297	184	6	l.	l.	PROPN
fcis-30297	184	7	(	(	PUNCT
fcis-30297	184	8	1998	1998	NUM
fcis-30297	184	9	)	)	PUNCT
fcis-30297	184	10	.	.	PUNCT
fcis-30297	185	1	early	early	ADJ
fcis-30297	185	2	stopping	stopping	NOUN
fcis-30297	185	3	but	but	CCONJ
fcis-30297	185	4	when	when	SCONJ
fcis-30297	185	5	?	?	PUNCT
fcis-30297	186	1	in	in	ADP
fcis-30297	186	2	g.	g.	PROPN
fcis-30297	186	3	b.	b.	PROPN
fcis-30297	186	4	orr	orr	PROPN
fcis-30297	186	5	&	&	CCONJ
fcis-30297	186	6	k.-r	k.-r	PROPN
fcis-30297	186	7	.	.	PUNCT
fcis-30297	187	1	müller	müller	PROPN
fcis-30297	187	2	(	(	PUNCT
fcis-30297	187	3	eds	eds	PROPN
fcis-30297	187	4	.	.	PUNCT
fcis-30297	187	5	)	)	PUNCT
fcis-30297	187	6	,	,	PUNCT
fcis-30297	187	7	neural	neural	ADJ
fcis-30297	187	8	networks	network	NOUN
fcis-30297	187	9	:	:	PUNCT
fcis-30297	187	10	tricks	trick	NOUN
fcis-30297	187	11	of	of	ADP
fcis-30297	187	12	the	the	DET
fcis-30297	187	13	trade	trade	NOUN
fcis-30297	187	14	(	(	PUNCT
fcis-30297	187	15	pp	pp	ADJ
fcis-30297	187	16	.	.	PUNCT
fcis-30297	188	1	55	55	NUM
fcis-30297	188	2	-	-	SYM
fcis-30297	188	3	69	69	NUM
fcis-30297	188	4	)	)	PUNCT
fcis-30297	188	5	.	.	PUNCT
fcis-30297	189	1	springer	springer	NOUN
fcis-30297	189	2	.	.	PUNCT
fcis-30297	190	1	[	[	X
fcis-30297	190	2	14	14	NUM
fcis-30297	190	3	]	]	X
fcis-30297	190	4	dosovitskiy	dosovitskiy	NOUN
fcis-30297	190	5	,	,	PUNCT
fcis-30297	190	6	a.	a.	PROPN
fcis-30297	190	7	,	,	PUNCT
fcis-30297	190	8	beyer	beyer	PROPN
fcis-30297	190	9	,	,	PUNCT
fcis-30297	190	10	l.	l.	PROPN
fcis-30297	190	11	,	,	PUNCT
fcis-30297	190	12	kolesnikov	kolesnikov	PROPN
fcis-30297	190	13	,	,	PUNCT
fcis-30297	190	14	a.	a.	NOUN
fcis-30297	190	15	,	,	PUNCT
fcis-30297	190	16	weissenborn	weissenborn	ADJ
fcis-30297	190	17	,	,	PUNCT
fcis-30297	190	18	d.	d.	PROPN
fcis-30297	190	19	,	,	PUNCT
fcis-30297	190	20	et	et	PROPN
fcis-30297	190	21	al	al	PROPN
fcis-30297	190	22	.	.	PROPN
fcis-30297	190	23	(	(	PUNCT
fcis-30297	190	24	2020	2020	NUM
fcis-30297	190	25	)	)	PUNCT
fcis-30297	190	26	.	.	PUNCT
fcis-30297	191	1	an	an	DET
fcis-30297	191	2	image	image	NOUN
fcis-30297	191	3	is	be	AUX
fcis-30297	191	4	worth	worth	ADJ
fcis-30297	191	5	16x16	16x16	NUM
fcis-30297	191	6	words	word	NOUN
fcis-30297	191	7	:	:	PUNCT
fcis-30297	191	8	transformers	transformer	NOUN
fcis-30297	191	9	for	for	ADP
fcis-30297	191	10	image	image	NOUN
fcis-30297	191	11	recognition	recognition	NOUN
fcis-30297	191	12	at	at	ADP
fcis-30297	191	13	scale	scale	NOUN
fcis-30297	191	14	.	.	PUNCT
fcis-30297	192	1	proceedings	proceeding	NOUN
fcis-30297	192	2	of	of	ADP
fcis-30297	192	3	the	the	DET
fcis-30297	192	4	international	international	ADJ
fcis-30297	192	5	conference	conference	NOUN
fcis-30297	192	6	on	on	ADP
fcis-30297	192	7	learning	learn	VERB
fcis-30297	192	8	representations	representation	NOUN
fcis-30297	192	9	(	(	PUNCT
fcis-30297	192	10	iclr	iclr	NOUN
fcis-30297	192	11	)	)	PUNCT
fcis-30297	192	12	.	.	PUNCT