id	sid	tid	token	lemma	pos
ajst-3607	1	1	academic	academic	ADJ
ajst-3607	1	2	journal	journal	NOUN
ajst-3607	1	3	of	of	ADP
ajst-3607	1	4	science	science	NOUN
ajst-3607	1	5	and	and	CCONJ
ajst-3607	1	6	technology	technology	NOUN
ajst-3607	1	7	issn	issn	NOUN
ajst-3607	1	8	:	:	PUNCT
ajst-3607	1	9	2771	2771	NUM
ajst-3607	1	10	-	-	SYM
ajst-3607	1	11	3032	3032	NUM
ajst-3607	1	12	|	|	NOUN
ajst-3607	1	13	vol	vol	NOUN
ajst-3607	1	14	.	.	PROPN
ajst-3607	2	1	4	4	NUM
ajst-3607	2	2	,	,	PUNCT
ajst-3607	2	3	no	no	INTJ
ajst-3607	2	4	.	.	NOUN
ajst-3607	2	5	1	1	NUM
ajst-3607	2	6	,	,	PUNCT
ajst-3607	2	7	2022	2022	NUM
ajst-3607	2	8	149	149	NUM
ajst-3607	2	9	analysis	analysis	NOUN
ajst-3607	2	10	of	of	ADP
ajst-3607	2	11	interaction	interaction	NOUN
ajst-3607	2	12	grouping	group	VERB
ajst-3607	2	13	modeling	modeling	NOUN
ajst-3607	2	14	fusion	fusion	NOUN
ajst-3607	2	15	group	group	NOUN
ajst-3607	2	16	behavior	behavior	NOUN
ajst-3607	2	17	recognition	recognition	NOUN
ajst-3607	2	18	algorithm	algorithm	PROPN
ajst-3607	2	19	bin	bin	PROPN
ajst-3607	2	20	li	li	PROPN
ajst-3607	2	21	,	,	PUNCT
ajst-3607	2	22	fan	fan	PROPN
ajst-3607	2	23	zhang	zhang	PROPN
ajst-3607	2	24	china	china	PROPN
ajst-3607	2	25	auto	auto	PROPN
ajst-3607	2	26	information	information	NOUN
ajst-3607	2	27	technology	technology	NOUN
ajst-3607	2	28	(	(	PUNCT
ajst-3607	2	29	tianjin	tianjin	PROPN
ajst-3607	2	30	)	)	PUNCT
ajst-3607	2	31	co.	co.	PROPN
ajst-3607	2	32	,	,	PUNCT
ajst-3607	2	33	ltd	ltd	PROPN
ajst-3607	2	34	,	,	PUNCT
ajst-3607	2	35	china	china	PROPN
ajst-3607	2	36	abstract	abstract	NOUN
ajst-3607	2	37	:	:	PUNCT
ajst-3607	2	38	in	in	ADP
ajst-3607	2	39	order	order	NOUN
ajst-3607	2	40	to	to	PART
ajst-3607	2	41	make	make	VERB
ajst-3607	2	42	full	full	ADJ
ajst-3607	2	43	use	use	NOUN
ajst-3607	2	44	of	of	ADP
ajst-3607	2	45	the	the	DET
ajst-3607	2	46	effective	effective	ADJ
ajst-3607	2	47	information	information	NOUN
ajst-3607	2	48	in	in	ADP
ajst-3607	2	49	the	the	DET
ajst-3607	2	50	video	video	NOUN
ajst-3607	2	51	,	,	PUNCT
ajst-3607	2	52	this	this	DET
ajst-3607	2	53	paper	paper	NOUN
ajst-3607	2	54	proposes	propose	VERB
ajst-3607	2	55	a	a	DET
ajst-3607	2	56	multi	multi	ADJ
ajst-3607	2	57	-	-	ADJ
ajst-3607	2	58	model	model	ADJ
ajst-3607	2	59	interactive	interactive	ADJ
ajst-3607	2	60	video	video	NOUN
ajst-3607	2	61	behavior	behavior	NOUN
ajst-3607	2	62	recognition	recognition	NOUN
ajst-3607	2	63	method	method	NOUN
ajst-3607	2	64	.	.	PUNCT
ajst-3607	3	1	in	in	ADP
ajst-3607	3	2	order	order	NOUN
ajst-3607	3	3	to	to	PART
ajst-3607	3	4	solve	solve	VERB
ajst-3607	3	5	the	the	DET
ajst-3607	3	6	problems	problem	NOUN
ajst-3607	3	7	of	of	ADP
ajst-3607	3	8	incomplete	incomplete	ADJ
ajst-3607	3	9	human	human	ADJ
ajst-3607	3	10	target	target	NOUN
ajst-3607	3	11	detection	detection	NOUN
ajst-3607	3	12	and	and	CCONJ
ajst-3607	3	13	redundant	redundant	ADJ
ajst-3607	3	14	feature	feature	NOUN
ajst-3607	3	15	extraction	extraction	NOUN
ajst-3607	3	16	,	,	PUNCT
ajst-3607	3	17	yolo_v4	yolo_v4	NOUN
ajst-3607	3	18	is	be	AUX
ajst-3607	3	19	used	use	VERB
ajst-3607	3	20	to	to	PART
ajst-3607	3	21	detect	detect	VERB
ajst-3607	3	22	the	the	DET
ajst-3607	3	23	human	human	ADJ
ajst-3607	3	24	body	body	NOUN
ajst-3607	3	25	and	and	CCONJ
ajst-3607	3	26	remove	remove	VERB
ajst-3607	3	27	the	the	DET
ajst-3607	3	28	redundant	redundant	ADJ
ajst-3607	3	29	background	background	NOUN
ajst-3607	3	30	information	information	NOUN
ajst-3607	3	31	.	.	PUNCT
ajst-3607	4	1	then	then	ADV
ajst-3607	4	2	,	,	PUNCT
ajst-3607	4	3	it	it	PRON
ajst-3607	4	4	is	be	AUX
ajst-3607	4	5	proposed	propose	VERB
ajst-3607	4	6	to	to	PART
ajst-3607	4	7	introduce	introduce	VERB
ajst-3607	4	8	the	the	DET
ajst-3607	4	9	channel	channel	NOUN
ajst-3607	4	10	attention	attention	NOUN
ajst-3607	4	11	model	model	NOUN
ajst-3607	4	12	se	se	PROPN
ajst-3607	4	13	-	-	NOUN
ajst-3607	4	14	net	net	NOUN
ajst-3607	4	15	into	into	ADP
ajst-3607	4	16	the	the	DET
ajst-3607	4	17	inception_v3	inception_v3	ADJ
ajst-3607	4	18	network	network	NOUN
ajst-3607	4	19	,	,	PUNCT
ajst-3607	4	20	so	so	SCONJ
ajst-3607	4	21	as	as	SCONJ
ajst-3607	4	22	to	to	PART
ajst-3607	4	23	strengthen	strengthen	VERB
ajst-3607	4	24	the	the	DET
ajst-3607	4	25	extraction	extraction	NOUN
ajst-3607	4	26	of	of	ADP
ajst-3607	4	27	key	key	ADJ
ajst-3607	4	28	features	feature	NOUN
ajst-3607	4	29	and	and	CCONJ
ajst-3607	4	30	make	make	VERB
ajst-3607	4	31	the	the	DET
ajst-3607	4	32	network	network	NOUN
ajst-3607	4	33	pay	pay	VERB
ajst-3607	4	34	more	more	ADJ
ajst-3607	4	35	attention	attention	NOUN
ajst-3607	4	36	to	to	ADP
ajst-3607	4	37	the	the	DET
ajst-3607	4	38	details	detail	NOUN
ajst-3607	4	39	of	of	ADP
ajst-3607	4	40	key	key	ADJ
ajst-3607	4	41	features	feature	NOUN
ajst-3607	4	42	.	.	PUNCT
ajst-3607	5	1	finally	finally	ADV
ajst-3607	5	2	,	,	PUNCT
ajst-3607	5	3	the	the	DET
ajst-3607	5	4	feature	feature	NOUN
ajst-3607	5	5	information	information	NOUN
ajst-3607	5	6	is	be	AUX
ajst-3607	5	7	sent	send	VERB
ajst-3607	5	8	to	to	ADP
ajst-3607	5	9	lstm	lstm	NOUN
ajst-3607	5	10	network	network	NOUN
ajst-3607	5	11	with	with	ADP
ajst-3607	5	12	memory	memory	NOUN
ajst-3607	5	13	function	function	NOUN
ajst-3607	5	14	for	for	ADP
ajst-3607	5	15	action	action	NOUN
ajst-3607	5	16	recognition	recognition	NOUN
ajst-3607	5	17	and	and	CCONJ
ajst-3607	5	18	classification	classification	NOUN
ajst-3607	5	19	.	.	PUNCT
ajst-3607	6	1	the	the	DET
ajst-3607	6	2	multi	multi	ADJ
ajst-3607	6	3	-	-	ADJ
ajst-3607	6	4	model	model	ADJ
ajst-3607	6	5	mutual	mutual	ADJ
ajst-3607	6	6	fusion	fusion	NOUN
ajst-3607	6	7	algorithm	algorithm	NOUN
ajst-3607	6	8	proposed	propose	VERB
ajst-3607	6	9	in	in	ADP
ajst-3607	6	10	this	this	DET
ajst-3607	6	11	paper	paper	NOUN
ajst-3607	6	12	is	be	AUX
ajst-3607	6	13	tested	test	VERB
ajst-3607	6	14	and	and	CCONJ
ajst-3607	6	15	verified	verify	VERB
ajst-3607	6	16	on	on	ADP
ajst-3607	6	17	an	an	DET
ajst-3607	6	18	internationally	internationally	ADV
ajst-3607	6	19	published	publish	VERB
ajst-3607	6	20	ut	ut	NOUN
ajst-3607	6	21	-	-	PUNCT
ajst-3607	6	22	interaction	interaction	NOUN
ajst-3607	6	23	data	datum	NOUN
ajst-3607	6	24	set	set	VERB
ajst-3607	6	25	.	.	PUNCT
ajst-3607	7	1	the	the	DET
ajst-3607	7	2	experimental	experimental	ADJ
ajst-3607	7	3	results	result	NOUN
ajst-3607	7	4	show	show	VERB
ajst-3607	7	5	that	that	SCONJ
ajst-3607	7	6	the	the	DET
ajst-3607	7	7	accuracy	accuracy	NOUN
ajst-3607	7	8	of	of	ADP
ajst-3607	7	9	interactive	interactive	ADJ
ajst-3607	7	10	behavior	behavior	NOUN
ajst-3607	7	11	recognition	recognition	NOUN
ajst-3607	7	12	is	be	AUX
ajst-3607	7	13	improved	improve	VERB
ajst-3607	7	14	,	,	PUNCT
ajst-3607	7	15	and	and	CCONJ
ajst-3607	7	16	the	the	DET
ajst-3607	7	17	improved	improved	ADJ
ajst-3607	7	18	accuracy	accuracy	NOUN
ajst-3607	7	19	is	be	AUX
ajst-3607	7	20	85.1	85.1	NUM
ajst-3607	7	21	%	%	NOUN
ajst-3607	7	22	,	,	PUNCT
ajst-3607	7	23	which	which	PRON
ajst-3607	7	24	indicates	indicate	VERB
ajst-3607	7	25	that	that	SCONJ
ajst-3607	7	26	the	the	DET
ajst-3607	7	27	multi	multi	ADJ
ajst-3607	7	28	-	-	ADJ
ajst-3607	7	29	model	model	ADJ
ajst-3607	7	30	fusion	fusion	NOUN
ajst-3607	7	31	method	method	NOUN
ajst-3607	7	32	has	have	VERB
ajst-3607	7	33	higher	high	ADJ
ajst-3607	7	34	accuracy	accuracy	NOUN
ajst-3607	7	35	.	.	PUNCT
ajst-3607	8	1	keywords	keyword	NOUN
ajst-3607	8	2	:	:	PUNCT
ajst-3607	8	3	two	two	NUM
ajst-3607	8	4	-	-	PUNCT
ajst-3607	8	5	person	person	NOUN
ajst-3607	8	6	interaction	interaction	NOUN
ajst-3607	8	7	,	,	PUNCT
ajst-3607	8	8	behavior	behavior	NOUN
ajst-3607	8	9	recognition	recognition	NOUN
ajst-3607	8	10	,	,	PUNCT
ajst-3607	8	11	multi	multi	ADJ
ajst-3607	8	12	-	-	ADJ
ajst-3607	8	13	model	model	ADJ
ajst-3607	8	14	fusion	fusion	NOUN
ajst-3607	8	15	,	,	PUNCT
ajst-3607	8	16	deep	deep	ADJ
ajst-3607	8	17	learning	learning	NOUN
ajst-3607	8	18	,	,	PUNCT
ajst-3607	8	19	feature	feature	NOUN
ajst-3607	8	20	extraction	extraction	NOUN
ajst-3607	8	21	,	,	PUNCT
ajst-3607	8	22	modelling	model	VERB
ajst-3607	8	23	verification	verification	NOUN
ajst-3607	8	24	.	.	PUNCT
ajst-3607	9	1	1	1	X
ajst-3607	9	2	.	.	X
ajst-3607	9	3	introduction	introduction	NOUN
ajst-3607	9	4	with	with	ADP
ajst-3607	9	5	the	the	DET
ajst-3607	9	6	rapid	rapid	ADJ
ajst-3607	9	7	development	development	NOUN
ajst-3607	9	8	of	of	ADP
ajst-3607	9	9	the	the	DET
ajst-3607	9	10	"	"	PUNCT
ajst-3607	9	11	14th	14th	ADJ
ajst-3607	9	12	five	five	NUM
ajst-3607	9	13	-	-	PUNCT
ajst-3607	9	14	year	year	NOUN
ajst-3607	9	15	plan	plan	NOUN
ajst-3607	9	16	"	"	PUNCT
ajst-3607	9	17	,	,	PUNCT
ajst-3607	9	18	in	in	ADP
ajst-3607	9	19	order	order	NOUN
ajst-3607	9	20	to	to	PART
ajst-3607	9	21	maintain	maintain	VERB
ajst-3607	9	22	social	social	ADJ
ajst-3607	9	23	stability	stability	NOUN
ajst-3607	9	24	,	,	PUNCT
ajst-3607	9	25	the	the	DET
ajst-3607	9	26	construction	construction	NOUN
ajst-3607	9	27	of	of	ADP
ajst-3607	9	28	video	video	NOUN
ajst-3607	9	29	surveillance	surveillance	NOUN
ajst-3607	9	30	system	system	NOUN
ajst-3607	9	31	has	have	AUX
ajst-3607	9	32	been	be	AUX
ajst-3607	9	33	gradually	gradually	ADV
ajst-3607	9	34	promoted	promote	VERB
ajst-3607	9	35	nationwide	nationwide	ADV
ajst-3607	9	36	[	[	X
ajst-3607	9	37	1	1	NUM
ajst-3607	9	38	]	]	PUNCT
ajst-3607	9	39	.	.	PUNCT
ajst-3607	10	1	accurately	accurately	ADV
ajst-3607	10	2	predicting	predict	VERB
ajst-3607	10	3	and	and	CCONJ
ajst-3607	10	4	identifying	identify	VERB
ajst-3607	10	5	behaviors	behavior	NOUN
ajst-3607	10	6	in	in	ADP
ajst-3607	10	7	videos	video	NOUN
ajst-3607	10	8	is	be	AUX
ajst-3607	10	9	of	of	ADP
ajst-3607	10	10	great	great	ADJ
ajst-3607	10	11	significance	significance	NOUN
ajst-3607	10	12	to	to	ADP
ajst-3607	10	13	people	people	NOUN
ajst-3607	10	14	's	's	PART
ajst-3607	10	15	safety	safety	NOUN
ajst-3607	10	16	.	.	PUNCT
ajst-3607	11	1	the	the	DET
ajst-3607	11	2	behavior	behavior	NOUN
ajst-3607	11	3	prediction	prediction	NOUN
ajst-3607	11	4	and	and	CCONJ
ajst-3607	11	5	recognition	recognition	NOUN
ajst-3607	11	6	of	of	ADP
ajst-3607	11	7	video	video	NOUN
ajst-3607	11	8	interaction	interaction	NOUN
ajst-3607	11	9	between	between	ADP
ajst-3607	11	10	two	two	NUM
ajst-3607	11	11	people	people	NOUN
ajst-3607	11	12	has	have	AUX
ajst-3607	11	13	always	always	ADV
ajst-3607	11	14	been	be	AUX
ajst-3607	11	15	a	a	DET
ajst-3607	11	16	research	research	NOUN
ajst-3607	11	17	hotspot	hotspot	NOUN
ajst-3607	11	18	in	in	ADP
ajst-3607	11	19	the	the	DET
ajst-3607	11	20	field	field	NOUN
ajst-3607	11	21	of	of	ADP
ajst-3607	11	22	intelligent	intelligent	ADJ
ajst-3607	11	23	video	video	NOUN
ajst-3607	11	24	surveillance	surveillance	NOUN
ajst-3607	11	25	.	.	PUNCT
ajst-3607	12	1	literature	literature	NOUN
ajst-3607	13	1	[	[	X
ajst-3607	13	2	2	2	X
ajst-3607	13	3	]	]	PUNCT
ajst-3607	13	4	puts	put	VERB
ajst-3607	13	5	forward	forward	ADV
ajst-3607	13	6	a	a	DET
ajst-3607	13	7	method	method	NOUN
ajst-3607	13	8	of	of	ADP
ajst-3607	13	9	double	double	ADJ
ajst-3607	13	10	-	-	PUNCT
ajst-3607	13	11	person	person	NOUN
ajst-3607	13	12	behavior	behavior	NOUN
ajst-3607	13	13	recognition	recognition	NOUN
ajst-3607	13	14	of	of	ADP
ajst-3607	13	15	key	key	ADJ
ajst-3607	13	16	frames	frame	NOUN
ajst-3607	13	17	,	,	PUNCT
ajst-3607	13	18	but	but	CCONJ
ajst-3607	13	19	its	its	PRON
ajst-3607	13	20	key	key	ADJ
ajst-3607	13	21	frames	frame	NOUN
ajst-3607	13	22	have	have	VERB
ajst-3607	13	23	the	the	DET
ajst-3607	13	24	problem	problem	NOUN
ajst-3607	13	25	of	of	ADP
ajst-3607	13	26	background	background	NOUN
ajst-3607	13	27	interference	interference	NOUN
ajst-3607	13	28	,	,	PUNCT
ajst-3607	13	29	so	so	CCONJ
ajst-3607	13	30	it	it	PRON
ajst-3607	13	31	is	be	AUX
ajst-3607	13	32	easy	easy	ADJ
ajst-3607	13	33	to	to	PART
ajst-3607	13	34	mistake	mistake	VERB
ajst-3607	13	35	the	the	DET
ajst-3607	13	36	background	background	NOUN
ajst-3607	13	37	with	with	ADP
ajst-3607	13	38	the	the	DET
ajst-3607	13	39	target	target	NOUN
ajst-3607	13	40	.	.	PUNCT
ajst-3607	14	1	in	in	ADP
ajst-3607	14	2	order	order	NOUN
ajst-3607	14	3	to	to	PART
ajst-3607	14	4	fully	fully	ADV
ajst-3607	14	5	obtain	obtain	VERB
ajst-3607	14	6	the	the	DET
ajst-3607	14	7	effective	effective	ADJ
ajst-3607	14	8	information	information	NOUN
ajst-3607	14	9	of	of	ADP
ajst-3607	14	10	two	two	NUM
ajst-3607	14	11	-	-	PUNCT
ajst-3607	14	12	person	person	NOUN
ajst-3607	14	13	interaction	interaction	NOUN
ajst-3607	14	14	behavior	behavior	NOUN
ajst-3607	14	15	on	on	ADP
ajst-3607	14	16	the	the	DET
ajst-3607	14	17	characteristics	characteristic	NOUN
ajst-3607	14	18	of	of	ADP
ajst-3607	14	19	spatio	spatio	PROPN
ajst-3607	14	20	-	-	PUNCT
ajst-3607	14	21	temporal	temporal	ADJ
ajst-3607	14	22	network	network	NOUN
ajst-3607	14	23	,	,	PUNCT
ajst-3607	14	24	reference[3	reference[3	PROPN
ajst-3607	14	25	]	]	PUNCT
ajst-3607	14	26	proposed	propose	VERB
ajst-3607	14	27	to	to	PART
ajst-3607	14	28	construct	construct	VERB
ajst-3607	14	29	a	a	DET
ajst-3607	14	30	multi	multi	ADJ
ajst-3607	14	31	-	-	ADJ
ajst-3607	14	32	layer	layer	ADJ
ajst-3607	14	33	convolutional	convolutional	ADJ
ajst-3607	14	34	neural	neural	ADJ
ajst-3607	14	35	network	network	NOUN
ajst-3607	14	36	in	in	ADP
ajst-3607	14	37	three	three	NUM
ajst-3607	14	38	-	-	PUNCT
ajst-3607	14	39	dimensional	dimensional	ADJ
ajst-3607	14	40	space	space	NOUN
ajst-3607	14	41	,	,	PUNCT
ajst-3607	14	42	using	use	VERB
ajst-3607	14	43	the	the	DET
ajst-3607	14	44	way	way	NOUN
ajst-3607	14	45	that	that	PRON
ajst-3607	14	46	two	two	NUM
ajst-3607	14	47	layers	layer	NOUN
ajst-3607	14	48	of	of	ADP
ajst-3607	14	49	convolutional	convolutional	ADJ
ajst-3607	14	50	networks	network	NOUN
ajst-3607	14	51	are	be	AUX
ajst-3607	14	52	superimposed	superimpose	VERB
ajst-3607	14	53	to	to	PART
ajst-3607	14	54	form	form	VERB
ajst-3607	14	55	an	an	DET
ajst-3607	14	56	independent	independent	ADJ
ajst-3607	14	57	sub	sub	ADJ
ajst-3607	14	58	-	-	ADJ
ajst-3607	14	59	three	three	NUM
ajst-3607	14	60	-	-	PUNCT
ajst-3607	14	61	dimensional	dimensional	ADJ
ajst-3607	14	62	space	space	NOUN
ajst-3607	14	63	as	as	ADP
ajst-3607	14	64	the	the	DET
ajst-3607	14	65	model	model	NOUN
ajst-3607	14	66	of	of	ADP
ajst-3607	14	67	action	action	NOUN
ajst-3607	14	68	recognition	recognition	NOUN
ajst-3607	14	69	.	.	PUNCT
ajst-3607	15	1	however	however	ADV
ajst-3607	15	2	,	,	PUNCT
ajst-3607	15	3	the	the	DET
ajst-3607	15	4	multi	multi	ADJ
ajst-3607	15	5	-	-	ADJ
ajst-3607	15	6	layer	layer	ADJ
ajst-3607	15	7	convolutional	convolutional	ADJ
ajst-3607	15	8	neural	neural	ADJ
ajst-3607	15	9	network	network	NOUN
ajst-3607	15	10	constructed	construct	VERB
ajst-3607	15	11	in	in	ADP
ajst-3607	15	12	three	three	NUM
ajst-3607	15	13	-	-	PUNCT
ajst-3607	15	14	dimensional	dimensional	ADJ
ajst-3607	15	15	space	space	NOUN
ajst-3607	15	16	has	have	VERB
ajst-3607	15	17	complex	complex	ADJ
ajst-3607	15	18	structure	structure	NOUN
ajst-3607	15	19	,	,	PUNCT
ajst-3607	15	20	large	large	ADJ
ajst-3607	15	21	parameters	parameter	NOUN
ajst-3607	15	22	and	and	CCONJ
ajst-3607	15	23	high	high	ADJ
ajst-3607	15	24	hardware	hardware	NOUN
ajst-3607	15	25	requirements	requirement	NOUN
ajst-3607	15	26	.	.	PUNCT
ajst-3607	16	1	literature[4	literature[4	PROPN
ajst-3607	16	2	]	]	PUNCT
ajst-3607	16	3	proposes	propose	VERB
ajst-3607	16	4	a	a	DET
ajst-3607	16	5	multi	multi	ADJ
ajst-3607	16	6	-	-	ADJ
ajst-3607	16	7	dimensional	dimensional	ADJ
ajst-3607	16	8	spatio	spatio	NOUN
ajst-3607	16	9	-	-	PUNCT
ajst-3607	16	10	temporal	temporal	ADJ
ajst-3607	16	11	two	two	NUM
ajst-3607	16	12	-	-	PUNCT
ajst-3607	16	13	person	person	NOUN
ajst-3607	16	14	information	information	NOUN
ajst-3607	16	15	system	system	NOUN
ajst-3607	16	16	fusion	fusion	NOUN
ajst-3607	16	17	method	method	NOUN
ajst-3607	16	18	based	base	VERB
ajst-3607	16	19	on	on	ADP
ajst-3607	16	20	multi	multi	ADJ
ajst-3607	16	21	-	-	ADJ
ajst-3607	16	22	information	information	ADJ
ajst-3607	16	23	channels	channel	NOUN
ajst-3607	16	24	,	,	PUNCT
ajst-3607	16	25	which	which	PRON
ajst-3607	16	26	mainly	mainly	ADV
ajst-3607	16	27	fuses	fuse	VERB
ajst-3607	16	28	different	different	ADJ
ajst-3607	16	29	sequence	sequence	NOUN
ajst-3607	16	30	information	information	NOUN
ajst-3607	16	31	in	in	ADP
ajst-3607	16	32	the	the	DET
ajst-3607	16	33	skeleton	skeleton	NOUN
ajst-3607	16	34	network	network	NOUN
ajst-3607	16	35	of	of	ADP
ajst-3607	16	36	two	two	NUM
ajst-3607	16	37	-	-	PUNCT
ajst-3607	16	38	person	person	NOUN
ajst-3607	16	39	interaction	interaction	NOUN
ajst-3607	16	40	system	system	NOUN
ajst-3607	16	41	,	,	PUNCT
ajst-3607	16	42	but	but	CCONJ
ajst-3607	16	43	the	the	DET
ajst-3607	16	44	multi	multi	ADJ
ajst-3607	16	45	-	-	ADJ
ajst-3607	16	46	channel	channel	ADJ
ajst-3607	16	47	spatio	spatio	NOUN
ajst-3607	16	48	-	-	PUNCT
ajst-3607	16	49	temporal	temporal	ADJ
ajst-3607	16	50	fusion	fusion	NOUN
ajst-3607	16	51	network	network	NOUN
ajst-3607	16	52	has	have	VERB
ajst-3607	16	53	too	too	ADV
ajst-3607	16	54	many	many	ADJ
ajst-3607	16	55	network	network	NOUN
ajst-3607	16	56	parameters	parameter	NOUN
ajst-3607	16	57	to	to	PART
ajst-3607	16	58	be	be	AUX
ajst-3607	16	59	applied	apply	VERB
ajst-3607	16	60	in	in	ADP
ajst-3607	16	61	practice	practice	NOUN
ajst-3607	16	62	.	.	PUNCT
ajst-3607	17	1	literature[5	literature[5	PUNCT
ajst-3607	17	2	]	]	PUNCT
ajst-3607	17	3	puts	put	VERB
ajst-3607	17	4	forward	forward	ADV
ajst-3607	17	5	the	the	DET
ajst-3607	17	6	behavior	behavior	NOUN
ajst-3607	17	7	probability	probability	NOUN
ajst-3607	17	8	modeling	modeling	NOUN
ajst-3607	17	9	method	method	NOUN
ajst-3607	17	10	of	of	ADP
ajst-3607	17	11	twoperson	twoperson	NOUN
ajst-3607	17	12	dynamic	dynamic	ADJ
ajst-3607	17	13	bow(bag	bow(bag	NOUN
ajst-3607	17	14	of	of	ADP
ajst-3607	17	15	word	word	NOUN
ajst-3607	17	16	)	)	PUNCT
ajst-3607	17	17	,	,	PUNCT
ajst-3607	17	18	which	which	PRON
ajst-3607	17	19	can	can	AUX
ajst-3607	17	20	effectively	effectively	ADV
ajst-3607	17	21	solve	solve	VERB
ajst-3607	17	22	the	the	DET
ajst-3607	17	23	behavior	behavior	NOUN
ajst-3607	17	24	probability	probability	NOUN
ajst-3607	17	25	prediction	prediction	NOUN
ajst-3607	17	26	problem	problem	NOUN
ajst-3607	17	27	of	of	ADP
ajst-3607	17	28	twoperson	twoperson	NOUN
ajst-3607	17	29	collective	collective	ADJ
ajst-3607	17	30	interaction	interaction	NOUN
ajst-3607	17	31	.	.	PUNCT
ajst-3607	18	1	in	in	ADP
ajst-3607	18	2	this	this	DET
ajst-3607	18	3	method	method	NOUN
ajst-3607	18	4	,	,	PUNCT
ajst-3607	18	5	a	a	DET
ajst-3607	18	6	whole	whole	ADJ
ajst-3607	18	7	square	square	ADJ
ajst-3607	18	8	graph	graph	NOUN
ajst-3607	18	9	is	be	AUX
ajst-3607	18	10	used	use	VERB
ajst-3607	18	11	to	to	PART
ajst-3607	18	12	accurately	accurately	ADV
ajst-3607	18	13	express	express	VERB
ajst-3607	18	14	the	the	DET
ajst-3607	18	15	whole	whole	ADJ
ajst-3607	18	16	spatio	spatio	ADJ
ajst-3607	18	17	-	-	PUNCT
ajst-3607	18	18	temporal	temporal	ADJ
ajst-3607	18	19	characteristics	characteristic	NOUN
ajst-3607	18	20	of	of	ADP
ajst-3607	18	21	two	two	NUM
ajst-3607	18	22	-	-	PUNCT
ajst-3607	18	23	person	person	NOUN
ajst-3607	18	24	behavior	behavior	NOUN
ajst-3607	18	25	,	,	PUNCT
ajst-3607	18	26	and	and	CCONJ
ajst-3607	18	27	then	then	ADV
ajst-3607	18	28	a	a	DET
ajst-3607	18	29	probability	probability	NOUN
ajst-3607	18	30	model	model	NOUN
ajst-3607	18	31	is	be	AUX
ajst-3607	18	32	built	build	VERB
ajst-3607	18	33	to	to	PART
ajst-3607	18	34	predict	predict	VERB
ajst-3607	18	35	the	the	DET
ajst-3607	18	36	probability	probability	NOUN
ajst-3607	18	37	of	of	ADP
ajst-3607	18	38	two	two	NUM
ajst-3607	18	39	-	-	PUNCT
ajst-3607	18	40	person	person	NOUN
ajst-3607	18	41	behavior	behavior	NOUN
ajst-3607	18	42	.	.	PUNCT
ajst-3607	19	1	literature[6	literature[6	NOUN
ajst-3607	19	2	]	]	PUNCT
ajst-3607	19	3	proposed	propose	VERB
ajst-3607	19	4	a	a	DET
ajst-3607	19	5	prediction	prediction	NOUN
ajst-3607	19	6	framework	framework	NOUN
ajst-3607	19	7	based	base	VERB
ajst-3607	19	8	on	on	ADP
ajst-3607	19	9	the	the	DET
ajst-3607	19	10	combination	combination	NOUN
ajst-3607	19	11	of	of	ADP
ajst-3607	19	12	local	local	ADJ
ajst-3607	19	13	interest	interest	NOUN
ajst-3607	19	14	points	point	NOUN
ajst-3607	19	15	and	and	CCONJ
ajst-3607	19	16	sparse	sparse	ADJ
ajst-3607	19	17	representation	representation	NOUN
ajst-3607	19	18	.	.	PUNCT
ajst-3607	20	1	each	each	DET
ajst-3607	20	2	video	video	NOUN
ajst-3607	20	3	is	be	AUX
ajst-3607	20	4	divided	divide	VERB
ajst-3607	20	5	into	into	ADP
ajst-3607	20	6	multiple	multiple	ADJ
ajst-3607	20	7	time	time	NOUN
ajst-3607	20	8	periods	period	NOUN
ajst-3607	20	9	,	,	PUNCT
ajst-3607	20	10	and	and	CCONJ
ajst-3607	20	11	the	the	DET
ajst-3607	20	12	category	category	NOUN
ajst-3607	20	13	of	of	ADP
ajst-3607	20	14	two	two	NUM
ajst-3607	20	15	-	-	PUNCT
ajst-3607	20	16	person	person	NOUN
ajst-3607	20	17	activities	activity	NOUN
ajst-3607	20	18	is	be	AUX
ajst-3607	20	19	determined	determine	VERB
ajst-3607	20	20	by	by	ADP
ajst-3607	20	21	constructing	construct	VERB
ajst-3607	20	22	a	a	DET
ajst-3607	20	23	sparse	sparse	ADJ
ajst-3607	20	24	word	word	NOUN
ajst-3607	20	25	packet	packet	NOUN
ajst-3607	20	26	in	in	ADP
ajst-3607	20	27	each	each	DET
ajst-3607	20	28	time	time	NOUN
ajst-3607	20	29	period	period	NOUN
ajst-3607	20	30	.	.	PUNCT
ajst-3607	21	1	because	because	SCONJ
ajst-3607	21	2	each	each	DET
ajst-3607	21	3	video	video	NOUN
ajst-3607	21	4	is	be	AUX
ajst-3607	21	5	divided	divide	VERB
ajst-3607	21	6	into	into	ADP
ajst-3607	21	7	multiple	multiple	ADJ
ajst-3607	21	8	time	time	NOUN
ajst-3607	21	9	periods	period	NOUN
ajst-3607	21	10	,	,	PUNCT
ajst-3607	21	11	the	the	DET
ajst-3607	21	12	time	time	NOUN
ajst-3607	21	13	information	information	NOUN
ajst-3607	21	14	and	and	CCONJ
ajst-3607	21	15	space	space	NOUN
ajst-3607	21	16	information	information	NOUN
ajst-3607	21	17	are	be	AUX
ajst-3607	21	18	separated	separate	VERB
ajst-3607	21	19	,	,	PUNCT
ajst-3607	21	20	and	and	CCONJ
ajst-3607	21	21	the	the	DET
ajst-3607	21	22	key	key	ADJ
ajst-3607	21	23	information	information	NOUN
ajst-3607	21	24	of	of	ADP
ajst-3607	21	25	the	the	DET
ajst-3607	21	26	video	video	NOUN
ajst-3607	21	27	can	can	AUX
ajst-3607	21	28	not	not	PART
ajst-3607	21	29	be	be	AUX
ajst-3607	21	30	fully	fully	ADV
ajst-3607	21	31	utilized	utilize	VERB
ajst-3607	21	32	,	,	PUNCT
ajst-3607	21	33	which	which	PRON
ajst-3607	21	34	leads	lead	VERB
ajst-3607	21	35	to	to	ADP
ajst-3607	21	36	a	a	DET
ajst-3607	21	37	low	low	ADJ
ajst-3607	21	38	recognition	recognition	NOUN
ajst-3607	21	39	rate	rate	NOUN
ajst-3607	21	40	.	.	PUNCT
ajst-3607	22	1	in	in	ADP
ajst-3607	22	2	reference	reference	NOUN
ajst-3607	22	3	[	[	X
ajst-3607	22	4	7	7	NUM
ajst-3607	22	5	]	]	PUNCT
ajst-3607	22	6	,	,	PUNCT
ajst-3607	22	7	a	a	DET
ajst-3607	22	8	frame	frame	NOUN
ajst-3607	22	9	of	of	ADP
ajst-3607	22	10	motion	motion	NOUN
ajst-3607	22	11	recognition	recognition	NOUN
ajst-3607	22	12	and	and	CCONJ
ajst-3607	22	13	prediction	prediction	NOUN
ajst-3607	22	14	model	model	NOUN
ajst-3607	22	15	based	base	VERB
ajst-3607	22	16	on	on	ADP
ajst-3607	22	17	key	key	ADJ
ajst-3607	22	18	frames	frame	NOUN
ajst-3607	22	19	is	be	AUX
ajst-3607	22	20	proposed	propose	VERB
ajst-3607	22	21	.	.	PUNCT
ajst-3607	23	1	by	by	ADP
ajst-3607	23	2	extracting	extract	VERB
ajst-3607	23	3	the	the	DET
ajst-3607	23	4	key	key	ADJ
ajst-3607	23	5	frames	frame	NOUN
ajst-3607	23	6	of	of	ADP
ajst-3607	23	7	motion	motion	NOUN
ajst-3607	23	8	video	video	NOUN
ajst-3607	23	9	as	as	ADP
ajst-3607	23	10	the	the	DET
ajst-3607	23	11	state	state	NOUN
ajst-3607	23	12	nodes	node	NOUN
ajst-3607	23	13	of	of	ADP
ajst-3607	23	14	the	the	DET
ajst-3607	23	15	motion	motion	NOUN
ajst-3607	23	16	model	model	NOUN
ajst-3607	23	17	,	,	PUNCT
ajst-3607	23	18	the	the	DET
ajst-3607	23	19	frame	frame	NOUN
ajst-3607	23	20	of	of	ADP
ajst-3607	23	21	two	two	NUM
ajst-3607	23	22	-	-	PUNCT
ajst-3607	23	23	person	person	NOUN
ajst-3607	23	24	behavior	behavior	NOUN
ajst-3607	23	25	recognition	recognition	NOUN
ajst-3607	23	26	and	and	CCONJ
ajst-3607	23	27	prediction	prediction	NOUN
ajst-3607	23	28	is	be	AUX
ajst-3607	23	29	realized	realize	VERB
ajst-3607	23	30	.	.	PUNCT
ajst-3607	24	1	reference	reference	NOUN
ajst-3607	24	2	[	[	X
ajst-3607	24	3	8	8	NUM
ajst-3607	24	4	]	]	PUNCT
ajst-3607	24	5	proposed	propose	VERB
ajst-3607	24	6	a	a	DET
ajst-3607	24	7	multi	multi	ADJ
ajst-3607	24	8	-	-	ADJ
ajst-3607	24	9	scale	scale	ADJ
ajst-3607	24	10	kernel	kernel	NOUN
ajst-3607	24	11	model	model	NOUN
ajst-3607	24	12	method	method	NOUN
ajst-3607	24	13	.	.	PUNCT
ajst-3607	25	1	the	the	DET
ajst-3607	25	2	local	local	ADJ
ajst-3607	25	3	progress	progress	NOUN
ajst-3607	25	4	model	model	NOUN
ajst-3607	25	5	and	and	CCONJ
ajst-3607	25	6	the	the	DET
ajst-3607	25	7	global	global	ADJ
ajst-3607	25	8	progress	progress	NOUN
ajst-3607	25	9	model	model	NOUN
ajst-3607	25	10	are	be	AUX
ajst-3607	25	11	used	use	VERB
ajst-3607	25	12	to	to	PART
ajst-3607	25	13	capture	capture	VERB
ajst-3607	25	14	the	the	DET
ajst-3607	25	15	relationship	relationship	NOUN
ajst-3607	25	16	between	between	ADP
ajst-3607	25	17	the	the	DET
ajst-3607	25	18	time	time	NOUN
ajst-3607	25	19	progress	progress	NOUN
ajst-3607	25	20	and	and	CCONJ
ajst-3607	25	21	the	the	DET
ajst-3607	25	22	global	global	ADJ
ajst-3607	25	23	progress	progress	NOUN
ajst-3607	25	24	,	,	PUNCT
ajst-3607	25	25	so	so	SCONJ
ajst-3607	25	26	as	as	SCONJ
ajst-3607	25	27	to	to	PART
ajst-3607	25	28	complete	complete	VERB
ajst-3607	25	29	the	the	DET
ajst-3607	25	30	action	action	NOUN
ajst-3607	25	31	recognition	recognition	NOUN
ajst-3607	25	32	of	of	ADP
ajst-3607	25	33	the	the	DET
ajst-3607	25	34	local	local	ADJ
ajst-3607	25	35	video	video	NOUN
ajst-3607	25	36	.	.	PUNCT
ajst-3607	26	1	the	the	DET
ajst-3607	26	2	existing	exist	VERB
ajst-3607	26	3	problems	problem	NOUN
ajst-3607	26	4	of	of	ADP
ajst-3607	26	5	double	double	ADJ
ajst-3607	26	6	-	-	PUNCT
ajst-3607	26	7	person	person	NOUN
ajst-3607	26	8	behavior	behavior	NOUN
ajst-3607	26	9	recognition	recognition	NOUN
ajst-3607	26	10	mainly	mainly	ADV
ajst-3607	26	11	include	include	VERB
ajst-3607	26	12	:	:	PUNCT
ajst-3607	26	13	video	video	NOUN
ajst-3607	26	14	background	background	NOUN
ajst-3607	26	15	interference	interference	NOUN
ajst-3607	26	16	,	,	PUNCT
ajst-3607	26	17	incomplete	incomplete	ADJ
ajst-3607	26	18	human	human	ADJ
ajst-3607	26	19	target	target	NOUN
ajst-3607	26	20	detection	detection	NOUN
ajst-3607	26	21	,	,	PUNCT
ajst-3607	26	22	false	false	ADJ
ajst-3607	26	23	target	target	NOUN
ajst-3607	26	24	detection	detection	NOUN
ajst-3607	26	25	,	,	PUNCT
ajst-3607	26	26	complex	complex	ADJ
ajst-3607	26	27	network	network	NOUN
ajst-3607	26	28	,	,	PUNCT
ajst-3607	26	29	low	low	ADJ
ajst-3607	26	30	recognition	recognition	NOUN
ajst-3607	26	31	rate	rate	NOUN
ajst-3607	26	32	and	and	CCONJ
ajst-3607	26	33	so	so	ADV
ajst-3607	26	34	on	on	ADV
ajst-3607	26	35	.	.	PUNCT
ajst-3607	27	1	in	in	ADP
ajst-3607	27	2	order	order	NOUN
ajst-3607	27	3	to	to	PART
ajst-3607	27	4	solve	solve	VERB
ajst-3607	27	5	the	the	DET
ajst-3607	27	6	above	above	ADJ
ajst-3607	27	7	problems	problem	NOUN
ajst-3607	27	8	,	,	PUNCT
ajst-3607	27	9	yolo_v4	yolo_v4	NOUN
ajst-3607	27	10	is	be	AUX
ajst-3607	27	11	used	use	VERB
ajst-3607	27	12	to	to	PART
ajst-3607	27	13	detect	detect	VERB
ajst-3607	27	14	the	the	DET
ajst-3607	27	15	human	human	ADJ
ajst-3607	27	16	target	target	NOUN
ajst-3607	27	17	and	and	CCONJ
ajst-3607	27	18	remove	remove	VERB
ajst-3607	27	19	the	the	DET
ajst-3607	27	20	redundant	redundant	ADJ
ajst-3607	27	21	background	background	NOUN
ajst-3607	27	22	.	.	PUNCT
ajst-3607	28	1	after	after	ADP
ajst-3607	28	2	that	that	PRON
ajst-3607	28	3	,	,	PUNCT
ajst-3607	28	4	se	se	ADJ
ajst-3607	28	5	-	-	ADJ
ajst-3607	28	6	se	se	ADJ
ajst-3607	28	7	-	-	ADJ
ajst-3607	28	8	inception_v3	inception_v3	ADJ
ajst-3607	28	9	network	network	NOUN
ajst-3607	28	10	is	be	AUX
ajst-3607	28	11	proposed	propose	VERB
ajst-3607	28	12	to	to	PART
ajst-3607	28	13	recognize	recognize	VERB
ajst-3607	28	14	the	the	DET
ajst-3607	28	15	human	human	ADJ
ajst-3607	28	16	target	target	NOUN
ajst-3607	28	17	behavior	behavior	NOUN
ajst-3607	28	18	,	,	PUNCT
ajst-3607	28	19	and	and	CCONJ
ajst-3607	28	20	its	its	PRON
ajst-3607	28	21	advantages	advantage	NOUN
ajst-3607	28	22	are	be	AUX
ajst-3607	28	23	as	as	SCONJ
ajst-3607	28	24	follows	follow	VERB
ajst-3607	28	25	:	:	PUNCT
ajst-3607	28	26	1)inception_v3	1)inception_v3	NUM
ajst-3607	28	27	network	network	NOUN
ajst-3607	28	28	can	can	AUX
ajst-3607	28	29	extract	extract	VERB
ajst-3607	28	30	different	different	ADJ
ajst-3607	28	31	features	feature	NOUN
ajst-3607	28	32	by	by	ADP
ajst-3607	28	33	designing	design	VERB
ajst-3607	28	34	different	different	ADJ
ajst-3607	28	35	convolution	convolution	NOUN
ajst-3607	28	36	modules	module	NOUN
ajst-3607	28	37	,	,	PUNCT
ajst-3607	28	38	thus	thus	ADV
ajst-3607	28	39	effectively	effectively	ADV
ajst-3607	28	40	solving	solve	VERB
ajst-3607	28	41	the	the	DET
ajst-3607	28	42	problem	problem	NOUN
ajst-3607	28	43	of	of	ADP
ajst-3607	28	44	incomplete	incomplete	ADJ
ajst-3607	28	45	human	human	ADJ
ajst-3607	28	46	target	target	NOUN
ajst-3607	28	47	detection	detection	NOUN
ajst-3607	28	48	;	;	PUNCT
ajst-3607	28	49	2	2	X
ajst-3607	28	50	)	)	PUNCT
ajst-3607	28	51	the	the	DET
ajst-3607	28	52	channel	channel	NOUN
ajst-3607	28	53	attention	attention	NOUN
ajst-3607	28	54	model	model	NOUN
ajst-3607	28	55	se	se	PROPN
ajst-3607	28	56	-	-	NOUN
ajst-3607	28	57	net	net	ADJ
ajst-3607	28	58	can	can	AUX
ajst-3607	28	59	solve	solve	VERB
ajst-3607	28	60	the	the	DET
ajst-3607	28	61	problem	problem	NOUN
ajst-3607	28	62	of	of	ADP
ajst-3607	28	63	false	false	ADJ
ajst-3607	28	64	detection	detection	NOUN
ajst-3607	28	65	of	of	ADP
ajst-3607	28	66	targets	target	NOUN
ajst-3607	28	67	;	;	PUNCT
ajst-3607	28	68	3	3	X
ajst-3607	28	69	)	)	PUNCT
ajst-3607	28	70	the	the	DET
ajst-3607	28	71	extracted	extract	VERB
ajst-3607	28	72	features	feature	NOUN
ajst-3607	28	73	are	be	AUX
ajst-3607	28	74	rich	rich	ADJ
ajst-3607	28	75	,	,	PUNCT
ajst-3607	28	76	and	and	CCONJ
ajst-3607	28	77	the	the	DET
ajst-3607	28	78	network	network	NOUN
ajst-3607	28	79	reasoning	reasoning	NOUN
ajst-3607	28	80	speed	speed	NOUN
ajst-3607	28	81	is	be	AUX
ajst-3607	28	82	fast	fast	ADJ
ajst-3607	28	83	.	.	PUNCT
ajst-3607	29	1	2	2	X
ajst-3607	29	2	.	.	X
ajst-3607	29	3	two	two	NUM
ajst-3607	29	4	-	-	PUNCT
ajst-3607	29	5	person	person	NOUN
ajst-3607	29	6	interaction	interaction	NOUN
ajst-3607	29	7	behavior	behavior	NOUN
ajst-3607	29	8	recognition	recognition	NOUN
ajst-3607	29	9	2.1	2.1	NUM
ajst-3607	29	10	.	.	PUNCT
ajst-3607	30	1	human	human	ADJ
ajst-3607	30	2	target	target	NOUN
ajst-3607	30	3	detection	detection	NOUN
ajst-3607	30	4	network	network	NOUN
ajst-3607	30	5	-	-	PUNCT
ajst-3607	30	6	yolo	yolo	NOUN
ajst-3607	30	7	_	_	PRON
ajst-3607	30	8	v4	v4	NOUN
ajst-3607	30	9	at	at	ADP
ajst-3607	30	10	present	present	ADJ
ajst-3607	30	11	,	,	PUNCT
ajst-3607	30	12	yolo_v4[9	yolo_v4[9	PROPN
ajst-3607	30	13	]	]	PUNCT
ajst-3607	30	14	is	be	AUX
ajst-3607	30	15	selected	select	VERB
ajst-3607	30	16	as	as	ADP
ajst-3607	30	17	the	the	DET
ajst-3607	30	18	target	target	NOUN
ajst-3607	30	19	detection	detection	NOUN
ajst-3607	30	20	network	network	NOUN
ajst-3607	30	21	.	.	PUNCT
ajst-3607	31	1	the	the	DET
ajst-3607	31	2	principle	principle	NOUN
ajst-3607	31	3	of	of	ADP
ajst-3607	31	4	yolo_v4	yolo_v4	NOUN
ajst-3607	31	5	target	target	NOUN
ajst-3607	31	6	detection	detection	NOUN
ajst-3607	31	7	is	be	AUX
ajst-3607	31	8	to	to	ADP
ajst-3607	31	9	150	150	NUM
ajst-3607	31	10	transform	transform	VERB
ajst-3607	31	11	the	the	DET
ajst-3607	31	12	target	target	NOUN
ajst-3607	31	13	detection	detection	NOUN
ajst-3607	31	14	results	result	NOUN
ajst-3607	31	15	into	into	ADP
ajst-3607	31	16	the	the	DET
ajst-3607	31	17	regression	regression	NOUN
ajst-3607	31	18	probability	probability	NOUN
ajst-3607	31	19	of	of	ADP
ajst-3607	31	20	nodes	node	NOUN
ajst-3607	31	21	,	,	PUNCT
ajst-3607	31	22	and	and	CCONJ
ajst-3607	31	23	the	the	DET
ajst-3607	31	24	convolution	convolution	NOUN
ajst-3607	31	25	method	method	NOUN
ajst-3607	31	26	and	and	CCONJ
ajst-3607	31	27	neural	neural	ADJ
ajst-3607	31	28	network	network	NOUN
ajst-3607	31	29	algorithm	algorithm	NOUN
ajst-3607	31	30	are	be	AUX
ajst-3607	31	31	used	use	VERB
ajst-3607	31	32	to	to	PART
ajst-3607	31	33	predict	predict	VERB
ajst-3607	31	34	the	the	DET
ajst-3607	31	35	boundary	boundary	ADJ
ajst-3607	31	36	box	box	NOUN
ajst-3607	31	37	of	of	ADP
ajst-3607	31	38	nodes	node	NOUN
ajst-3607	31	39	and	and	CCONJ
ajst-3607	31	40	the	the	DET
ajst-3607	31	41	regression	regression	NOUN
ajst-3607	31	42	probability	probability	NOUN
ajst-3607	31	43	of	of	ADP
ajst-3607	31	44	the	the	DET
ajst-3607	31	45	target	target	NOUN
ajst-3607	31	46	category	category	NOUN
ajst-3607	31	47	to	to	PART
ajst-3607	31	48	which	which	PRON
ajst-3607	31	49	the	the	DET
ajst-3607	31	50	nodes	node	NOUN
ajst-3607	31	51	belong	belong	VERB
ajst-3607	31	52	.	.	PUNCT
ajst-3607	32	1	yolo_v4	yolo_v4	NOUN
ajst-3607	32	2	is	be	AUX
ajst-3607	32	3	used	use	VERB
ajst-3607	32	4	to	to	PART
ajst-3607	32	5	divide	divide	VERB
ajst-3607	32	6	the	the	DET
ajst-3607	32	7	input	input	NOUN
ajst-3607	32	8	object	object	NOUN
ajst-3607	32	9	image	image	NOUN
ajst-3607	32	10	first	first	ADV
ajst-3607	32	11	,	,	PUNCT
ajst-3607	32	12	and	and	CCONJ
ajst-3607	32	13	then	then	ADV
ajst-3607	32	14	s×s	s×s	PROPN
ajst-3607	32	15	grids	grid	NOUN
ajst-3607	32	16	are	be	AUX
ajst-3607	32	17	formed	form	VERB
ajst-3607	32	18	.	.	PUNCT
ajst-3607	33	1	then	then	ADV
ajst-3607	33	2	,	,	PUNCT
ajst-3607	33	3	each	each	DET
ajst-3607	33	4	grid	grid	NOUN
ajst-3607	33	5	detects	detect	NOUN
ajst-3607	33	6	whether	whether	SCONJ
ajst-3607	33	7	an	an	DET
ajst-3607	33	8	object	object	NOUN
ajst-3607	33	9	falls	fall	VERB
ajst-3607	33	10	into	into	ADP
ajst-3607	33	11	the	the	DET
ajst-3607	33	12	grid	grid	NOUN
ajst-3607	33	13	.	.	PUNCT
ajst-3607	34	1	when	when	SCONJ
ajst-3607	34	2	one	one	NUM
ajst-3607	34	3	of	of	ADP
ajst-3607	34	4	the	the	DET
ajst-3607	34	5	center	center	NOUN
ajst-3607	34	6	coordinates	coordinate	NOUN
ajst-3607	34	7	of	of	ADP
ajst-3607	34	8	the	the	DET
ajst-3607	34	9	detected	detect	VERB
ajst-3607	34	10	object	object	NOUN
ajst-3607	34	11	falls	fall	VERB
ajst-3607	34	12	in	in	ADP
ajst-3607	34	13	the	the	DET
ajst-3607	34	14	grid	grid	NOUN
ajst-3607	34	15	,	,	PUNCT
ajst-3607	34	16	it	it	PRON
ajst-3607	34	17	means	mean	VERB
ajst-3607	34	18	that	that	SCONJ
ajst-3607	34	19	the	the	DET
ajst-3607	34	20	grid	grid	NOUN
ajst-3607	34	21	can	can	AUX
ajst-3607	34	22	detect	detect	VERB
ajst-3607	34	23	the	the	DET
ajst-3607	34	24	object	object	NOUN
ajst-3607	34	25	.	.	PUNCT
ajst-3607	35	1	each	each	DET
ajst-3607	35	2	grid	grid	NOUN
ajst-3607	35	3	is	be	AUX
ajst-3607	35	4	responsible	responsible	ADJ
ajst-3607	35	5	for	for	ADP
ajst-3607	35	6	predicting	predict	VERB
ajst-3607	35	7	b	b	ADP
ajst-3607	35	8	bounding	bounding	NOUN
ajst-3607	35	9	boxes	box	NOUN
ajst-3607	35	10	and	and	CCONJ
ajst-3607	35	11	the	the	DET
ajst-3607	35	12	confidence	confidence	NOUN
ajst-3607	35	13	score	score	NOUN
ajst-3607	35	14	of	of	ADP
ajst-3607	35	15	these	these	DET
ajst-3607	35	16	bounding	bounding	NOUN
ajst-3607	35	17	boxes	box	NOUN
ajst-3607	35	18	.	.	PUNCT
ajst-3607	36	1	the	the	DET
ajst-3607	36	2	confidence	confidence	NOUN
ajst-3607	36	3	score	score	NOUN
ajst-3607	36	4	reflects	reflect	VERB
ajst-3607	36	5	whether	whether	SCONJ
ajst-3607	36	6	the	the	DET
ajst-3607	36	7	grid	grid	NOUN
ajst-3607	36	8	has	have	AUX
ajst-3607	36	9	detected	detect	VERB
ajst-3607	36	10	objects	object	NOUN
ajst-3607	36	11	and	and	CCONJ
ajst-3607	36	12	the	the	DET
ajst-3607	36	13	accuracy	accuracy	NOUN
ajst-3607	36	14	of	of	ADP
ajst-3607	36	15	the	the	DET
ajst-3607	36	16	detected	detect	VERB
ajst-3607	36	17	objects	object	NOUN
ajst-3607	36	18	.	.	PUNCT
ajst-3607	37	1	confidence	confidence	NOUN
ajst-3607	37	2	is	be	AUX
ajst-3607	37	3	defined	define	VERB
ajst-3607	37	4	as	as	ADP
ajst-3607	37	5	:	:	PUNCT
ajst-3607	37	6	after	after	ADP
ajst-3607	37	7	obtaining	obtain	VERB
ajst-3607	37	8	the	the	DET
ajst-3607	37	9	confidence	confidence	NOUN
ajst-3607	37	10	score	score	NOUN
ajst-3607	37	11	,	,	PUNCT
ajst-3607	37	12	compared	compare	VERB
ajst-3607	37	13	with	with	ADP
ajst-3607	37	14	the	the	DET
ajst-3607	37	15	manually	manually	ADV
ajst-3607	37	16	set	set	VERB
ajst-3607	37	17	threshold	threshold	NOUN
ajst-3607	37	18	,	,	PUNCT
ajst-3607	37	19	if	if	SCONJ
ajst-3607	37	20	the	the	DET
ajst-3607	37	21	confidence	confidence	NOUN
ajst-3607	37	22	score	score	NOUN
ajst-3607	37	23	is	be	AUX
ajst-3607	37	24	greater	great	ADJ
ajst-3607	37	25	than	than	ADP
ajst-3607	37	26	the	the	DET
ajst-3607	37	27	manually	manually	ADV
ajst-3607	37	28	set	set	VERB
ajst-3607	37	29	threshold	threshold	NOUN
ajst-3607	37	30	,	,	PUNCT
ajst-3607	37	31	the	the	DET
ajst-3607	37	32	prediction	prediction	NOUN
ajst-3607	37	33	box	box	NOUN
ajst-3607	37	34	will	will	AUX
ajst-3607	37	35	be	be	AUX
ajst-3607	37	36	kept	keep	VERB
ajst-3607	37	37	.	.	PUNCT
ajst-3607	38	1	when	when	SCONJ
ajst-3607	38	2	the	the	DET
ajst-3607	38	3	confidence	confidence	NOUN
ajst-3607	38	4	score	score	NOUN
ajst-3607	38	5	is	be	AUX
ajst-3607	38	6	less	less	ADJ
ajst-3607	38	7	than	than	ADP
ajst-3607	38	8	the	the	DET
ajst-3607	38	9	manually	manually	ADV
ajst-3607	38	10	set	set	VERB
ajst-3607	38	11	threshold	threshold	NOUN
ajst-3607	38	12	,	,	PUNCT
ajst-3607	38	13	the	the	DET
ajst-3607	38	14	prediction	prediction	NOUN
ajst-3607	38	15	box	box	NOUN
ajst-3607	38	16	is	be	AUX
ajst-3607	38	17	discarded	discard	VERB
ajst-3607	38	18	.	.	PUNCT
ajst-3607	39	1	yolo_v4	yolo_v4	NOUN
ajst-3607	39	2	is	be	AUX
ajst-3607	39	3	used	use	VERB
ajst-3607	39	4	to	to	PART
ajst-3607	39	5	detect	detect	VERB
ajst-3607	39	6	the	the	DET
ajst-3607	39	7	human	human	ADJ
ajst-3607	39	8	target	target	NOUN
ajst-3607	39	9	in	in	ADP
ajst-3607	39	10	the	the	DET
ajst-3607	39	11	video	video	NOUN
ajst-3607	39	12	first	first	ADV
ajst-3607	39	13	,	,	PUNCT
ajst-3607	39	14	and	and	CCONJ
ajst-3607	39	15	remove	remove	VERB
ajst-3607	39	16	the	the	DET
ajst-3607	39	17	interference	interference	NOUN
ajst-3607	39	18	of	of	ADP
ajst-3607	39	19	background	background	NOUN
ajst-3607	39	20	redundant	redundant	ADJ
ajst-3607	39	21	information	information	NOUN
ajst-3607	39	22	.	.	PUNCT
ajst-3607	40	1	2.2	2.2	NUM
ajst-3607	40	2	.	.	PUNCT
ajst-3607	41	1	feature	feature	NOUN
ajst-3607	41	2	extraction	extraction	NOUN
ajst-3607	41	3	network	network	NOUN
ajst-3607	41	4	-	-	PUNCT
ajst-3607	41	5	se	se	NOUN
ajst-3607	41	6	-	-	NOUN
ajst-3607	41	7	incidence	incidence	NOUN
ajst-3607	41	8	_	_	PUNCT
ajst-3607	41	9	v3	v3	PROPN
ajst-3607	41	10	in	in	ADP
ajst-3607	41	11	the	the	DET
ajst-3607	41	12	process	process	NOUN
ajst-3607	41	13	of	of	ADP
ajst-3607	41	14	feature	feature	NOUN
ajst-3607	41	15	extraction	extraction	NOUN
ajst-3607	41	16	,	,	PUNCT
ajst-3607	41	17	inception_v3	inception_v3	VERB
ajst-3607	41	18	uses	use	VERB
ajst-3607	41	19	convolution	convolution	NOUN
ajst-3607	41	20	kernels	kernel	NOUN
ajst-3607	41	21	of	of	ADP
ajst-3607	41	22	different	different	ADJ
ajst-3607	41	23	sizes	size	NOUN
ajst-3607	41	24	to	to	PART
ajst-3607	41	25	obtain	obtain	VERB
ajst-3607	41	26	features	feature	NOUN
ajst-3607	41	27	of	of	ADP
ajst-3607	41	28	different	different	ADJ
ajst-3607	41	29	sizes	size	NOUN
ajst-3607	41	30	,	,	PUNCT
ajst-3607	41	31	so	so	SCONJ
ajst-3607	41	32	as	as	SCONJ
ajst-3607	41	33	to	to	PART
ajst-3607	41	34	extract	extract	VERB
ajst-3607	41	35	rich	rich	ADJ
ajst-3607	41	36	features	feature	NOUN
ajst-3607	41	37	.	.	PUNCT
ajst-3607	42	1	in	in	ADP
ajst-3607	42	2	order	order	NOUN
ajst-3607	42	3	to	to	PART
ajst-3607	42	4	pay	pay	VERB
ajst-3607	42	5	more	more	ADJ
ajst-3607	42	6	attention	attention	NOUN
ajst-3607	42	7	to	to	ADP
ajst-3607	42	8	important	important	ADJ
ajst-3607	42	9	features	feature	NOUN
ajst-3607	42	10	and	and	CCONJ
ajst-3607	42	11	prevent	prevent	VERB
ajst-3607	42	12	secondary	secondary	ADJ
ajst-3607	42	13	features	feature	NOUN
ajst-3607	42	14	,	,	PUNCT
ajst-3607	42	15	this	this	DET
ajst-3607	42	16	paper	paper	NOUN
ajst-3607	42	17	adds	add	VERB
ajst-3607	42	18	an	an	DET
ajst-3607	42	19	se	se	ADJ
ajst-3607	42	20	-	-	NOUN
ajst-3607	42	21	module	module	NOUN
ajst-3607	42	22	module	module	NOUN
ajst-3607	42	23	after	after	ADP
ajst-3607	42	24	each	each	DET
ajst-3607	42	25	sub	sub	NOUN
ajst-3607	42	26	-	-	NOUN
ajst-3607	42	27	module	module	NOUN
ajst-3607	42	28	of	of	ADP
ajst-3607	42	29	incidence	incidence	NOUN
ajst-3607	42	30	_	_	PUNCT
ajst-3607	43	1	v3	v3	PROPN
ajst-3607	43	2	,	,	PUNCT
ajst-3607	43	3	which	which	PRON
ajst-3607	43	4	is	be	AUX
ajst-3607	43	5	called	call	VERB
ajst-3607	43	6	se	se	ADJ
ajst-3607	43	7	-	-	ADJ
ajst-3607	43	8	incidence	incidence	ADJ
ajst-3607	43	9	feature	feature	NOUN
ajst-3607	43	10	extraction	extraction	NOUN
ajst-3607	43	11	network	network	NOUN
ajst-3607	43	12	[	[	X
ajst-3607	43	13	10	10	NUM
ajst-3607	43	14	]	]	PUNCT
ajst-3607	43	15	.	.	PUNCT
ajst-3607	44	1	its	its	PRON
ajst-3607	44	2	overall	overall	ADJ
ajst-3607	44	3	structure	structure	NOUN
ajst-3607	44	4	is	be	AUX
ajst-3607	44	5	shown	show	VERB
ajst-3607	44	6	in	in	ADP
ajst-3607	44	7	figure	figure	NOUN
ajst-3607	44	8	1	1	NUM
ajst-3607	44	9	.	.	PUNCT
ajst-3607	44	10	figure	figure	NOUN
ajst-3607	44	11	1	1	NUM
ajst-3607	44	12	.	.	PUNCT
ajst-3607	44	13	general	general	ADJ
ajst-3607	44	14	structure	structure	NOUN
ajst-3607	44	15	diagram	diagram	NOUN
ajst-3607	44	16	of	of	ADP
ajst-3607	44	17	se	se	NOUN
ajst-3607	44	18	-	-	NOUN
ajst-3607	44	19	incidence	incidence	ADJ
ajst-3607	44	20	_	_	PUNCT
ajst-3607	45	1	v3	v3	PROPN
ajst-3607	45	2	figure	figure	NOUN
ajst-3607	45	3	.	.	PUNCT
ajst-3607	46	1	2	2	NUM
ajst-3607	46	2	is	be	AUX
ajst-3607	46	3	a	a	DET
ajst-3607	46	4	detailed	detailed	ADJ
ajst-3607	46	5	structure	structure	NOUN
ajst-3607	46	6	diagram	diagram	NOUN
ajst-3607	46	7	of	of	ADP
ajst-3607	46	8	se	se	NOUN
ajst-3607	46	9	-	-	NOUN
ajst-3607	46	10	inception	inception	NOUN
ajst-3607	46	11	.	.	PUNCT
ajst-3607	47	1	firstly	firstly	ADV
ajst-3607	47	2	,	,	PUNCT
ajst-3607	47	3	the	the	DET
ajst-3607	47	4	input	input	NOUN
ajst-3607	47	5	network	network	NOUN
ajst-3607	47	6	feature	feature	NOUN
ajst-3607	47	7	information	information	NOUN
ajst-3607	47	8	x	x	PUNCT
ajst-3607	47	9	is	be	AUX
ajst-3607	47	10	subjected	subject	VERB
ajst-3607	47	11	to	to	ADP
ajst-3607	47	12	a	a	DET
ajst-3607	47	13	global	global	ADJ
ajst-3607	47	14	pooling	pooling	NOUN
ajst-3607	47	15	operation	operation	NOUN
ajst-3607	47	16	,	,	PUNCT
ajst-3607	47	17	and	and	CCONJ
ajst-3607	47	18	then	then	ADV
ajst-3607	47	19	the	the	DET
ajst-3607	47	20	feature	feature	NOUN
ajst-3607	47	21	dimension	dimension	NOUN
ajst-3607	47	22	is	be	AUX
ajst-3607	47	23	reduced	reduce	VERB
ajst-3607	47	24	to	to	ADP
ajst-3607	47	25	the	the	DET
ajst-3607	47	26	original	original	ADJ
ajst-3607	47	27	r	r	NOUN
ajst-3607	47	28	1	1	NUM
ajst-3607	47	29	,	,	PUNCT
ajst-3607	47	30	through	through	ADP
ajst-3607	47	31	the	the	DET
ajst-3607	47	32	full	full	ADJ
ajst-3607	47	33	connection	connection	NOUN
ajst-3607	47	34	layer	layer	NOUN
ajst-3607	47	35	(	(	PUNCT
ajst-3607	47	36	fc	fc	PROPN
ajst-3607	47	37	)	)	PUNCT
ajst-3607	47	38	;	;	PUNCT
ajst-3607	47	39	then	then	ADV
ajst-3607	47	40	,	,	PUNCT
ajst-3607	47	41	the	the	DET
ajst-3607	47	42	feature	feature	NOUN
ajst-3607	47	43	dimension	dimension	NOUN
ajst-3607	47	44	is	be	AUX
ajst-3607	47	45	activated	activate	VERB
ajst-3607	47	46	by	by	ADP
ajst-3607	47	47	relu	relu	NOUN
ajst-3607	47	48	function	function	NOUN
ajst-3607	47	49	,	,	PUNCT
ajst-3607	47	50	and	and	CCONJ
ajst-3607	47	51	then	then	ADV
ajst-3607	47	52	transformed	transform	VERB
ajst-3607	47	53	into	into	ADP
ajst-3607	47	54	the	the	DET
ajst-3607	47	55	original	original	ADJ
ajst-3607	47	56	dimension	dimension	NOUN
ajst-3607	47	57	c	c	NOUN
ajst-3607	47	58	through	through	ADP
ajst-3607	47	59	a	a	DET
ajst-3607	47	60	global	global	ADJ
ajst-3607	47	61	complete	complete	ADJ
ajst-3607	47	62	connection	connection	NOUN
ajst-3607	47	63	layer	layer	NOUN
ajst-3607	47	64	(	(	PUNCT
ajst-3607	47	65	fc	fc	PROPN
ajst-3607	47	66	)	)	PUNCT
ajst-3607	47	67	;	;	PUNCT
ajst-3607	47	68	finally	finally	ADV
ajst-3607	47	69	,	,	PUNCT
ajst-3607	47	70	it	it	PRON
ajst-3607	47	71	is	be	AUX
ajst-3607	47	72	converted	convert	VERB
ajst-3607	47	73	into	into	ADP
ajst-3607	47	74	a	a	DET
ajst-3607	47	75	normalized	normalize	VERB
ajst-3607	47	76	weight	weight	NOUN
ajst-3607	47	77	of	of	ADP
ajst-3607	47	78	[	[	X
ajst-3607	47	79	0,1	0,1	NUM
ajst-3607	47	80	]	]	PUNCT
ajst-3607	47	81	by	by	ADP
ajst-3607	47	82	sigmoid	sigmoid	NOUN
ajst-3607	47	83	function	function	NOUN
ajst-3607	47	84	.	.	PUNCT
ajst-3607	48	1	figure	figure	NOUN
ajst-3607	48	2	2	2	NUM
ajst-3607	48	3	.	.	PUNCT
ajst-3607	48	4	structure	structure	NOUN
ajst-3607	48	5	diagram	diagram	NOUN
ajst-3607	48	6	of	of	ADP
ajst-3607	48	7	se	se	ADJ
ajst-3607	48	8	-	-	NOUN
ajst-3607	48	9	induction	induction	NOUN
ajst-3607	48	10	module	module	NOUN
ajst-3607	48	11	incident	incident	NOUN
ajst-3607	48	12	_	_	PROPN
ajst-3607	49	1	v3	v3	PROPN
ajst-3607	49	2	submodule	submodule	PROPN
ajst-3607	50	1	[	[	X
ajst-3607	50	2	11	11	NUM
ajst-3607	50	3	]	]	SYM
ajst-3607	50	4	:	:	PUNCT
ajst-3607	50	5	incident	incident	NOUN
ajst-3607	50	6	_	_	PROPN
ajst-3607	50	7	v3	v3	PROPN
ajst-3607	50	8	includes	include	VERB
ajst-3607	50	9	submodules	submodule	NOUN
ajst-3607	50	10	a	a	PRON
ajst-3607	50	11	,	,	PUNCT
ajst-3607	50	12	b	b	PROPN
ajst-3607	50	13	and	and	CCONJ
ajst-3607	50	14	c	c	NOUN
ajst-3607	50	15	,	,	PUNCT
ajst-3607	50	16	as	as	SCONJ
ajst-3607	50	17	shown	show	VERB
ajst-3607	50	18	in	in	ADP
ajst-3607	50	19	figure	figure	NOUN
ajst-3607	50	20	3	3	NUM
ajst-3607	50	21	.	.	PUNCT
ajst-3607	51	1	there	there	PRON
ajst-3607	51	2	are	be	VERB
ajst-3607	51	3	three	three	NUM
ajst-3607	51	4	sub	sub	NOUN
ajst-3607	51	5	-	-	NOUN
ajst-3607	51	6	modules	module	NOUN
ajst-3607	51	7	a.	a.	NOUN
ajst-3607	51	8	the	the	DET
ajst-3607	51	9	basic	basic	ADJ
ajst-3607	51	10	design	design	NOUN
ajst-3607	51	11	idea	idea	NOUN
ajst-3607	51	12	is	be	AUX
ajst-3607	51	13	to	to	PART
ajst-3607	51	14	solve	solve	VERB
ajst-3607	51	15	the	the	DET
ajst-3607	51	16	large	large	ADJ
ajst-3607	51	17	volume	volume	NOUN
ajst-3607	51	18	integral	integral	ADJ
ajst-3607	51	19	into	into	ADP
ajst-3607	51	20	several	several	ADJ
ajst-3607	51	21	small	small	ADJ
ajst-3607	51	22	convolutions	convolution	NOUN
ajst-3607	51	23	,	,	PUNCT
ajst-3607	51	24	and	and	CCONJ
ajst-3607	51	25	use	use	VERB
ajst-3607	51	26	two	two	NUM
ajst-3607	51	27	3×3	3×3	NUM
ajst-3607	51	28	convolution	convolution	NOUN
ajst-3607	51	29	kernels	kernel	NOUN
ajst-3607	51	30	instead	instead	ADV
ajst-3607	51	31	of	of	ADP
ajst-3607	51	32	one	one	NUM
ajst-3607	51	33	5×5	5×5	NUM
ajst-3607	51	34	convolution	convolution	NOUN
ajst-3607	51	35	kernel	kernel	NOUN
ajst-3607	51	36	.	.	PUNCT
ajst-3607	52	1	one	one	NUM
ajst-3607	52	2	of	of	ADP
ajst-3607	52	3	its	its	PRON
ajst-3607	52	4	main	main	ADJ
ajst-3607	52	5	features	feature	NOUN
ajst-3607	52	6	is	be	AUX
ajst-3607	52	7	to	to	PART
ajst-3607	52	8	effectively	effectively	ADV
ajst-3607	52	9	reduce	reduce	VERB
ajst-3607	52	10	the	the	DET
ajst-3607	52	11	151	151	NUM
ajst-3607	52	12	parameters	parameter	NOUN
ajst-3607	52	13	of	of	ADP
ajst-3607	52	14	the	the	DET
ajst-3607	52	15	model	model	NOUN
ajst-3607	52	16	while	while	SCONJ
ajst-3607	52	17	ensuring	ensure	VERB
ajst-3607	52	18	the	the	DET
ajst-3607	52	19	same	same	ADJ
ajst-3607	52	20	receptive	receptive	ADJ
ajst-3607	52	21	field	field	NOUN
ajst-3607	52	22	,	,	PUNCT
ajst-3607	52	23	as	as	SCONJ
ajst-3607	52	24	shown	show	VERB
ajst-3607	52	25	in	in	ADP
ajst-3607	52	26	figure	figure	NOUN
ajst-3607	52	27	3a	3a	NUM
ajst-3607	52	28	)	)	PUNCT
ajst-3607	52	29	;	;	PUNCT
ajst-3607	52	30	there	there	PRON
ajst-3607	52	31	are	be	VERB
ajst-3607	52	32	five	five	NUM
ajst-3607	52	33	sub	sub	NOUN
ajst-3607	52	34	-	-	NOUN
ajst-3607	52	35	modules	module	NOUN
ajst-3607	52	36	b.	b.	NOUN
ajst-3607	52	37	the	the	DET
ajst-3607	52	38	basic	basic	ADJ
ajst-3607	52	39	design	design	NOUN
ajst-3607	52	40	principle	principle	NOUN
ajst-3607	52	41	of	of	ADP
ajst-3607	52	42	sub	sub	NOUN
ajst-3607	52	43	-	-	NOUN
ajst-3607	52	44	module	module	ADJ
ajst-3607	52	45	b	b	NOUN
ajst-3607	52	46	is	be	AUX
ajst-3607	52	47	to	to	PART
ajst-3607	52	48	decompose	decompose	VERB
ajst-3607	52	49	spatial	spatial	ADJ
ajst-3607	52	50	convolution	convolution	NOUN
ajst-3607	52	51	into	into	ADP
ajst-3607	52	52	asymmetric	asymmetric	ADJ
ajst-3607	52	53	convolution	convolution	NOUN
ajst-3607	52	54	in	in	ADP
ajst-3607	52	55	series	series	NOUN
ajst-3607	52	56	.	.	PUNCT
ajst-3607	53	1	one	one	NUM
ajst-3607	53	2	of	of	ADP
ajst-3607	53	3	its	its	PRON
ajst-3607	53	4	main	main	ADJ
ajst-3607	53	5	features	feature	NOUN
ajst-3607	53	6	is	be	AUX
ajst-3607	53	7	to	to	PART
ajst-3607	53	8	decompose	decompose	VERB
ajst-3607	53	9	an	an	DET
ajst-3607	53	10	n×n	n×n	PROPN
ajst-3607	53	11	convolution	convolution	NOUN
ajst-3607	53	12	kernel	kernel	NOUN
ajst-3607	53	13	into	into	ADP
ajst-3607	53	14	a	a	DET
ajst-3607	53	15	series	series	NOUN
ajst-3607	53	16	combination	combination	NOUN
ajst-3607	53	17	of	of	ADP
ajst-3607	53	18	1×n	1×n	NUM
ajst-3607	53	19	and	and	CCONJ
ajst-3607	53	20	n×1	n×1	NOUN
ajst-3607	53	21	convolution	convolution	NOUN
ajst-3607	53	22	kernels	kernel	NOUN
ajst-3607	53	23	,	,	PUNCT
ajst-3607	53	24	which	which	PRON
ajst-3607	53	25	reduces	reduce	VERB
ajst-3607	53	26	the	the	DET
ajst-3607	53	27	computational	computational	ADJ
ajst-3607	53	28	load	load	NOUN
ajst-3607	53	29	and	and	CCONJ
ajst-3607	53	30	greatly	greatly	ADV
ajst-3607	53	31	reduces	reduce	VERB
ajst-3607	53	32	the	the	DET
ajst-3607	53	33	computational	computational	ADJ
ajst-3607	53	34	burden	burden	NOUN
ajst-3607	53	35	of	of	ADP
ajst-3607	53	36	hardware	hardware	NOUN
ajst-3607	53	37	,	,	PUNCT
ajst-3607	53	38	as	as	SCONJ
ajst-3607	53	39	shown	show	VERB
ajst-3607	53	40	in	in	ADP
ajst-3607	53	41	figure	figure	NOUN
ajst-3607	53	42	3b	3b	NUM
ajst-3607	53	43	)	)	PUNCT
ajst-3607	53	44	.	.	PUNCT
ajst-3607	54	1	there	there	PRON
ajst-3607	54	2	are	be	VERB
ajst-3607	54	3	two	two	NUM
ajst-3607	54	4	sub	sub	ADJ
ajst-3607	54	5	-	-	ADJ
ajst-3607	54	6	modules	module	NOUN
ajst-3607	54	7	c.	c.	NOUN
ajst-3607	54	8	one	one	NUM
ajst-3607	54	9	of	of	ADP
ajst-3607	54	10	the	the	DET
ajst-3607	54	11	main	main	ADJ
ajst-3607	54	12	features	feature	NOUN
ajst-3607	54	13	of	of	ADP
ajst-3607	54	14	sub	sub	NOUN
ajst-3607	54	15	-	-	NOUN
ajst-3607	54	16	module	module	ADJ
ajst-3607	54	17	c	c	NOUN
ajst-3607	54	18	is	be	AUX
ajst-3607	54	19	that	that	SCONJ
ajst-3607	54	20	the	the	DET
ajst-3607	54	21	3×3	3×3	NUM
ajst-3607	54	22	convolution	convolution	NOUN
ajst-3607	54	23	kernel	kernel	NOUN
ajst-3607	54	24	is	be	AUX
ajst-3607	54	25	decomposed	decompose	VERB
ajst-3607	54	26	into	into	ADP
ajst-3607	54	27	1×3	1×3	NUM
ajst-3607	54	28	and	and	CCONJ
ajst-3607	54	29	3×1	3×1	NUM
ajst-3607	54	30	convolution	convolution	NOUN
ajst-3607	54	31	kernels	kernel	NOUN
ajst-3607	54	32	and	and	CCONJ
ajst-3607	54	33	combined	combine	VERB
ajst-3607	54	34	in	in	ADP
ajst-3607	54	35	parallel	parallel	NOUN
ajst-3607	54	36	,	,	PUNCT
ajst-3607	54	37	which	which	PRON
ajst-3607	54	38	reduces	reduce	VERB
ajst-3607	54	39	the	the	DET
ajst-3607	54	40	workload	workload	NOUN
ajst-3607	54	41	of	of	ADP
ajst-3607	54	42	parameters	parameter	NOUN
ajst-3607	54	43	and	and	CCONJ
ajst-3607	54	44	operations	operation	NOUN
ajst-3607	54	45	,	,	PUNCT
ajst-3607	54	46	as	as	SCONJ
ajst-3607	54	47	shown	show	VERB
ajst-3607	54	48	in	in	ADP
ajst-3607	54	49	figure	figure	NOUN
ajst-3607	54	50	3c	3c	NUM
ajst-3607	54	51	)	)	PUNCT
ajst-3607	54	52	.	.	PUNCT
ajst-3607	55	1	figure	figure	VERB
ajst-3607	55	2	3	3	NUM
ajst-3607	55	3	.	.	PUNCT
ajst-3607	55	4	structure	structure	NOUN
ajst-3607	55	5	diagram	diagram	NOUN
ajst-3607	55	6	of	of	ADP
ajst-3607	55	7	incidence	incidence	NOUN
ajst-3607	55	8	_	_	PUNCT
ajst-3607	56	1	v3	v3	PROPN
ajst-3607	56	2	submodule	submodule	PROPN
ajst-3607	56	3	inception_v3	inception_v3	VERB
ajst-3607	56	4	is	be	AUX
ajst-3607	56	5	suitable	suitable	ADJ
ajst-3607	56	6	for	for	ADP
ajst-3607	56	7	smaller	small	ADJ
ajst-3607	56	8	data	datum	NOUN
ajst-3607	56	9	sets	set	NOUN
ajst-3607	56	10	because	because	SCONJ
ajst-3607	56	11	of	of	ADP
ajst-3607	56	12	its	its	PRON
ajst-3607	56	13	multi	multi	ADJ
ajst-3607	56	14	-	-	ADJ
ajst-3607	56	15	scale	scale	ADJ
ajst-3607	56	16	features	feature	NOUN
ajst-3607	56	17	and	and	CCONJ
ajst-3607	56	18	fewer	few	ADJ
ajst-3607	56	19	parameters	parameter	NOUN
ajst-3607	56	20	.	.	PUNCT
ajst-3607	57	1	two	two	NUM
ajst-3607	57	2	-	-	PUNCT
ajst-3607	57	3	person	person	NOUN
ajst-3607	57	4	interaction	interaction	NOUN
ajst-3607	57	5	ut_interaction	ut_interaction	NOUN
ajst-3607	57	6	data	datum	NOUN
ajst-3607	57	7	set	set	VERB
ajst-3607	57	8	is	be	AUX
ajst-3607	57	9	small	small	ADJ
ajst-3607	57	10	,	,	PUNCT
ajst-3607	57	11	so	so	SCONJ
ajst-3607	57	12	it	it	PRON
ajst-3607	57	13	is	be	AUX
ajst-3607	57	14	easy	easy	ADJ
ajst-3607	57	15	to	to	PART
ajst-3607	57	16	produce	produce	VERB
ajst-3607	57	17	over	over	ADP
ajst-3607	57	18	-	-	PUNCT
ajst-3607	57	19	fitting	fit	VERB
ajst-3607	57	20	phenomenon	phenomenon	NOUN
ajst-3607	57	21	when	when	SCONJ
ajst-3607	57	22	using	use	VERB
ajst-3607	57	23	large	large	ADJ
ajst-3607	57	24	network	network	NOUN
ajst-3607	57	25	.	.	PUNCT
ajst-3607	58	1	inception_v3	inception_v3	PROPN
ajst-3607	58	2	is	be	AUX
ajst-3607	58	3	used	use	VERB
ajst-3607	58	4	as	as	ADP
ajst-3607	58	5	the	the	DET
ajst-3607	58	6	backbone	backbone	NOUN
ajst-3607	58	7	network	network	NOUN
ajst-3607	58	8	of	of	ADP
ajst-3607	58	9	feature	feature	NOUN
ajst-3607	58	10	extraction	extraction	NOUN
ajst-3607	58	11	,	,	PUNCT
ajst-3607	58	12	which	which	PRON
ajst-3607	58	13	can	can	AUX
ajst-3607	58	14	extract	extract	VERB
ajst-3607	58	15	rich	rich	ADJ
ajst-3607	58	16	features	feature	NOUN
ajst-3607	58	17	and	and	CCONJ
ajst-3607	58	18	make	make	VERB
ajst-3607	58	19	the	the	DET
ajst-3607	58	20	network	network	NOUN
ajst-3607	58	21	reasoning	reasoning	NOUN
ajst-3607	58	22	faster	fast	ADV
ajst-3607	58	23	,	,	PUNCT
ajst-3607	58	24	thus	thus	ADV
ajst-3607	58	25	making	make	VERB
ajst-3607	58	26	preparations	preparation	NOUN
ajst-3607	58	27	for	for	ADP
ajst-3607	58	28	the	the	DET
ajst-3607	58	29	next	next	ADJ
ajst-3607	58	30	behavior	behavior	NOUN
ajst-3607	58	31	prediction	prediction	NOUN
ajst-3607	58	32	.	.	PUNCT
ajst-3607	59	1	2.3	2.3	NUM
ajst-3607	59	2	.	.	PUNCT
ajst-3607	60	1	behavior	behavior	NOUN
ajst-3607	60	2	prediction	prediction	NOUN
ajst-3607	60	3	and	and	CCONJ
ajst-3607	60	4	classification	classification	NOUN
ajst-3607	60	5	network	network	NOUN
ajst-3607	60	6	-	-	PUNCT
ajst-3607	60	7	lstm	lstm	NOUN
ajst-3607	60	8	network	network	NOUN
ajst-3607	60	9	lstm	lstm	NOUN
ajst-3607	60	10	(	(	PUNCT
ajst-3607	60	11	long	long	ADJ
ajst-3607	60	12	short	short	ADJ
ajst-3607	60	13	term	term	NOUN
ajst-3607	60	14	memory)[12	memory)[12	NOUN
ajst-3607	60	15	]	]	PUNCT
ajst-3607	60	16	is	be	AUX
ajst-3607	60	17	a	a	DET
ajst-3607	60	18	kind	kind	NOUN
ajst-3607	60	19	of	of	ADP
ajst-3607	60	20	recurrent	recurrent	ADJ
ajst-3607	60	21	neural	neural	ADJ
ajst-3607	60	22	network	network	NOUN
ajst-3607	60	23	(	(	PUNCT
ajst-3607	60	24	rnn	rnn	PROPN
ajst-3607	60	25	)	)	PUNCT
ajst-3607	60	26	,	,	PUNCT
ajst-3607	60	27	which	which	PRON
ajst-3607	60	28	has	have	VERB
ajst-3607	60	29	many	many	ADJ
ajst-3607	60	30	cyclic	cyclic	ADJ
ajst-3607	60	31	memory	memory	NOUN
ajst-3607	60	32	functions	function	NOUN
ajst-3607	60	33	that	that	PRON
ajst-3607	60	34	traditional	traditional	ADJ
ajst-3607	60	35	neural	neural	ADJ
ajst-3607	60	36	networks	network	NOUN
ajst-3607	60	37	do	do	AUX
ajst-3607	60	38	not	not	PART
ajst-3607	60	39	have	have	VERB
ajst-3607	60	40	,	,	PUNCT
ajst-3607	60	41	and	and	CCONJ
ajst-3607	60	42	can	can	AUX
ajst-3607	60	43	find	find	VERB
ajst-3607	60	44	the	the	DET
ajst-3607	60	45	sequence	sequence	NOUN
ajst-3607	60	46	relationship	relationship	NOUN
ajst-3607	60	47	between	between	ADP
ajst-3607	60	48	features	feature	NOUN
ajst-3607	60	49	.	.	PUNCT
ajst-3607	61	1	lstm	lstm	NOUN
ajst-3607	61	2	is	be	AUX
ajst-3607	61	3	effective	effective	ADJ
ajst-3607	61	4	in	in	ADP
ajst-3607	61	5	extracting	extract	VERB
ajst-3607	61	6	the	the	DET
ajst-3607	61	7	temporal	temporal	ADJ
ajst-3607	61	8	features	feature	NOUN
ajst-3607	61	9	of	of	ADP
ajst-3607	61	10	images	image	NOUN
ajst-3607	61	11	,	,	PUNCT
ajst-3607	61	12	so	so	SCONJ
ajst-3607	61	13	it	it	PRON
ajst-3607	61	14	has	have	VERB
ajst-3607	61	15	a	a	DET
ajst-3607	61	16	good	good	ADJ
ajst-3607	61	17	detection	detection	NOUN
ajst-3607	61	18	and	and	CCONJ
ajst-3607	61	19	recognition	recognition	NOUN
ajst-3607	61	20	effect	effect	NOUN
ajst-3607	61	21	in	in	ADP
ajst-3607	61	22	the	the	DET
ajst-3607	61	23	interaction	interaction	NOUN
ajst-3607	61	24	between	between	ADP
ajst-3607	61	25	two	two	NUM
ajst-3607	61	26	people	people	NOUN
ajst-3607	61	27	.	.	PUNCT
ajst-3607	62	1	the	the	DET
ajst-3607	62	2	main	main	ADJ
ajst-3607	62	3	structure	structure	NOUN
ajst-3607	62	4	of	of	ADP
ajst-3607	62	5	lstm	lstm	NOUN
ajst-3607	62	6	network	network	NOUN
ajst-3607	62	7	system	system	NOUN
ajst-3607	62	8	consists	consist	VERB
ajst-3607	62	9	of	of	ADP
ajst-3607	62	10	three	three	NUM
ajst-3607	62	11	parts	part	NOUN
ajst-3607	62	12	:	:	PUNCT
ajst-3607	62	13	the	the	DET
ajst-3607	62	14	first	first	ADJ
ajst-3607	62	15	part	part	NOUN
ajst-3607	62	16	is	be	AUX
ajst-3607	62	17	the	the	DET
ajst-3607	62	18	forgetting	forget	VERB
ajst-3607	62	19	gate	gate	NOUN
ajst-3607	62	20	of	of	ADP
ajst-3607	62	21	data	datum	NOUN
ajst-3607	62	22	input	input	NOUN
ajst-3607	62	23	,	,	PUNCT
ajst-3607	62	24	the	the	DET
ajst-3607	62	25	second	second	ADJ
ajst-3607	62	26	part	part	NOUN
ajst-3607	62	27	is	be	AUX
ajst-3607	62	28	the	the	DET
ajst-3607	62	29	input	input	NOUN
ajst-3607	62	30	gate	gate	NOUN
ajst-3607	62	31	of	of	ADP
ajst-3607	62	32	data	datum	NOUN
ajst-3607	62	33	input	input	NOUN
ajst-3607	62	34	,	,	PUNCT
ajst-3607	62	35	and	and	CCONJ
ajst-3607	62	36	the	the	DET
ajst-3607	62	37	third	third	ADJ
ajst-3607	62	38	part	part	NOUN
ajst-3607	62	39	is	be	AUX
ajst-3607	62	40	the	the	DET
ajst-3607	62	41	output	output	NOUN
ajst-3607	62	42	gate	gate	NOUN
ajst-3607	62	43	of	of	ADP
ajst-3607	62	44	data	datum	NOUN
ajst-3607	62	45	output	output	NOUN
ajst-3607	62	46	.	.	PUNCT
ajst-3607	63	1	intelligent	intelligent	ADJ
ajst-3607	63	2	building	building	NOUN
ajst-3607	63	3	is	be	AUX
ajst-3607	63	4	not	not	PART
ajst-3607	63	5	only	only	ADV
ajst-3607	63	6	a	a	DET
ajst-3607	63	7	simple	simple	ADJ
ajst-3607	63	8	application	application	NOUN
ajst-3607	63	9	of	of	ADP
ajst-3607	63	10	intelligent	intelligent	ADJ
ajst-3607	63	11	technology	technology	NOUN
ajst-3607	63	12	in	in	ADP
ajst-3607	63	13	architectural	architectural	ADJ
ajst-3607	63	14	engineering	engineering	NOUN
ajst-3607	63	15	,	,	PUNCT
ajst-3607	63	16	but	but	CCONJ
ajst-3607	63	17	also	also	ADV
ajst-3607	63	18	based	base	VERB
ajst-3607	63	19	on	on	ADP
ajst-3607	63	20	the	the	DET
ajst-3607	63	21	deep	deep	ADJ
ajst-3607	63	22	integration	integration	NOUN
ajst-3607	63	23	of	of	ADP
ajst-3607	63	24	architectural	architectural	ADJ
ajst-3607	63	25	design	design	NOUN
ajst-3607	63	26	,	,	PUNCT
ajst-3607	63	27	construction	construction	NOUN
ajst-3607	63	28	,	,	PUNCT
ajst-3607	63	29	operation	operation	NOUN
ajst-3607	63	30	,	,	PUNCT
ajst-3607	63	31	service	service	NOUN
ajst-3607	63	32	and	and	CCONJ
ajst-3607	63	33	high	high	ADJ
ajst-3607	63	34	-	-	PUNCT
ajst-3607	63	35	tech	tech	NOUN
ajst-3607	63	36	development	development	NOUN
ajst-3607	63	37	,	,	PUNCT
ajst-3607	63	38	embedding	embed	VERB
ajst-3607	63	39	and	and	CCONJ
ajst-3607	63	40	use	use	NOUN
ajst-3607	63	41	.	.	PUNCT
ajst-3607	64	1	it	it	PRON
ajst-3607	64	2	needs	need	VERB
ajst-3607	64	3	the	the	DET
ajst-3607	64	4	joint	joint	ADJ
ajst-3607	64	5	efforts	effort	NOUN
ajst-3607	64	6	and	and	CCONJ
ajst-3607	64	7	cooperation	cooperation	NOUN
ajst-3607	64	8	of	of	ADP
ajst-3607	64	9	multi	multi	NOUN
ajst-3607	64	10	-	-	NOUN
ajst-3607	64	11	disciplines	discipline	NOUN
ajst-3607	64	12	,	,	PUNCT
ajst-3607	64	13	multi	multi	NOUN
ajst-3607	64	14	-	-	NOUN
ajst-3607	64	15	types	type	NOUN
ajst-3607	64	16	of	of	ADP
ajst-3607	64	17	work	work	NOUN
ajst-3607	64	18	to	to	PART
ajst-3607	64	19	finally	finally	ADV
ajst-3607	64	20	achieve	achieve	VERB
ajst-3607	64	21	high	high	ADJ
ajst-3607	64	22	-	-	PUNCT
ajst-3607	64	23	quality	quality	NOUN
ajst-3607	64	24	development	development	NOUN
ajst-3607	64	25	.	.	PUNCT
ajst-3607	65	1	the	the	DET
ajst-3607	65	2	forgetting	forget	VERB
ajst-3607	65	3	gate	gate	NOUN
ajst-3607	65	4	determines	determine	NOUN
ajst-3607	65	5	which	which	DET
ajst-3607	65	6	information	information	NOUN
ajst-3607	65	7	the	the	DET
ajst-3607	65	8	model	model	NOUN
ajst-3607	65	9	will	will	AUX
ajst-3607	65	10	delete	delete	VERB
ajst-3607	65	11	from	from	ADP
ajst-3607	65	12	which	which	DET
ajst-3607	65	13	state	state	NOUN
ajst-3607	65	14	,	,	PUNCT
ajst-3607	65	15	and	and	CCONJ
ajst-3607	65	16	the	the	DET
ajst-3607	65	17	calculation	calculation	NOUN
ajst-3607	65	18	formula	formula	NOUN
ajst-3607	65	19	of	of	ADP
ajst-3607	65	20	forgetting	forget	VERB
ajst-3607	65	21	gate	gate	NOUN
ajst-3607	65	22	is	be	AUX
ajst-3607	65	23	:	:	PUNCT
ajst-3607	65	24	the	the	DET
ajst-3607	65	25	input	input	NOUN
ajst-3607	65	26	gate	gate	NOUN
ajst-3607	65	27	mainly	mainly	ADV
ajst-3607	65	28	finds	find	VERB
ajst-3607	65	29	the	the	DET
ajst-3607	65	30	cell	cell	NOUN
ajst-3607	65	31	state	state	NOUN
ajst-3607	65	32	to	to	PART
ajst-3607	65	33	be	be	AUX
ajst-3607	65	34	updated	update	VERB
ajst-3607	65	35	and	and	CCONJ
ajst-3607	65	36	updates	update	VERB
ajst-3607	65	37	the	the	DET
ajst-3607	65	38	information	information	NOUN
ajst-3607	65	39	to	to	ADP
ajst-3607	65	40	the	the	DET
ajst-3607	65	41	cell	cell	NOUN
ajst-3607	65	42	state	state	NOUN
ajst-3607	65	43	.	.	PUNCT
ajst-3607	66	1	the	the	DET
ajst-3607	66	2	calculation	calculation	NOUN
ajst-3607	66	3	formula	formula	NOUN
ajst-3607	66	4	of	of	ADP
ajst-3607	66	5	the	the	DET
ajst-3607	66	6	input	input	NOUN
ajst-3607	66	7	gate	gate	NOUN
ajst-3607	66	8	is	be	AUX
ajst-3607	66	9	:	:	PUNCT
ajst-3607	66	10	the	the	DET
ajst-3607	66	11	calculation	calculation	NOUN
ajst-3607	66	12	formula	formula	NOUN
ajst-3607	66	13	of	of	ADP
ajst-3607	66	14	timely	timely	ADJ
ajst-3607	66	15	cell	cell	NOUN
ajst-3607	66	16	status	status	NOUN
ajst-3607	66	17	pagenumber	pagenumber	NOUN
ajst-3607	66	18	_	_	PUNCT
ajst-3607	67	1	ebook	ebook	NOUN
ajst-3607	67	2	=	=	NOUN
ajst-3607	67	3	185	185	NUM
ajst-3607	67	4	and	and	CCONJ
ajst-3607	67	5	pagenumber	pagenumber	VERB
ajst-3607	67	6	_	_	DET
ajst-3607	67	7	book	book	NOUN
ajst-3607	67	8	=	=	NOUN
ajst-3607	67	9	177	177	NUM
ajst-3607	67	10	is	be	AUX
ajst-3607	67	11	:	:	PUNCT
ajst-3607	67	12	the	the	DET
ajst-3607	67	13	information	information	NOUN
ajst-3607	67	14	stored	store	VERB
ajst-3607	67	15	in	in	ADP
ajst-3607	67	16	the	the	DET
ajst-3607	67	17	long	long	ADJ
ajst-3607	67	18	-	-	PUNCT
ajst-3607	67	19	term	term	NOUN
ajst-3607	67	20	memory	memory	NOUN
ajst-3607	67	21	cell	cell	NOUN
ajst-3607	67	22	state	state	NOUN
ajst-3607	67	23	ct	ct	PROPN
ajst-3607	67	24	is	be	AUX
ajst-3607	67	25	:	:	PUNCT
ajst-3607	67	26	the	the	DET
ajst-3607	67	27	output	output	NOUN
ajst-3607	67	28	gate	gate	NOUN
ajst-3607	67	29	mainly	mainly	ADV
ajst-3607	67	30	determines	determine	VERB
ajst-3607	67	31	which	which	DET
ajst-3607	67	32	part	part	NOUN
ajst-3607	67	33	needs	need	VERB
ajst-3607	67	34	to	to	PART
ajst-3607	67	35	be	be	AUX
ajst-3607	67	36	152	152	NUM
ajst-3607	67	37	output	output	NOUN
ajst-3607	67	38	,	,	PUNCT
ajst-3607	67	39	and	and	CCONJ
ajst-3607	67	40	the	the	DET
ajst-3607	67	41	calculation	calculation	NOUN
ajst-3607	67	42	formula	formula	NOUN
ajst-3607	67	43	of	of	ADP
ajst-3607	67	44	the	the	DET
ajst-3607	67	45	output	output	NOUN
ajst-3607	67	46	gate	gate	NOUN
ajst-3607	67	47	is	be	AUX
ajst-3607	67	48	:	:	PUNCT
ajst-3607	67	49	x	x	X
ajst-3607	67	50	the	the	DET
ajst-3607	67	51	output	output	NOUN
ajst-3607	67	52	value	value	NOUN
ajst-3607	67	53	of	of	ADP
ajst-3607	67	54	lstm	lstm	NOUN
ajst-3607	67	55	is	be	AUX
ajst-3607	67	56	:	:	PUNCT
ajst-3607	67	57	where	where	SCONJ
ajst-3607	67	58	:	:	PUNCT
ajst-3607	67	59	σ	σ	NOUN
ajst-3607	67	60	is	be	AUX
ajst-3607	67	61	sigmoid	sigmoid	NOUN
ajst-3607	67	62	function	function	NOUN
ajst-3607	67	63	;	;	PUNCT
ajst-3607	67	64	said	say	VERB
ajst-3607	67	65	xt	xt	PROPN
ajst-3607	67	66	t	t	PROPN
ajst-3607	67	67	time	time	NOUN
ajst-3607	67	68	input	input	NOUN
ajst-3607	67	69	characteristic	characteristic	ADJ
ajst-3607	67	70	matrix	matrix	NOUN
ajst-3607	67	71	;	;	PUNCT
ajst-3607	67	72	wi	wi	PROPN
ajst-3607	67	73	represents	represent	VERB
ajst-3607	67	74	the	the	DET
ajst-3607	67	75	input	input	NOUN
ajst-3607	67	76	layer	layer	NOUN
ajst-3607	67	77	to	to	ADP
ajst-3607	67	78	the	the	DET
ajst-3607	67	79	input	input	NOUN
ajst-3607	67	80	gate	gate	NOUN
ajst-3607	67	81	;	;	PUNCT
ajst-3607	67	82	wf	wf	PROPN
ajst-3607	67	83	means	mean	VERB
ajst-3607	67	84	forgetting	forget	VERB
ajst-3607	67	85	gate	gate	NOUN
ajst-3607	67	86	;	;	PUNCT
ajst-3607	67	87	wc	wc	PROPN
ajst-3607	67	88	represents	represent	VERB
ajst-3607	67	89	a	a	DET
ajst-3607	67	90	memory	memory	NOUN
ajst-3607	67	91	cell	cell	NOUN
ajst-3607	67	92	;	;	PUNCT
ajst-3607	67	93	wo	will	AUX
ajst-3607	67	94	represents	represent	VERB
ajst-3607	67	95	the	the	DET
ajst-3607	67	96	weight	weight	NOUN
ajst-3607	67	97	matrix	matrix	NOUN
ajst-3607	67	98	between	between	ADP
ajst-3607	67	99	output	output	NOUN
ajst-3607	67	100	gates	gate	NOUN
ajst-3607	67	101	;	;	PUNCT
ajst-3607	67	102	said	say	VERB
ajst-3607	67	103	bi	bi	PROPN
ajst-3607	67	104	input	input	NOUN
ajst-3607	67	105	gate	gate	NOUN
ajst-3607	67	106	;	;	PUNCT
ajst-3607	67	107	bf	bf	NOUN
ajst-3607	67	108	means	mean	VERB
ajst-3607	67	109	forgetting	forget	VERB
ajst-3607	67	110	gate	gate	NOUN
ajst-3607	67	111	;	;	PUNCT
ajst-3607	67	112	bc	bc	PROPN
ajst-3607	67	113	represents	represent	VERB
ajst-3607	67	114	a	a	DET
ajst-3607	67	115	storage	storage	NOUN
ajst-3607	67	116	unit	unit	NOUN
ajst-3607	67	117	(	(	PUNCT
ajst-3607	67	118	cell	cell	NOUN
ajst-3607	67	119	)	)	PUNCT
ajst-3607	67	120	;	;	PUNCT
ajst-3607	67	121	bo	bo	PROPN
ajst-3607	67	122	indicates	indicate	VERB
ajst-3607	67	123	the	the	DET
ajst-3607	67	124	offset	offset	NOUN
ajst-3607	67	125	value	value	NOUN
ajst-3607	67	126	of	of	ADP
ajst-3607	67	127	the	the	DET
ajst-3607	67	128	output	output	NOUN
ajst-3607	67	129	gate	gate	NOUN
ajst-3607	67	130	.	.	PUNCT
ajst-3607	68	1	the	the	DET
ajst-3607	68	2	model	model	NOUN
ajst-3607	68	3	sends	send	VERB
ajst-3607	68	4	the	the	DET
ajst-3607	68	5	output	output	NOUN
ajst-3607	68	6	result	result	NOUN
ajst-3607	68	7	of	of	ADP
ajst-3607	68	8	the	the	DET
ajst-3607	68	9	previous	previous	ADJ
ajst-3607	68	10	step	step	NOUN
ajst-3607	68	11	seinception_v3	seinception_v3	NOUN
ajst-3607	68	12	to	to	ADP
ajst-3607	68	13	the	the	DET
ajst-3607	68	14	lstm	lstm	ADJ
ajst-3607	68	15	unit	unit	NOUN
ajst-3607	68	16	of	of	ADP
ajst-3607	68	17	the	the	DET
ajst-3607	68	18	time	time	NOUN
ajst-3607	68	19	series	series	NOUN
ajst-3607	68	20	module	module	NOUN
ajst-3607	68	21	,	,	PUNCT
ajst-3607	68	22	and	and	CCONJ
ajst-3607	68	23	then	then	ADV
ajst-3607	68	24	averages	average	VERB
ajst-3607	68	25	the	the	DET
ajst-3607	68	26	probability	probability	NOUN
ajst-3607	68	27	distribution	distribution	NOUN
ajst-3607	68	28	of	of	ADP
ajst-3607	68	29	all	all	DET
ajst-3607	68	30	frames	frame	NOUN
ajst-3607	68	31	,	,	PUNCT
ajst-3607	68	32	and	and	CCONJ
ajst-3607	68	33	selects	select	VERB
ajst-3607	68	34	the	the	DET
ajst-3607	68	35	most	most	ADV
ajst-3607	68	36	possible	possible	ADJ
ajst-3607	68	37	label	label	NOUN
ajst-3607	68	38	,	,	PUNCT
ajst-3607	68	39	thus	thus	ADV
ajst-3607	68	40	completing	complete	VERB
ajst-3607	68	41	the	the	DET
ajst-3607	68	42	recognition	recognition	NOUN
ajst-3607	68	43	of	of	ADP
ajst-3607	68	44	human	human	ADJ
ajst-3607	68	45	motion	motion	NOUN
ajst-3607	68	46	in	in	ADP
ajst-3607	68	47	video	video	NOUN
ajst-3607	68	48	.	.	PUNCT
ajst-3607	69	1	3	3	X
ajst-3607	69	2	.	.	X
ajst-3607	69	3	experimental	experimental	ADJ
ajst-3607	69	4	structure	structure	NOUN
ajst-3607	69	5	3.1	3.1	NUM
ajst-3607	69	6	.	.	PUNCT
ajst-3607	70	1	algorithm	algorithm	NOUN
ajst-3607	70	2	framework	framework	NOUN
ajst-3607	70	3	the	the	DET
ajst-3607	70	4	algorithm	algorithm	NOUN
ajst-3607	70	5	framework	framework	NOUN
ajst-3607	70	6	is	be	AUX
ajst-3607	70	7	shown	show	VERB
ajst-3607	70	8	in	in	ADP
ajst-3607	70	9	figure	figure	NOUN
ajst-3607	70	10	4	4	NUM
ajst-3607	70	11	.	.	PUNCT
ajst-3607	70	12	figure	figure	VERB
ajst-3607	70	13	4	4	NUM
ajst-3607	70	14	.	.	PUNCT
ajst-3607	71	1	algorithm	algorithm	NOUN
ajst-3607	71	2	framework	framework	VERB
ajst-3607	71	3	the	the	DET
ajst-3607	71	4	processing	processing	NOUN
ajst-3607	71	5	flow	flow	NOUN
ajst-3607	71	6	of	of	ADP
ajst-3607	71	7	this	this	DET
ajst-3607	71	8	algorithm	algorithm	NOUN
ajst-3607	71	9	is	be	AUX
ajst-3607	71	10	as	as	SCONJ
ajst-3607	71	11	follows	follow	VERB
ajst-3607	71	12	:	:	PUNCT
ajst-3607	71	13	1	1	X
ajst-3607	71	14	)	)	PUNCT
ajst-3607	71	15	yolo_v4	yolo_v4	NOUN
ajst-3607	71	16	is	be	AUX
ajst-3607	71	17	used	use	VERB
ajst-3607	71	18	to	to	PART
ajst-3607	71	19	detect	detect	VERB
ajst-3607	71	20	the	the	DET
ajst-3607	71	21	human	human	ADJ
ajst-3607	71	22	target	target	NOUN
ajst-3607	71	23	in	in	ADP
ajst-3607	71	24	the	the	DET
ajst-3607	71	25	video	video	NOUN
ajst-3607	71	26	,	,	PUNCT
ajst-3607	71	27	and	and	CCONJ
ajst-3607	71	28	the	the	DET
ajst-3607	71	29	detected	detect	VERB
ajst-3607	71	30	video	video	NOUN
ajst-3607	71	31	is	be	AUX
ajst-3607	71	32	processed	process	VERB
ajst-3607	71	33	at	at	ADP
ajst-3607	71	34	frame	frame	NOUN
ajst-3607	71	35	level	level	NOUN
ajst-3607	71	36	after	after	SCONJ
ajst-3607	71	37	the	the	DET
ajst-3607	71	38	human	human	ADJ
ajst-3607	71	39	target	target	NOUN
ajst-3607	71	40	is	be	AUX
ajst-3607	71	41	detected	detect	VERB
ajst-3607	71	42	.	.	PUNCT
ajst-3607	72	1	2	2	X
ajst-3607	72	2	)	)	PUNCT
ajst-3607	72	3	use	use	VERB
ajst-3607	72	4	se	se	NOUN
ajst-3607	72	5	-	-	NOUN
ajst-3607	72	6	incidence	incidence	ADJ
ajst-3607	72	7	_	_	PUNCT
ajst-3607	73	1	v3	v3	PROPN
ajst-3607	73	2	for	for	ADP
ajst-3607	73	3	feature	feature	NOUN
ajst-3607	73	4	extraction	extraction	NOUN
ajst-3607	73	5	.	.	PUNCT
ajst-3607	74	1	inception_v3	inception_v3	ADJ
ajst-3607	74	2	network	network	NOUN
ajst-3607	74	3	can	can	AUX
ajst-3607	74	4	extract	extract	VERB
ajst-3607	74	5	features	feature	NOUN
ajst-3607	74	6	of	of	ADP
ajst-3607	74	7	different	different	ADJ
ajst-3607	74	8	scales	scale	NOUN
ajst-3607	74	9	,	,	PUNCT
ajst-3607	74	10	and	and	CCONJ
ajst-3607	74	11	channel	channel	NOUN
ajst-3607	74	12	attention	attention	NOUN
ajst-3607	74	13	mechanism	mechanism	NOUN
ajst-3607	74	14	(	(	PUNCT
ajst-3607	74	15	se	se	ADJ
ajst-3607	74	16	-	-	NOUN
ajst-3607	74	17	net	net	NOUN
ajst-3607	74	18	)	)	PUNCT
ajst-3607	74	19	can	can	AUX
ajst-3607	74	20	assign	assign	VERB
ajst-3607	74	21	weights	weight	NOUN
ajst-3607	74	22	to	to	ADP
ajst-3607	74	23	different	different	ADJ
ajst-3607	74	24	convolution	convolution	NOUN
ajst-3607	74	25	kernels	kernel	NOUN
ajst-3607	74	26	to	to	PART
ajst-3607	74	27	make	make	VERB
ajst-3607	74	28	them	they	PRON
ajst-3607	74	29	pay	pay	VERB
ajst-3607	74	30	more	more	ADJ
ajst-3607	74	31	attention	attention	NOUN
ajst-3607	74	32	to	to	ADP
ajst-3607	74	33	the	the	DET
ajst-3607	74	34	main	main	ADJ
ajst-3607	74	35	features	feature	NOUN
ajst-3607	74	36	.	.	PUNCT
ajst-3607	75	1	3	3	X
ajst-3607	75	2	)	)	PUNCT
ajst-3607	75	3	through	through	ADP
ajst-3607	75	4	lstm	lstm	NOUN
ajst-3607	75	5	for	for	ADP
ajst-3607	75	6	prediction	prediction	NOUN
ajst-3607	75	7	and	and	CCONJ
ajst-3607	75	8	recognition	recognition	NOUN
ajst-3607	75	9	,	,	PUNCT
ajst-3607	75	10	test	test	NOUN
ajst-3607	75	11	action	action	NOUN
ajst-3607	75	12	video	video	NOUN
ajst-3607	75	13	frames	frame	NOUN
ajst-3607	75	14	of	of	ADP
ajst-3607	75	15	different	different	ADJ
ajst-3607	75	16	types	type	NOUN
ajst-3607	75	17	of	of	ADP
ajst-3607	75	18	interactive	interactive	ADJ
ajst-3607	75	19	actions	action	NOUN
ajst-3607	75	20	are	be	AUX
ajst-3607	75	21	sent	send	VERB
ajst-3607	75	22	into	into	ADP
ajst-3607	75	23	the	the	DET
ajst-3607	75	24	trained	train	VERB
ajst-3607	75	25	test	test	NOUN
ajst-3607	75	26	models	model	NOUN
ajst-3607	75	27	respectively	respectively	ADV
ajst-3607	75	28	,	,	PUNCT
ajst-3607	75	29	and	and	CCONJ
ajst-3607	75	30	the	the	DET
ajst-3607	75	31	probability	probability	NOUN
ajst-3607	75	32	value	value	NOUN
ajst-3607	75	33	of	of	ADP
ajst-3607	75	34	each	each	DET
ajst-3607	75	35	type	type	NOUN
ajst-3607	75	36	of	of	ADP
ajst-3607	75	37	test	test	NOUN
ajst-3607	75	38	action	action	NOUN
ajst-3607	75	39	is	be	AUX
ajst-3607	75	40	obtained	obtain	VERB
ajst-3607	75	41	,	,	PUNCT
ajst-3607	75	42	so	so	SCONJ
ajst-3607	75	43	as	as	SCONJ
ajst-3607	75	44	to	to	PART
ajst-3607	75	45	realize	realize	VERB
ajst-3607	75	46	the	the	DET
ajst-3607	75	47	prediction	prediction	NOUN
ajst-3607	75	48	and	and	CCONJ
ajst-3607	75	49	recognition	recognition	NOUN
ajst-3607	75	50	of	of	ADP
ajst-3607	75	51	two	two	NUM
ajst-3607	75	52	-	-	PUNCT
ajst-3607	75	53	person	person	NOUN
ajst-3607	75	54	interactive	interactive	ADJ
ajst-3607	75	55	behavior	behavior	NOUN
ajst-3607	75	56	.	.	PUNCT
ajst-3607	76	1	3.2	3.2	NUM
ajst-3607	76	2	.	.	PUNCT
ajst-3607	77	1	network	network	NOUN
ajst-3607	77	2	training	train	VERB
ajst-3607	77	3	the	the	DET
ajst-3607	77	4	network	network	NOUN
ajst-3607	77	5	of	of	ADP
ajst-3607	77	6	this	this	DET
ajst-3607	77	7	paper	paper	NOUN
ajst-3607	77	8	is	be	AUX
ajst-3607	77	9	built	build	VERB
ajst-3607	77	10	on	on	ADP
ajst-3607	77	11	tensorflow	tensorflow	NOUN
ajst-3607	77	12	.	.	PUNCT
ajst-3607	78	1	hardware	hardware	NOUN
ajst-3607	78	2	:	:	PUNCT
ajst-3607	78	3	the	the	DET
ajst-3607	78	4	processor	processor	NOUN
ajst-3607	78	5	is	be	AUX
ajst-3607	78	6	i9	i9	NOUN
ajst-3607	78	7	-	-	PUNCT
ajst-3607	78	8	9700k	9700k	NUM
ajst-3607	78	9	；	；	NOUN
ajst-3607	78	10	;	;	PUNCT
ajst-3607	78	11	gpu	gpu	X
ajst-3607	78	12	is	be	AUX
ajst-3607	78	13	nivda	nivda	NOUN
ajst-3607	78	14	-	-	PUNCT
ajst-3607	78	15	rtx2080	rtx2080	NOUN
ajst-3607	78	16	-	-	PUNCT
ajst-3607	78	17	ti	ti	NOUN
ajst-3607	78	18	;	;	PUNCT
ajst-3607	78	19	the	the	DET
ajst-3607	78	20	gpu	gpu	NOUN
ajst-3607	78	21	memory	memory	NOUN
ajst-3607	78	22	size	size	NOUN
ajst-3607	78	23	is	be	AUX
ajst-3607	78	24	11	11	NUM
ajst-3607	78	25	gb	gb	NOUN
ajst-3607	78	26	.	.	PUNCT
ajst-3607	79	1	software	software	NOUN
ajst-3607	79	2	:	:	PUNCT
ajst-3607	79	3	the	the	DET
ajst-3607	79	4	operating	operate	VERB
ajst-3607	79	5	system	system	NOUN
ajst-3607	79	6	is	be	AUX
ajst-3607	79	7	ubuntu	ubuntu	ADJ
ajst-3607	79	8	18.04	18.04	NUM
ajst-3607	79	9	；	；	NOUN
ajst-3607	79	10	;	;	PUNCT
ajst-3607	79	11	the	the	DET
ajst-3607	79	12	language	language	NOUN
ajst-3607	79	13	is	be	AUX
ajst-3607	79	14	python	python	NOUN
ajst-3607	79	15	.	.	PUNCT
ajst-3607	80	1	the	the	DET
ajst-3607	80	2	yolo_v4	yolo_v4	NOUN
ajst-3607	80	3	network	network	NOUN
ajst-3607	80	4	design	design	NOUN
ajst-3607	80	5	uses	use	VERB
ajst-3607	80	6	mish	mish	PROPN
ajst-3607	80	7	as	as	ADP
ajst-3607	80	8	an	an	DET
ajst-3607	80	9	active	active	ADJ
ajst-3607	80	10	gradient	gradient	NOUN
ajst-3607	80	11	function	function	NOUN
ajst-3607	80	12	,	,	PUNCT
ajst-3607	80	13	with	with	ADP
ajst-3607	80	14	the	the	DET
ajst-3607	80	15	initial	initial	ADJ
ajst-3607	80	16	random	random	ADJ
ajst-3607	80	17	learning	learning	NOUN
ajst-3607	80	18	rate	rate	NOUN
ajst-3607	80	19	of	of	ADP
ajst-3607	80	20	0.1	0.1	NUM
ajst-3607	80	21	,	,	PUNCT
ajst-3607	80	22	weight	weight	NOUN
ajst-3607	80	23	coefficient	coefficient	NOUN
ajst-3607	80	24	of	of	ADP
ajst-3607	80	25	0.005	0.005	NUM
ajst-3607	80	26	and	and	CCONJ
ajst-3607	80	27	momentum	momentum	NOUN
ajst-3607	80	28	coefficient	coefficient	NOUN
ajst-3607	80	29	of	of	ADP
ajst-3607	80	30	0.9	0.9	NUM
ajst-3607	80	31	.	.	PUNCT
ajst-3607	81	1	se	se	ADJ
ajst-3607	81	2	-	-	NOUN
ajst-3607	81	3	incidence	incidence	ADJ
ajst-3607	81	4	_	_	PUNCT
ajst-3607	81	5	v3	v3	PROPN
ajst-3607	81	6	network	network	NOUN
ajst-3607	81	7	adopts	adopt	VERB
ajst-3607	81	8	random	random	ADJ
ajst-3607	81	9	gradient	gradient	ADJ
ajst-3607	81	10	descent	descent	NOUN
ajst-3607	81	11	method	method	NOUN
ajst-3607	81	12	(	(	PUNCT
ajst-3607	81	13	sgd	sgd	PROPN
ajst-3607	81	14	)	)	PUNCT
ajst-3607	81	15	with	with	ADP
ajst-3607	81	16	high	high	ADJ
ajst-3607	81	17	learning	learning	NOUN
ajst-3607	81	18	rate	rate	NOUN
ajst-3607	81	19	,	,	PUNCT
ajst-3607	81	20	the	the	DET
ajst-3607	81	21	batch_size	batch_size	NOUN
ajst-3607	81	22	is	be	AUX
ajst-3607	81	23	set	set	VERB
ajst-3607	81	24	to	to	ADP
ajst-3607	81	25	16	16	NUM
ajst-3607	81	26	,	,	PUNCT
ajst-3607	81	27	and	and	CCONJ
ajst-3607	81	28	the	the	DET
ajst-3607	81	29	activation	activation	NOUN
ajst-3607	81	30	function	function	NOUN
ajst-3607	81	31	is	be	AUX
ajst-3607	81	32	relu	relu	NOUN
ajst-3607	81	33	.	.	PUNCT
ajst-3607	82	1	in	in	ADP
ajst-3607	82	2	order	order	NOUN
ajst-3607	82	3	to	to	PART
ajst-3607	82	4	avoid	avoid	VERB
ajst-3607	82	5	over	over	ADP
ajst-3607	82	6	-	-	PUNCT
ajst-3607	82	7	fitting	fitting	NOUN
ajst-3607	82	8	in	in	ADP
ajst-3607	82	9	training	training	NOUN
ajst-3607	82	10	,	,	PUNCT
ajst-3607	82	11	a	a	DET
ajst-3607	82	12	dropout	dropout	NOUN
ajst-3607	82	13	with	with	ADP
ajst-3607	82	14	a	a	DET
ajst-3607	82	15	threshold	threshold	NOUN
ajst-3607	82	16	of	of	ADP
ajst-3607	82	17	0.5	0.5	NUM
ajst-3607	82	18	is	be	AUX
ajst-3607	82	19	set	set	VERB
ajst-3607	82	20	,	,	PUNCT
ajst-3607	82	21	the	the	DET
ajst-3607	82	22	initial	initial	ADJ
ajst-3607	82	23	learning	learning	NOUN
ajst-3607	82	24	rate	rate	NOUN
ajst-3607	82	25	lr=0.001	lr=0.001	NOUN
ajst-3607	82	26	,	,	PUNCT
ajst-3607	82	27	and	and	CCONJ
ajst-3607	82	28	the	the	DET
ajst-3607	82	29	number	number	NOUN
ajst-3607	82	30	of	of	ADP
ajst-3607	82	31	iterations	iteration	NOUN
ajst-3607	82	32	nb_epoch=50	nb_epoch=50	PROPN
ajst-3607	82	33	.	.	PUNCT
ajst-3607	82	34	3.3	3.3	NUM
ajst-3607	82	35	.	.	PUNCT
ajst-3607	83	1	data	datum	NOUN
ajst-3607	83	2	set	set	VERB
ajst-3607	83	3	a	a	DET
ajst-3607	83	4	channel	channel	NOUN
ajst-3607	83	5	attention	attention	NOUN
ajst-3607	83	6	model	model	NOUN
ajst-3607	83	7	se	se	NOUN
ajst-3607	83	8	-	-	NOUN
ajst-3607	83	9	perception	perception	NOUN
ajst-3607	83	10	_	_	PROPN
ajst-3607	83	11	v3	v3	PROPN
ajst-3607	83	12	based	base	VERB
ajst-3607	83	13	on	on	ADP
ajst-3607	83	14	perception	perception	NOUN
ajst-3607	83	15	_	_	PUNCT
ajst-3607	84	1	v3	v3	PROPN
ajst-3607	84	2	is	be	AUX
ajst-3607	84	3	proposed	propose	VERB
ajst-3607	84	4	,	,	PUNCT
ajst-3607	84	5	which	which	PRON
ajst-3607	84	6	is	be	AUX
ajst-3607	84	7	used	use	VERB
ajst-3607	84	8	for	for	ADP
ajst-3607	84	9	behavior	behavior	NOUN
ajst-3607	84	10	recognition	recognition	NOUN
ajst-3607	84	11	algorithm	algorithm	NOUN
ajst-3607	84	12	of	of	ADP
ajst-3607	84	13	two	two	NUM
ajst-3607	84	14	-	-	PUNCT
ajst-3607	84	15	person	person	NOUN
ajst-3607	84	16	interaction	interaction	NOUN
ajst-3607	84	17	.	.	PUNCT
ajst-3607	85	1	in	in	ADP
ajst-3607	85	2	order	order	NOUN
ajst-3607	85	3	to	to	PART
ajst-3607	85	4	prove	prove	VERB
ajst-3607	85	5	the	the	DET
ajst-3607	85	6	feasibility	feasibility	NOUN
ajst-3607	85	7	of	of	ADP
ajst-3607	85	8	the	the	DET
ajst-3607	85	9	algorithm	algorithm	NOUN
ajst-3607	85	10	,	,	PUNCT
ajst-3607	85	11	it	it	PRON
ajst-3607	85	12	is	be	AUX
ajst-3607	85	13	tested	test	VERB
ajst-3607	85	14	on	on	ADP
ajst-3607	85	15	the	the	DET
ajst-3607	85	16	published	publish	VERB
ajst-3607	85	17	ut	ut	NOUN
ajst-3607	85	18	-	-	PUNCT
ajst-3607	85	19	interaction	interaction	NOUN
ajst-3607	85	20	datebase	datebase	NOUN
ajst-3607	85	21	database	database	NOUN
ajst-3607	85	22	[	[	X
ajst-3607	85	23	13	13	NUM
ajst-3607	85	24	]	]	PUNCT
ajst-3607	85	25	segmented_set2	segmented_set2	PROPN
ajst-3607	85	26	data	datum	NOUN
ajst-3607	85	27	set	set	VERB
ajst-3607	85	28	.	.	PUNCT
ajst-3607	86	1	this	this	DET
ajst-3607	86	2	data	datum	NOUN
ajst-3607	86	3	set	set	VERB
ajst-3607	86	4	includes	include	VERB
ajst-3607	86	5	six	six	NUM
ajst-3607	86	6	kinds	kind	NOUN
ajst-3607	86	7	of	of	ADP
ajst-3607	86	8	actions	action	NOUN
ajst-3607	86	9	:	:	PUNCT
ajst-3607	86	10	shaking	shake	VERB
ajst-3607	86	11	,	,	PUNCT
ajst-3607	86	12	hugging	hug	VERB
ajst-3607	86	13	,	,	PUNCT
ajst-3607	86	14	pointing	pointing	NOUN
ajst-3607	86	15	,	,	PUNCT
ajst-3607	86	16	kicking	kicking	NOUN
ajst-3607	86	17	,	,	PUNCT
ajst-3607	86	18	punching	punching	NOUN
ajst-3607	86	19	and	and	CCONJ
ajst-3607	86	20	pushing	push	VERB
ajst-3607	86	21	.	.	PUNCT
ajst-3607	87	1	in	in	ADP
ajst-3607	87	2	addition	addition	NOUN
ajst-3607	87	3	,	,	PUNCT
ajst-3607	87	4	each	each	DET
ajst-3607	87	5	type	type	NOUN
ajst-3607	87	6	of	of	ADP
ajst-3607	87	7	action	action	NOUN
ajst-3607	87	8	in	in	ADP
ajst-3607	87	9	this	this	DET
ajst-3607	87	10	database	database	NOUN
ajst-3607	87	11	is	be	AUX
ajst-3607	87	12	completed	complete	VERB
ajst-3607	87	13	by	by	ADP
ajst-3607	87	14	different	different	ADJ
ajst-3607	87	15	operators	operator	NOUN
ajst-3607	87	16	,	,	PUNCT
ajst-3607	87	17	and	and	CCONJ
ajst-3607	87	18	there	there	PRON
ajst-3607	87	19	are	be	VERB
ajst-3607	87	20	60	60	NUM
ajst-3607	87	21	groups	group	NOUN
ajst-3607	87	22	of	of	ADP
ajst-3607	87	23	interactive	interactive	ADJ
ajst-3607	87	24	actions	action	NOUN
ajst-3607	87	25	.	.	PUNCT
ajst-3607	88	1	because	because	SCONJ
ajst-3607	88	2	there	there	PRON
ajst-3607	88	3	is	be	VERB
ajst-3607	88	4	no	no	DET
ajst-3607	88	5	periodic	periodic	ADJ
ajst-3607	88	6	rule	rule	NOUN
ajst-3607	88	7	among	among	ADP
ajst-3607	88	8	all	all	DET
ajst-3607	88	9	kinds	kind	NOUN
ajst-3607	88	10	of	of	ADP
ajst-3607	88	11	actions	action	NOUN
ajst-3607	88	12	in	in	ADP
ajst-3607	88	13	the	the	DET
ajst-3607	88	14	database	database	NOUN
ajst-3607	88	15	,	,	PUNCT
ajst-3607	88	16	and	and	CCONJ
ajst-3607	88	17	there	there	PRON
ajst-3607	88	18	are	be	VERB
ajst-3607	88	19	many	many	ADJ
ajst-3607	88	20	actions	action	NOUN
ajst-3607	88	21	with	with	ADP
ajst-3607	88	22	high	high	ADJ
ajst-3607	88	23	similarity	similarity	NOUN
ajst-3607	88	24	among	among	ADP
ajst-3607	88	25	the	the	DET
ajst-3607	88	26	types	type	NOUN
ajst-3607	88	27	of	of	ADP
ajst-3607	88	28	actions	action	NOUN
ajst-3607	88	29	in	in	ADP
ajst-3607	88	30	each	each	DET
ajst-3607	88	31	database	database	NOUN
ajst-3607	88	32	,	,	PUNCT
ajst-3607	88	33	it	it	PRON
ajst-3607	88	34	is	be	AUX
ajst-3607	88	35	more	more	ADV
ajst-3607	88	36	testable	testable	ADJ
ajst-3607	88	37	to	to	PART
ajst-3607	88	38	recognize	recognize	VERB
ajst-3607	88	39	interactive	interactive	ADJ
ajst-3607	88	40	behaviors	behavior	NOUN
ajst-3607	88	41	.	.	PUNCT
ajst-3607	89	1	the	the	DET
ajst-3607	89	2	data	data	NOUN
ajst-3607	89	3	concentration	concentration	NOUN
ajst-3607	89	4	action	action	NOUN
ajst-3607	89	5	is	be	AUX
ajst-3607	89	6	shown	show	VERB
ajst-3607	89	7	in	in	ADP
ajst-3607	89	8	figure	figure	NOUN
ajst-3607	89	9	5	5	NUM
ajst-3607	89	10	.	.	PUNCT
ajst-3607	90	1	figure	figure	NOUN
ajst-3607	90	2	5	5	NUM
ajst-3607	90	3	.	.	PUNCT
ajst-3607	90	4	ut_interaction	ut_interaction	NOUN
ajst-3607	90	5	data	datum	NOUN
ajst-3607	90	6	set	set	VERB
ajst-3607	90	7	3.4	3.4	NUM
ajst-3607	90	8	.	.	PUNCT
ajst-3607	91	1	comparison	comparison	NOUN
ajst-3607	91	2	and	and	CCONJ
ajst-3607	91	3	analysis	analysis	NOUN
ajst-3607	91	4	of	of	ADP
ajst-3607	91	5	experimental	experimental	ADJ
ajst-3607	91	6	results	result	NOUN
ajst-3607	91	7	comparison	comparison	NOUN
ajst-3607	91	8	of	of	ADP
ajst-3607	91	9	accuracy	accuracy	NOUN
ajst-3607	91	10	results	result	NOUN
ajst-3607	91	11	of	of	ADP
ajst-3607	91	12	different	different	ADJ
ajst-3607	91	13	network	network	NOUN
ajst-3607	91	14	models	model	NOUN
ajst-3607	91	15	is	be	AUX
ajst-3607	91	16	shown	show	VERB
ajst-3607	91	17	in	in	ADP
ajst-3607	91	18	table	table	NOUN
ajst-3607	91	19	1	1	NUM
ajst-3607	91	20	.	.	PUNCT
ajst-3607	91	21	table	table	NOUN
ajst-3607	91	22	1	1	NUM
ajst-3607	91	23	.	.	PUNCT
ajst-3607	91	24	comparison	comparison	NOUN
ajst-3607	91	25	of	of	ADP
ajst-3607	91	26	accuracy	accuracy	NOUN
ajst-3607	91	27	results	result	NOUN
ajst-3607	91	28	of	of	ADP
ajst-3607	91	29	different	different	ADJ
ajst-3607	91	30	network	network	NOUN
ajst-3607	91	31	models	model	NOUN
ajst-3607	91	32	network	network	NOUN
ajst-3607	91	33	model	model	NOUN
ajst-3607	91	34	accuracy(%	accuracy(%	PROPN
ajst-3607	91	35	)	)	PUNCT
ajst-3607	92	1	inception_v3+lstm	inception_v3+lstm	VERB
ajst-3607	92	2	82.10	82.10	NUM
ajst-3607	92	3	se_inception_v3+lstm	se_inception_v3+lstm	PROPN
ajst-3607	92	4	84.20	84.20	NUM
ajst-3607	92	5	yolo_v4+seinception_v3+lstm	yolo_v4+seinception_v3+lstm	NOUN
ajst-3607	92	6	85.10	85.10	NUM
ajst-3607	92	7	as	as	SCONJ
ajst-3607	92	8	can	can	AUX
ajst-3607	92	9	be	be	AUX
ajst-3607	92	10	seen	see	VERB
ajst-3607	92	11	from	from	ADP
ajst-3607	92	12	table	table	NOUN
ajst-3607	92	13	1	1	NUM
ajst-3607	92	14	,	,	PUNCT
ajst-3607	92	15	the	the	DET
ajst-3607	92	16	recognition	recognition	NOUN
ajst-3607	92	17	rate	rate	NOUN
ajst-3607	92	18	has	have	AUX
ajst-3607	92	19	been	be	AUX
ajst-3607	92	20	improved	improve	VERB
ajst-3607	92	21	by	by	ADP
ajst-3607	92	22	adding	add	VERB
ajst-3607	92	23	the	the	DET
ajst-3607	92	24	channel	channel	NOUN
ajst-3607	92	25	attention	attention	NOUN
ajst-3607	92	26	module	module	NOUN
ajst-3607	92	27	(	(	PUNCT
ajst-3607	92	28	se	se	NOUN
ajst-3607	92	29	-	-	NOUN
ajst-3607	92	30	net	net	ADJ
ajst-3607	92	31	)	)	PUNCT
ajst-3607	92	32	,	,	PUNCT
ajst-3607	92	33	and	and	CCONJ
ajst-3607	92	34	the	the	DET
ajst-3607	92	35	recognition	recognition	NOUN
ajst-3607	92	36	accuracy	accuracy	NOUN
ajst-3607	92	37	of	of	ADP
ajst-3607	92	38	two	two	NUM
ajst-3607	92	39	-	-	PUNCT
ajst-3607	92	40	person	person	NOUN
ajst-3607	92	41	interaction	interaction	NOUN
ajst-3607	92	42	behavior	behavior	NOUN
ajst-3607	92	43	is	be	AUX
ajst-3607	92	44	84.20	84.20	NUM
ajst-3607	92	45	%	%	NOUN
ajst-3607	92	46	;	;	PUNCT
ajst-3607	92	47	the	the	DET
ajst-3607	92	48	recognition	recognition	NOUN
ajst-3607	92	49	rate	rate	NOUN
ajst-3607	92	50	of	of	ADP
ajst-3607	92	51	human	human	ADJ
ajst-3607	92	52	body	body	NOUN
ajst-3607	92	53	detected	detect	VERB
ajst-3607	92	54	by	by	ADP
ajst-3607	92	55	yolo_v4	yolo_v4	NOUN
ajst-3607	92	56	and	and	CCONJ
ajst-3607	92	57	then	then	ADV
ajst-3607	92	58	sent	send	VERB
ajst-3607	92	59	to	to	ADP
ajst-3607	92	60	se	se	VERB
ajst-3607	92	61	-	-	VERB
ajst-3607	92	62	induction	induction	NOUN
ajst-3607	92	63	_	_	PUNCT
ajst-3607	92	64	v3	v3	PROPN
ajst-3607	92	65	and	and	CCONJ
ajst-3607	92	66	lstm	lstm	PROPN
ajst-3607	92	67	network	network	NOUN
ajst-3607	92	68	is	be	AUX
ajst-3607	92	69	higher	high	ADJ
ajst-3607	92	70	than	than	ADP
ajst-3607	92	71	the	the	DET
ajst-3607	92	72	original	original	ADJ
ajst-3607	92	73	recognition	recognition	NOUN
ajst-3607	92	74	rate	rate	NOUN
ajst-3607	92	75	,	,	PUNCT
ajst-3607	92	76	reaching	reach	VERB
ajst-3607	92	77	85.10	85.10	NUM
ajst-3607	92	78	%	%	NOUN
ajst-3607	92	79	.	.	PUNCT
ajst-3607	93	1	it	it	PRON
ajst-3607	93	2	shows	show	VERB
ajst-3607	93	3	that	that	SCONJ
ajst-3607	93	4	the	the	DET
ajst-3607	93	5	se	se	ADJ
ajst-3607	93	6	-	-	ADJ
ajst-3607	93	7	inception_v3	inception_v3	ADJ
ajst-3607	93	8	proposed	propose	VERB
ajst-3607	93	9	in	in	ADP
ajst-3607	93	10	this	this	DET
ajst-3607	93	11	paper	paper	NOUN
ajst-3607	93	12	has	have	VERB
ajst-3607	93	13	certain	certain	ADJ
ajst-3607	93	14	reliability	reliability	NOUN
ajst-3607	93	15	in	in	ADP
ajst-3607	93	16	improving	improve	VERB
ajst-3607	93	17	the	the	DET
ajst-3607	93	18	accuracy	accuracy	NOUN
ajst-3607	93	19	of	of	ADP
ajst-3607	93	20	two	two	NUM
ajst-3607	93	21	-	-	PUNCT
ajst-3607	93	22	person	person	NOUN
ajst-3607	93	23	interaction	interaction	NOUN
ajst-3607	93	24	behavior	behavior	NOUN
ajst-3607	93	25	recognition	recognition	NOUN
ajst-3607	93	26	,	,	PUNCT
ajst-3607	93	27	and	and	CCONJ
ajst-3607	93	28	it	it	PRON
ajst-3607	93	29	also	also	ADV
ajst-3607	93	30	proves	prove	VERB
ajst-3607	93	31	that	that	SCONJ
ajst-3607	93	32	it	it	PRON
ajst-3607	93	33	is	be	AUX
ajst-3607	93	34	feasible	feasible	ADJ
ajst-3607	93	35	to	to	PART
ajst-3607	93	36	improve	improve	VERB
ajst-3607	93	37	the	the	DET
ajst-3607	93	38	accuracy	accuracy	NOUN
ajst-3607	93	39	of	of	ADP
ajst-3607	93	40	later	later	ADJ
ajst-3607	93	41	behavior	behavior	NOUN
ajst-3607	93	42	recognition	recognition	NOUN
ajst-3607	93	43	through	through	ADP
ajst-3607	93	44	yolo_v4	yolo_v4	NOUN
ajst-3607	93	45	in	in	ADP
ajst-3607	93	46	the	the	DET
ajst-3607	93	47	early	early	ADJ
ajst-3607	93	48	stage	stage	NOUN
ajst-3607	93	49	.	.	PUNCT
ajst-3607	94	1	comparison	comparison	NOUN
ajst-3607	94	2	of	of	ADP
ajst-3607	94	3	recognition	recognition	NOUN
ajst-3607	94	4	rate	rate	NOUN
ajst-3607	94	5	results	result	NOUN
ajst-3607	94	6	of	of	ADP
ajst-3607	94	7	different	different	ADJ
ajst-3607	94	8	algorithms	algorithm	NOUN
ajst-3607	94	9	is	be	AUX
ajst-3607	94	10	shown	show	VERB
ajst-3607	94	11	in	in	ADP
ajst-3607	94	12	table	table	NOUN
ajst-3607	94	13	2	2	NUM
ajst-3607	94	14	.	.	PUNCT
ajst-3607	94	15	table	table	NOUN
ajst-3607	94	16	2	2	NUM
ajst-3607	94	17	.	.	PUNCT
ajst-3607	94	18	comparison	comparison	NOUN
ajst-3607	94	19	of	of	ADP
ajst-3607	94	20	recognition	recognition	NOUN
ajst-3607	94	21	rate	rate	NOUN
ajst-3607	94	22	results	result	NOUN
ajst-3607	94	23	of	of	ADP
ajst-3607	94	24	different	different	ADJ
ajst-3607	94	25	algorithms	algorithm	NOUN
ajst-3607	94	26	algorithm	algorithm	NOUN
ajst-3607	94	27	recognition	recognition	NOUN
ajst-3607	94	28	method	method	NOUN
ajst-3607	94	29	recognition	recognition	NOUN
ajst-3607	94	30	rate	rate	NOUN
ajst-3607	94	31	algorithm	algorithm	NOUN
ajst-3607	94	32	in	in	ADP
ajst-3607	94	33	literature[14	literature[14	PROPN
ajst-3607	94	34	]	]	PUNCT
ajst-3607	94	35	stip+implicit	stip+implicit	ADJ
ajst-3607	94	36	shape	shape	NOUN
ajst-3607	94	37	model	model	NOUN
ajst-3607	94	38	73.30	73.30	NUM
ajst-3607	94	39	algorithm	algorithm	NOUN
ajst-3607	94	40	in	in	ADP
ajst-3607	94	41	literature[15	literature[15	NOUN
ajst-3607	94	42	]	]	X
ajst-3607	95	1	stip+bp+svm	stip+bp+svm	ADJ
ajst-3607	95	2	83.30	83.30	NUM
ajst-3607	95	3	algorithm	algorithm	NOUN
ajst-3607	95	4	in	in	ADP
ajst-3607	95	5	literature[2	literature[2	PROPN
ajst-3607	95	6	]	]	PUNCT
ajst-3607	95	7	description	description	NOUN
ajst-3607	95	8	of	of	ADP
ajst-3607	95	9	keyframe	keyframe	NOUN
ajst-3607	95	10	feature	feature	NOUN
ajst-3607	95	11	library	library	NOUN
ajst-3607	95	12	85.00	85.00	NUM
ajst-3607	95	13	algorithm	algorithm	NOUN
ajst-3607	95	14	in	in	ADP
ajst-3607	95	15	this	this	DET
ajst-3607	95	16	paper	paper	NOUN
ajst-3607	95	17	yolo+se	yolo+se	ADV
ajst-3607	95	18	-	-	PUNCT
ajst-3607	95	19	inception_v3+lstm	inception_v3+lstm	NOUN
ajst-3607	95	20	85.10	85.10	NUM
ajst-3607	95	21	153	153	NUM
ajst-3607	95	22	compared	compare	VERB
ajst-3607	95	23	with	with	ADP
ajst-3607	95	24	literature	literature	NOUN
ajst-3607	95	25	[	[	X
ajst-3607	95	26	2	2	NUM
ajst-3607	95	27	]	]	PUNCT
ajst-3607	95	28	,	,	PUNCT
ajst-3607	95	29	literature	literature	NOUN
ajst-3607	95	30	[	[	X
ajst-3607	95	31	14	14	NUM
ajst-3607	95	32	]	]	PUNCT
ajst-3607	95	33	and	and	CCONJ
ajst-3607	95	34	literature	literature	NOUN
ajst-3607	95	35	[	[	X
ajst-3607	95	36	15	15	NUM
ajst-3607	95	37	]	]	PUNCT
ajst-3607	95	38	,	,	PUNCT
ajst-3607	95	39	the	the	DET
ajst-3607	95	40	recognition	recognition	NOUN
ajst-3607	95	41	rate	rate	NOUN
ajst-3607	95	42	of	of	ADP
ajst-3607	95	43	this	this	DET
ajst-3607	95	44	algorithm	algorithm	NOUN
ajst-3607	95	45	framework	framework	NOUN
ajst-3607	95	46	is	be	AUX
ajst-3607	95	47	better	well	ADJ
ajst-3607	95	48	.	.	PUNCT
ajst-3607	96	1	although	although	SCONJ
ajst-3607	96	2	the	the	DET
ajst-3607	96	3	recognition	recognition	NOUN
ajst-3607	96	4	rate	rate	NOUN
ajst-3607	96	5	of	of	ADP
ajst-3607	96	6	literature	literature	NOUN
ajst-3607	96	7	[	[	X
ajst-3607	96	8	2	2	NUM
ajst-3607	96	9	]	]	PUNCT
ajst-3607	96	10	is	be	AUX
ajst-3607	96	11	the	the	DET
ajst-3607	96	12	same	same	ADJ
ajst-3607	96	13	as	as	ADP
ajst-3607	96	14	that	that	PRON
ajst-3607	96	15	of	of	ADP
ajst-3607	96	16	the	the	DET
ajst-3607	96	17	algorithm	algorithm	NOUN
ajst-3607	96	18	in	in	ADP
ajst-3607	96	19	this	this	DET
ajst-3607	96	20	paper	paper	NOUN
ajst-3607	96	21	,	,	PUNCT
ajst-3607	96	22	the	the	DET
ajst-3607	96	23	extraction	extraction	NOUN
ajst-3607	96	24	of	of	ADP
ajst-3607	96	25	key	key	ADJ
ajst-3607	96	26	frames	frame	NOUN
ajst-3607	96	27	requires	require	VERB
ajst-3607	96	28	a	a	DET
ajst-3607	96	29	lot	lot	NOUN
ajst-3607	96	30	of	of	ADP
ajst-3607	96	31	computation	computation	NOUN
ajst-3607	96	32	,	,	PUNCT
ajst-3607	96	33	requires	require	VERB
ajst-3607	96	34	high	high	ADJ
ajst-3607	96	35	hardware	hardware	NOUN
ajst-3607	96	36	requirements	requirement	NOUN
ajst-3607	96	37	,	,	PUNCT
ajst-3607	96	38	and	and	CCONJ
ajst-3607	96	39	the	the	DET
ajst-3607	96	40	description	description	NOUN
ajst-3607	96	41	of	of	ADP
ajst-3607	96	42	feature	feature	NOUN
ajst-3607	96	43	library	library	NOUN
ajst-3607	96	44	is	be	AUX
ajst-3607	96	45	complicated	complicate	VERB
ajst-3607	96	46	.	.	PUNCT
ajst-3607	97	1	the	the	DET
ajst-3607	97	2	algorithms	algorithm	NOUN
ajst-3607	97	3	in	in	ADP
ajst-3607	97	4	literature	literature	NOUN
ajst-3607	97	5	[	[	X
ajst-3607	97	6	14	14	NUM
ajst-3607	97	7	]	]	PUNCT
ajst-3607	97	8	and	and	CCONJ
ajst-3607	97	9	literature	literature	NOUN
ajst-3607	97	10	[	[	X
ajst-3607	97	11	15	15	NUM
ajst-3607	97	12	]	]	PUNCT
ajst-3607	97	13	need	need	VERB
ajst-3607	97	14	to	to	PART
ajst-3607	97	15	construct	construct	VERB
ajst-3607	97	16	the	the	DET
ajst-3607	97	17	stip	stip	NOUN
ajst-3607	97	18	event	event	NOUN
ajst-3607	97	19	sequence	sequence	NOUN
ajst-3607	97	20	,	,	PUNCT
ajst-3607	97	21	and	and	CCONJ
ajst-3607	97	22	the	the	DET
ajst-3607	97	23	feature	feature	NOUN
ajst-3607	97	24	extraction	extraction	NOUN
ajst-3607	97	25	algorithm	algorithm	NOUN
ajst-3607	97	26	has	have	VERB
ajst-3607	97	27	high	high	ADJ
ajst-3607	97	28	complexity	complexity	NOUN
ajst-3607	97	29	and	and	CCONJ
ajst-3607	97	30	needs	need	VERB
ajst-3607	97	31	a	a	DET
ajst-3607	97	32	lot	lot	NOUN
ajst-3607	97	33	of	of	ADP
ajst-3607	97	34	training	training	NOUN
ajst-3607	97	35	.	.	PUNCT
ajst-3607	98	1	experiments	experiment	NOUN
ajst-3607	98	2	verify	verify	VERB
ajst-3607	98	3	the	the	DET
ajst-3607	98	4	feasibility	feasibility	NOUN
ajst-3607	98	5	of	of	ADP
ajst-3607	98	6	the	the	DET
ajst-3607	98	7	proposed	propose	VERB
ajst-3607	98	8	algorithm	algorithm	NOUN
ajst-3607	98	9	in	in	ADP
ajst-3607	98	10	two	two	NUM
ajst-3607	98	11	-	-	PUNCT
ajst-3607	98	12	person	person	NOUN
ajst-3607	98	13	interaction	interaction	NOUN
ajst-3607	98	14	.	.	PUNCT
ajst-3607	99	1	4	4	X
ajst-3607	99	2	.	.	X
ajst-3607	99	3	conclusion	conclusion	NOUN
ajst-3607	99	4	remarks	remark	VERB
ajst-3607	99	5	in	in	ADP
ajst-3607	99	6	this	this	DET
ajst-3607	99	7	paper	paper	NOUN
ajst-3607	99	8	,	,	PUNCT
ajst-3607	99	9	a	a	DET
ajst-3607	99	10	multi	multi	ADJ
ajst-3607	99	11	-	-	ADJ
ajst-3607	99	12	model	model	ADJ
ajst-3607	99	13	fusion	fusion	NOUN
ajst-3607	99	14	network	network	NOUN
ajst-3607	99	15	is	be	AUX
ajst-3607	99	16	proposed	propose	VERB
ajst-3607	99	17	.	.	PUNCT
ajst-3607	100	1	firstly	firstly	ADV
ajst-3607	100	2	,	,	PUNCT
ajst-3607	100	3	yolo_v4	yolo_v4	NOUN
ajst-3607	100	4	is	be	AUX
ajst-3607	100	5	used	use	VERB
ajst-3607	100	6	to	to	PART
ajst-3607	100	7	remove	remove	VERB
ajst-3607	100	8	redundant	redundant	ADJ
ajst-3607	100	9	background	background	NOUN
ajst-3607	100	10	information	information	NOUN
ajst-3607	100	11	and	and	CCONJ
ajst-3607	100	12	detect	detect	VERB
ajst-3607	100	13	the	the	DET
ajst-3607	100	14	target	target	NOUN
ajst-3607	100	15	human	human	ADJ
ajst-3607	100	16	body	body	NOUN
ajst-3607	100	17	.	.	PUNCT
ajst-3607	101	1	secondly	secondly	ADV
ajst-3607	101	2	,	,	PUNCT
ajst-3607	101	3	the	the	DET
ajst-3607	101	4	se	se	NOUN
ajst-3607	101	5	-	-	NOUN
ajst-3607	101	6	incidence	incidence	NOUN
ajst-3607	101	7	_	_	PUNCT
ajst-3607	102	1	v3	v3	PROPN
ajst-3607	102	2	proposed	propose	VERB
ajst-3607	102	3	in	in	ADP
ajst-3607	102	4	this	this	DET
ajst-3607	102	5	paper	paper	NOUN
ajst-3607	102	6	is	be	AUX
ajst-3607	102	7	used	use	VERB
ajst-3607	102	8	to	to	PART
ajst-3607	102	9	extract	extract	VERB
ajst-3607	102	10	features	feature	NOUN
ajst-3607	102	11	.	.	PUNCT
ajst-3607	103	1	finally	finally	ADV
ajst-3607	103	2	,	,	PUNCT
ajst-3607	103	3	lstm	lstm	ADJ
ajst-3607	103	4	network	network	NOUN
ajst-3607	103	5	is	be	AUX
ajst-3607	103	6	used	use	VERB
ajst-3607	103	7	to	to	PART
ajst-3607	103	8	predict	predict	VERB
ajst-3607	103	9	and	and	CCONJ
ajst-3607	103	10	classify	classify	VERB
ajst-3607	103	11	the	the	DET
ajst-3607	103	12	two	two	NUM
ajst-3607	103	13	-	-	PUNCT
ajst-3607	103	14	person	person	NOUN
ajst-3607	103	15	interaction	interaction	NOUN
ajst-3607	103	16	behavior	behavior	NOUN
ajst-3607	103	17	.	.	PUNCT
ajst-3607	104	1	experimental	experimental	ADJ
ajst-3607	104	2	results	result	NOUN
ajst-3607	104	3	show	show	VERB
ajst-3607	104	4	that	that	SCONJ
ajst-3607	104	5	the	the	DET
ajst-3607	104	6	model	model	NOUN
ajst-3607	104	7	algorithm	algorithm	NOUN
ajst-3607	104	8	proposed	propose	VERB
ajst-3607	104	9	in	in	ADP
ajst-3607	104	10	this	this	DET
ajst-3607	104	11	paper	paper	NOUN
ajst-3607	104	12	has	have	AUX
ajst-3607	104	13	achieved	achieve	VERB
ajst-3607	104	14	good	good	ADJ
ajst-3607	104	15	results	result	NOUN
ajst-3607	104	16	on	on	ADP
ajst-3607	104	17	international	international	ADJ
ajst-3607	104	18	public	public	ADJ
ajst-3607	104	19	data	data	NOUN
ajst-3607	104	20	sets	set	NOUN
ajst-3607	104	21	,	,	PUNCT
ajst-3607	104	22	and	and	CCONJ
ajst-3607	104	23	the	the	DET
ajst-3607	104	24	recognition	recognition	NOUN
ajst-3607	104	25	accuracy	accuracy	NOUN
ajst-3607	104	26	of	of	ADP
ajst-3607	104	27	this	this	DET
ajst-3607	104	28	algorithm	algorithm	NOUN
ajst-3607	104	29	has	have	AUX
ajst-3607	104	30	been	be	AUX
ajst-3607	104	31	significantly	significantly	ADV
ajst-3607	104	32	improved	improve	VERB
ajst-3607	104	33	.	.	PUNCT
ajst-3607	105	1	although	although	SCONJ
ajst-3607	105	2	the	the	DET
ajst-3607	105	3	algorithm	algorithm	NOUN
ajst-3607	105	4	in	in	ADP
ajst-3607	105	5	this	this	DET
ajst-3607	105	6	paper	paper	NOUN
ajst-3607	105	7	has	have	VERB
ajst-3607	105	8	a	a	DET
ajst-3607	105	9	good	good	ADJ
ajst-3607	105	10	effect	effect	NOUN
ajst-3607	105	11	in	in	ADP
ajst-3607	105	12	the	the	DET
ajst-3607	105	13	recognition	recognition	NOUN
ajst-3607	105	14	of	of	ADP
ajst-3607	105	15	two	two	NUM
ajst-3607	105	16	-	-	PUNCT
ajst-3607	105	17	person	person	NOUN
ajst-3607	105	18	interaction	interaction	NOUN
ajst-3607	105	19	behavior	behavior	NOUN
ajst-3607	105	20	,	,	PUNCT
ajst-3607	105	21	it	it	PRON
ajst-3607	105	22	relies	rely	VERB
ajst-3607	105	23	on	on	ADP
ajst-3607	105	24	early	early	ADJ
ajst-3607	105	25	human	human	ADJ
ajst-3607	105	26	target	target	NOUN
ajst-3607	105	27	detection	detection	NOUN
ajst-3607	105	28	and	and	CCONJ
ajst-3607	105	29	multi	multi	ADJ
ajst-3607	105	30	-	-	ADJ
ajst-3607	105	31	feature	feature	ADJ
ajst-3607	105	32	extraction	extraction	NOUN
ajst-3607	105	33	,	,	PUNCT
ajst-3607	105	34	and	and	CCONJ
ajst-3607	105	35	is	be	AUX
ajst-3607	105	36	not	not	PART
ajst-3607	105	37	suitable	suitable	ADJ
ajst-3607	105	38	for	for	ADP
ajst-3607	105	39	the	the	DET
ajst-3607	105	40	behavior	behavior	NOUN
ajst-3607	105	41	recognition	recognition	NOUN
ajst-3607	105	42	of	of	ADP
ajst-3607	105	43	three	three	NUM
ajst-3607	105	44	or	or	CCONJ
ajst-3607	105	45	more	more	ADJ
ajst-3607	105	46	people	people	NOUN
ajst-3607	105	47	,	,	PUNCT
ajst-3607	105	48	and	and	CCONJ
ajst-3607	105	49	the	the	DET
ajst-3607	105	50	recognition	recognition	NOUN
ajst-3607	105	51	of	of	ADP
ajst-3607	105	52	multi	multi	ADJ
ajst-3607	105	53	-	-	ADJ
ajst-3607	105	54	person	person	ADJ
ajst-3607	105	55	interaction	interaction	NOUN
ajst-3607	105	56	behavior	behavior	NOUN
ajst-3607	105	57	is	be	AUX
ajst-3607	105	58	the	the	DET
ajst-3607	105	59	key	key	ADJ
ajst-3607	105	60	direction	direction	NOUN
ajst-3607	105	61	of	of	ADP
ajst-3607	105	62	future	future	ADJ
ajst-3607	105	63	research	research	NOUN
ajst-3607	105	64	.	.	PUNCT
ajst-3607	106	1	references	reference	NOUN
ajst-3607	106	2	[	[	X
ajst-3607	106	3	1	1	NUM
ajst-3607	106	4	]	]	X
ajst-3607	106	5	carreira	carreira	PROPN
ajst-3607	106	6	j	j	PROPN
ajst-3607	106	7	,	,	PUNCT
ajst-3607	106	8	zisserman	zisserman	PROPN
ajst-3607	106	9	a.	a.	PROPN
ajst-3607	106	10	quo	quo	PROPN
ajst-3607	106	11	vadis	vadis	NOUN
ajst-3607	106	12	,	,	PUNCT
ajst-3607	106	13	action	action	NOUN
ajst-3607	106	14	recognition?a	recognition?a	NOUN
ajst-3607	106	15	new	new	ADJ
ajst-3607	106	16	model	model	NOUN
ajst-3607	106	17	and	and	CCONJ
ajst-3607	106	18	the	the	DET
ajst-3607	106	19	kinetics	kinetic	NOUN
ajst-3607	106	20	dataset	dataset	VERB
ajst-3607	106	21	[	[	PUNCT
ajst-3607	106	22	c]//	c]//	PROPN
ajst-3607	106	23	2017	2017	NUM
ajst-3607	106	24	ieee	ieee	NOUN
ajst-3607	106	25	conference	conference	NOUN
ajst-3607	106	26	on	on	ADP
ajst-3607	106	27	computer	computer	NOUN
ajst-3607	106	28	vision	vision	NOUN
ajst-3607	106	29	and	and	CCONJ
ajst-3607	106	30	pattern	pattern	NOUN
ajst-3607	106	31	recognition.honolulu	recognition.honolulu	VERB
ajst-3607	106	32	:	:	PUNCT
ajst-3607	106	33	ieee,2017:1063	ieee,2017:1063	NUM
ajst-3607	106	34	-	-	SYM
ajst-3607	106	35	1071	1071	NUM
ajst-3607	106	36	.	.	PUNCT
ajst-3607	107	1	[	[	X
ajst-3607	107	2	2	2	X
ajst-3607	107	3	]	]	X
ajst-3607	107	4	ji	ji	PROPN
ajst-3607	107	5	xiaofei	xiaofei	PROPN
ajst-3607	107	6	,	,	PUNCT
ajst-3607	107	7	zuo	zuo	PROPN
ajst-3607	107	8	xinmeng	xinmeng	PROPN
ajst-3607	107	9	.	.	PUNCT
ajst-3607	108	1	two	two	NUM
ajst-3607	108	2	-	-	PUNCT
ajst-3607	108	3	person	person	NOUN
ajst-3607	108	4	interaction	interaction	NOUN
ajst-3607	108	5	behavior	behavior	NOUN
ajst-3607	108	6	recognition	recognition	NOUN
ajst-3607	108	7	based	base	VERB
ajst-3607	108	8	on	on	ADP
ajst-3607	108	9	statistical	statistical	ADJ
ajst-3607	108	10	features	feature	NOUN
ajst-3607	108	11	of	of	ADP
ajst-3607	108	12	key	key	ADJ
ajst-3607	108	13	frame	frame	NOUN
ajst-3607	108	14	feature	feature	NOUN
ajst-3607	108	15	database	database	NOUN
ajst-3607	109	1	[	[	X
ajst-3607	109	2	j	j	X
ajst-3607	109	3	]	]	X
ajst-3607	109	4	.	.	PUNCT
ajst-3607	110	1	computer	computer	NOUN
ajst-3607	110	2	application	application	NOUN
ajst-3607	110	3	,	,	PUNCT
ajst-3607	110	4	2016,36	2016,36	NUM
ajst-3607	110	5	(	(	PUNCT
ajst-3607	110	6	8)	8)	NUM
ajst-3607	110	7	:	:	PUNCT
ajst-3607	110	8	2287	2287	NUM
ajst-3607	110	9	-	-	SYM
ajst-3607	110	10	2291	2291	NUM
ajst-3607	110	11	.	.	PUNCT
ajst-3607	111	1	[	[	X
ajst-3607	111	2	3	3	X
ajst-3607	111	3	]	]	X
ajst-3607	111	4	chen	chen	PROPN
ajst-3607	111	5	changhong	changhong	PROPN
ajst-3607	111	6	,	,	PUNCT
ajst-3607	111	7	liu	liu	PROPN
ajst-3607	111	8	yuan	yuan	PROPN
ajst-3607	111	9	.	.	PUNCT
ajst-3607	112	1	two	two	NUM
ajst-3607	112	2	-	-	PUNCT
ajst-3607	112	3	person	person	NOUN
ajst-3607	112	4	interaction	interaction	NOUN
ajst-3607	112	5	behavior	behavior	NOUN
ajst-3607	112	6	recognition	recognition	NOUN
ajst-3607	112	7	based	base	VERB
ajst-3607	112	8	on	on	ADP
ajst-3607	112	9	improved	improved	ADJ
ajst-3607	112	10	sum	sum	NOUN
ajst-3607	112	11	-	-	PUNCT
ajst-3607	112	12	product	product	NOUN
ajst-3607	112	13	network	network	NOUN
ajst-3607	112	14	[	[	X
ajst-3607	112	15	j	j	X
ajst-3607	112	16	]	]	X
ajst-3607	112	17	.	.	PUNCT
ajst-3607	113	1	computer	computer	NOUN
ajst-3607	113	2	technology	technology	NOUN
ajst-3607	113	3	and	and	CCONJ
ajst-3607	113	4	development	development	NOUN
ajst-3607	113	5	,	,	PUNCT
ajst-3607	113	6	2019,29	2019,29	NUM
ajst-3607	113	7	(	(	PUNCT
ajst-3607	113	8	10	10	NUM
ajst-3607	113	9	):	):	SYM
ajst-3607	113	10	157163	157163	NUM
ajst-3607	113	11	.	.	PUNCT
ajst-3607	114	1	[	[	X
ajst-3607	114	2	4	4	X
ajst-3607	114	3	]	]	X
ajst-3607	114	4	pei	pei	PROPN
ajst-3607	114	5	xiaomin	xiaomin	PROPN
ajst-3607	114	6	,	,	PUNCT
ajst-3607	114	7	fan	fan	PROPN
ajst-3607	114	8	huijie	huijie	PROPN
ajst-3607	114	9	,	,	PUNCT
ajst-3607	114	10	tang	tang	PROPN
ajst-3607	114	11	yandong	yandong	PROPN
ajst-3607	114	12	.	.	PUNCT
ajst-3607	115	1	recognition	recognition	NOUN
ajst-3607	115	2	of	of	ADP
ajst-3607	115	3	twoperson	twoperson	NOUN
ajst-3607	115	4	interaction	interaction	NOUN
ajst-3607	115	5	behavior	behavior	NOUN
ajst-3607	115	6	in	in	ADP
ajst-3607	115	7	multi	multi	ADJ
ajst-3607	115	8	-	-	ADJ
ajst-3607	115	9	channel	channel	ADJ
ajst-3607	115	10	spatio	spatio	NOUN
ajst-3607	115	11	-	-	PUNCT
ajst-3607	115	12	temporal	temporal	ADJ
ajst-3607	115	13	fusion	fusion	NOUN
ajst-3607	115	14	network	network	NOUN
ajst-3607	115	15	[	[	X
ajst-3607	115	16	j	j	X
ajst-3607	115	17	]	]	X
ajst-3607	115	18	.	.	PUNCT
ajst-3607	116	1	infrared	infrared	ADJ
ajst-3607	116	2	and	and	CCONJ
ajst-3607	116	3	laser	laser	NOUN
ajst-3607	116	4	engineering	engineering	NOUN
ajst-3607	116	5	,	,	PUNCT
ajst-3607	116	6	2020,49	2020,49	NUM
ajst-3607	116	7	(	(	PUNCT
ajst-3607	116	8	5	5	NUM
ajst-3607	116	9	):	):	PUNCT
ajst-3607	116	10	211	211	NUM
ajst-3607	116	11	-	-	SYM
ajst-3607	116	12	216	216	NUM
ajst-3607	116	13	.	.	PUNCT
ajst-3607	117	1	[	[	X
ajst-3607	117	2	5	5	NUM
ajst-3607	117	3	]	]	PUNCT
ajst-3607	117	4	ryoo	ryoo	NOUN
ajst-3607	117	5	m	m	PROPN
ajst-3607	117	6	s.	s.	PROPN
ajst-3607	117	7	human	human	ADJ
ajst-3607	117	8	activity	activity	NOUN
ajst-3607	117	9	prediction	prediction	NOUN
ajst-3607	117	10	:	:	PUNCT
ajst-3607	117	11	early	early	ADJ
ajst-3607	117	12	recognition	recognition	NOUN
ajst-3607	117	13	of	of	ADP
ajst-3607	117	14	ongoing	ongoing	ADJ
ajst-3607	117	15	activities	activity	NOUN
ajst-3607	117	16	from	from	ADP
ajst-3607	117	17	streaming	stream	VERB
ajst-3607	117	18	videos	video	NOUN
ajst-3607	117	19	[	[	X
ajst-3607	117	20	c]//	c]//	PROPN
ajst-3607	117	21	international	international	ADJ
ajst-3607	117	22	conference	conference	NOUN
ajst-3607	117	23	on	on	ADP
ajst-3607	117	24	computer	computer	NOUN
ajst-3607	117	25	vision	vision	NOUN
ajst-3607	117	26	.	.	PUNCT
ajst-3607	118	1	barcelona	barcelona	PROPN
ajst-3607	118	2	,	,	PUNCT
ajst-3607	118	3	spain	spain	PROPN
ajst-3607	118	4	:	:	PUNCT
ajst-3607	118	5	ieee	ieee	NOUN
ajst-3607	118	6	,	,	PUNCT
ajst-3607	118	7	2011:1036	2011:1036	PROPN
ajst-3607	118	8	-	-	SYM
ajst-3607	118	9	1043	1043	NUM
ajst-3607	118	10	.	.	PUNCT
ajst-3607	119	1	[	[	X
ajst-3607	119	2	6	6	NUM
ajst-3607	119	3	]	]	PUNCT
ajst-3607	119	4	xu	xu	PROPN
ajst-3607	120	1	k	k	PROPN
ajst-3607	120	2	,	,	PUNCT
ajst-3607	120	3	qin	qin	PROPN
ajst-3607	121	1	z	z	PROPN
ajst-3607	121	2	,	,	PUNCT
ajst-3607	121	3	wang	wang	PROPN
ajst-3607	121	4	g.	g.	PROPN
ajst-3607	121	5	human	human	ADJ
ajst-3607	121	6	activities	activity	NOUN
ajst-3607	121	7	prediction	prediction	NOUN
ajst-3607	121	8	by	by	ADP
ajst-3607	121	9	learning	learn	VERB
ajst-3607	121	10	combinatorial	combinatorial	ADJ
ajst-3607	121	11	sparse	sparse	ADJ
ajst-3607	121	12	representations	representation	NOUN
ajst-3607	121	13	[	[	PUNCT
ajst-3607	121	14	c]//	c]//	PROPN
ajst-3607	121	15	ieee	ieee	PROPN
ajst-3607	121	16	international	international	ADJ
ajst-3607	121	17	conference	conference	NOUN
ajst-3607	121	18	on	on	ADP
ajst-3607	121	19	image	image	NOUN
ajst-3607	121	20	processing	processing	NOUN
ajst-3607	121	21	.	.	PUNCT
ajst-3607	122	1	arizona	arizona	PROPN
ajst-3607	122	2	,	,	PUNCT
ajst-3607	122	3	usa	usa	PROPN
ajst-3607	122	4	:	:	PUNCT
ajst-3607	122	5	ieee	ieee	NOUN
ajst-3607	122	6	,	,	PUNCT
ajst-3607	122	7	2016:747	2016:747	NUM
ajst-3607	122	8	-	-	SYM
ajst-3607	122	9	748	748	NUM
ajst-3607	122	10	.	.	PUNCT
ajst-3607	123	1	[	[	X
ajst-3607	123	2	7	7	X
ajst-3607	123	3	]	]	X
ajst-3607	123	4	raptis	raptis	NOUN
ajst-3607	123	5	m	m	PROPN
ajst-3607	123	6	,	,	PUNCT
ajst-3607	123	7	sigal	sigal	PROPN
ajst-3607	123	8	l.	l.	PROPN
ajst-3607	123	9	poselet	poselet	VERB
ajst-3607	123	10	key	key	NOUN
ajst-3607	123	11	-	-	PUNCT
ajst-3607	123	12	framing	framing	NOUN
ajst-3607	123	13	:	:	PUNCT
ajst-3607	123	14	a	a	DET
ajst-3607	123	15	model	model	NOUN
ajst-3607	123	16	for	for	ADP
ajst-3607	123	17	human	human	ADJ
ajst-3607	123	18	activity	activity	NOUN
ajst-3607	123	19	recognition	recognition	NOUN
ajst-3607	123	20	[	[	X
ajst-3607	123	21	c]//	c]//	PROPN
ajst-3607	123	22	ieee	ieee	PROPN
ajst-3607	123	23	conference	conference	NOUN
ajst-3607	123	24	on	on	ADP
ajst-3607	123	25	computer	computer	NOUN
ajst-3607	123	26	vision	vision	NOUN
ajst-3607	123	27	&	&	CCONJ
ajst-3607	123	28	pattern	pattern	NOUN
ajst-3607	123	29	recognition	recognition	NOUN
ajst-3607	123	30	.	.	PUNCT
ajst-3607	124	1	oregon	oregon	PROPN
ajst-3607	124	2	,	,	PUNCT
ajst-3607	124	3	usa	usa	PROPN
ajst-3607	124	4	:	:	PUNCT
ajst-3607	124	5	ieee	ieee	NOUN
ajst-3607	124	6	,	,	PUNCT
ajst-3607	124	7	2013	2013	NUM
ajst-3607	124	8	:	:	PUNCT
ajst-3607	124	9	2650	2650	NUM
ajst-3607	124	10	-	-	SYM
ajst-3607	124	11	2657	2657	NUM
ajst-3607	124	12	.	.	PUNCT
ajst-3607	125	1	[	[	X
ajst-3607	125	2	8	8	NUM
ajst-3607	125	3	]	]	X
ajst-3607	125	4	kong	kong	PROPN
ajst-3607	125	5	y	y	PROPN
ajst-3607	125	6	,	,	PUNCT
ajst-3607	125	7	fu	fu	PROPN
ajst-3607	125	8	y.	y.	PROPN
ajst-3607	125	9	max	max	PROPN
ajst-3607	125	10	-	-	PUNCT
ajst-3607	125	11	margin	margin	NOUN
ajst-3607	125	12	action	action	NOUN
ajst-3607	125	13	prediction	prediction	NOUN
ajst-3607	125	14	machine	machine	NOUN
ajst-3607	125	15	[	[	X
ajst-3607	125	16	j].ieee	j].ieee	NOUN
ajst-3607	125	17	transactions	transaction	NOUN
ajst-3607	125	18	on	on	ADP
ajst-3607	125	19	pattern	pattern	NOUN
ajst-3607	125	20	analysis	analysis	NOUN
ajst-3607	125	21	&	&	CCONJ
ajst-3607	125	22	machine	machine	NOUN
ajst-3607	125	23	intelligence	intelligence	NOUN
ajst-3607	125	24	,	,	PUNCT
ajst-3607	125	25	2015	2015	NUM
ajst-3607	125	26	,	,	PUNCT
ajst-3607	125	27	38	38	NUM
ajst-3607	125	28	(	(	PUNCT
ajst-3607	125	29	9):1844	9):1844	NUM
ajst-3607	125	30	-	-	PUNCT
ajst-3607	125	31	1858	1858	NUM
ajst-3607	125	32	.	.	PUNCT
ajst-3607	126	1	[	[	X
ajst-3607	126	2	9	9	NUM
ajst-3607	126	3	]	]	X
ajst-3607	126	4	chen	chen	PROPN
ajst-3607	126	5	w	w	PROPN
ajst-3607	126	6	,	,	PUNCT
ajst-3607	126	7	lu	lu	PROPN
ajst-3607	126	8	s	s	PROPN
ajst-3607	126	9	,	,	PUNCT
ajst-3607	126	10	liu	liu	PROPN
ajst-3607	126	11	b	b	PROPN
ajst-3607	126	12	,	,	PUNCT
ajst-3607	126	13	et	et	PROPN
ajst-3607	126	14	al	al	PROPN
ajst-3607	126	15	.	.	PUNCT
ajst-3607	126	16	detecting	detect	VERB
ajst-3607	126	17	citrus	citrus	NOUN
ajst-3607	126	18	in	in	ADP
ajst-3607	126	19	orchard	orchard	ADJ
ajst-3607	126	20	environment	environment	NOUN
ajst-3607	126	21	by	by	ADP
ajst-3607	126	22	using	use	VERB
ajst-3607	126	23	improved	improved	ADJ
ajst-3607	126	24	yolov4	yolov4	NOUN
ajst-3607	127	1	[	[	X
ajst-3607	127	2	j	j	X
ajst-3607	127	3	]	]	X
ajst-3607	127	4	.	.	PUNCT
ajst-3607	127	5	scientific	scientific	ADJ
ajst-3607	127	6	programming	programming	NOUN
ajst-3607	127	7	,	,	PUNCT
ajst-3607	127	8	2020	2020	NUM
ajst-3607	127	9	,	,	PUNCT
ajst-3607	127	10	2020	2020	NUM
ajst-3607	127	11	(	(	PUNCT
ajst-3607	127	12	1):1	1):1	NUM
ajst-3607	127	13	-	-	SYM
ajst-3607	127	14	13	13	NUM
ajst-3607	127	15	.	.	PUNCT
ajst-3607	128	1	[	[	X
ajst-3607	128	2	10	10	NUM
ajst-3607	128	3	]	]	X
ajst-3607	128	4	jie	jie	PROPN
ajst-3607	128	5	h	h	PROPN
ajst-3607	128	6	,	,	PUNCT
ajst-3607	128	7	li	li	PROPN
ajst-3607	128	8	s	s	PROPN
ajst-3607	128	9	,	,	PUNCT
ajst-3607	128	10	albanie	albanie	PROPN
ajst-3607	128	11	a	a	PROPN
ajst-3607	128	12	,	,	PUNCT
ajst-3607	128	13	et	et	PROPN
ajst-3607	128	14	al	al	PROPN
ajst-3607	128	15	.	.	PUNCT
ajst-3607	128	16	squeeze	squeeze	NOUN
ajst-3607	128	17	-	-	PUNCT
ajst-3607	128	18	and	and	CCONJ
ajst-3607	128	19	-	-	PUNCT
ajst-3607	128	20	excitation	excitation	NOUN
ajst-3607	128	21	networks	network	NOUN
ajst-3607	128	22	[	[	X
ajst-3607	128	23	j	j	X
ajst-3607	128	24	]	]	X
ajst-3607	128	25	.	.	PUNCT
ajst-3607	129	1	ieee	ieee	NOUN
ajst-3607	129	2	transactions	transaction	NOUN
ajst-3607	129	3	on	on	ADP
ajst-3607	129	4	pattern	pattern	NOUN
ajst-3607	129	5	analysis	analysis	NOUN
ajst-3607	129	6	and	and	CCONJ
ajst-3607	129	7	machine	machine	NOUN
ajst-3607	129	8	intelligence,2020,42	intelligence,2020,42	NOUN
ajst-3607	129	9	(	(	PUNCT
ajst-3607	129	10	8)	8)	NUM
ajst-3607	129	11	:	:	PUNCT
ajst-3607	129	12	2011	2011	NUM
ajst-3607	129	13	-	-	SYM
ajst-3607	129	14	2023	2023	NUM
ajst-3607	129	15	.	.	PUNCT
ajst-3607	130	1	[	[	X
ajst-3607	130	2	11	11	NUM
ajst-3607	130	3	]	]	PUNCT
ajst-3607	130	4	szegedy	szegedy	VERB
ajst-3607	130	5	c	c	NOUN
ajst-3607	130	6	,	,	PUNCT
ajst-3607	130	7	vanhoucke	vanhoucke	NOUN
ajst-3607	130	8	v	v	NOUN
ajst-3607	130	9	,	,	PUNCT
ajst-3607	130	10	ioffe	ioffe	PROPN
ajst-3607	130	11	s.	s.	PROPN
ajst-3607	130	12	et	et	PROPN
ajst-3607	130	13	al	al	PROPN
ajst-3607	130	14	.	.	PUNCT
ajst-3607	131	1	rethinking	rethink	VERB
ajst-3607	131	2	the	the	DET
ajst-3607	131	3	inception	inception	ADJ
ajst-3607	131	4	architecture	architecture	NOUN
ajst-3607	131	5	for	for	ADP
ajst-3607	131	6	computer	computer	NOUN
ajst-3607	131	7	vision	vision	NOUN
ajst-3607	131	8	[	[	X
ajst-3607	131	9	c]//	c]//	PROPN
ajst-3607	131	10	2016	2016	NUM
ajst-3607	131	11	ieee	ieee	NOUN
ajst-3607	131	12	conference	conference	NOUN
ajst-3607	131	13	on	on	ADP
ajst-3607	131	14	computer	computer	NOUN
ajst-3607	131	15	vision	vision	NOUN
ajst-3607	131	16	and	and	CCONJ
ajst-3607	131	17	pattern	pattern	NOUN
ajst-3607	131	18	recognition	recognition	NOUN
ajst-3607	131	19	.	.	PUNCT
ajst-3607	132	1	las	las	PROPN
ajst-3607	132	2	vegas	vegas	PROPN
ajst-3607	132	3	:	:	PUNCT
ajst-3607	132	4	ieee,2016:2818	ieee,2016:2818	PROPN
ajst-3607	132	5	-	-	SYM
ajst-3607	132	6	2826	2826	NUM
ajst-3607	132	7	.	.	PUNCT
ajst-3607	133	1	[	[	X
ajst-3607	133	2	12	12	NUM
ajst-3607	133	3	]	]	X
ajst-3607	133	4	wang	wang	PROPN
ajst-3607	133	5	yi	yi	PROPN
ajst-3607	133	6	,	,	PUNCT
ajst-3607	133	7	ma	ma	PROPN
ajst-3607	133	8	cuihong	cuihong	PROPN
ajst-3607	133	9	,	,	PUNCT
ajst-3607	133	10	mao	mao	PROPN
ajst-3607	133	11	zhiqiang	zhiqiang	PROPN
ajst-3607	133	12	.	.	PUNCT
ajst-3607	134	1	research	research	NOUN
ajst-3607	134	2	on	on	ADP
ajst-3607	134	3	behavior	behavior	NOUN
ajst-3607	134	4	recognition	recognition	NOUN
ajst-3607	134	5	based	base	VERB
ajst-3607	134	6	on	on	ADP
ajst-3607	134	7	3d	3d	NUM
ajst-3607	134	8	convolution	convolution	NOUN
ajst-3607	134	9	and	and	CCONJ
ajst-3607	134	10	bidirectional	bidirectional	ADJ
ajst-3607	134	11	lstm	lstm	NOUN
ajst-3607	135	1	[	[	X
ajst-3607	135	2	j	j	X
ajst-3607	135	3	]	]	X
ajst-3607	135	4	.	.	PUNCT
ajst-3607	136	1	modern	modern	ADJ
ajst-3607	136	2	electronic	electronic	ADJ
ajst-3607	136	3	technology	technology	NOUN
ajst-3607	136	4	,	,	PUNCT
ajst-3607	136	5	2019,42	2019,42	NUM
ajst-3607	136	6	(	(	PUNCT
ajst-3607	136	7	14	14	NUM
ajst-3607	136	8	):	):	PUNCT
ajst-3607	136	9	78	78	NUM
ajst-3607	136	10	-	-	SYM
ajst-3607	136	11	82	82	NUM
ajst-3607	136	12	.	.	PUNCT
ajst-3607	137	1	[	[	X
ajst-3607	137	2	13	13	NUM
ajst-3607	137	3	]	]	X
ajst-3607	137	4	qian	qian	PROPN
ajst-3607	137	5	h	h	PROPN
ajst-3607	137	6	,	,	PUNCT
ajst-3607	137	7	zhou	zhou	PROPN
ajst-3607	137	8	x	x	PROPN
ajst-3607	137	9	,	,	PUNCT
ajst-3607	137	10	zheng	zheng	PROPN
ajst-3607	137	11	m.	m.	NOUN
ajst-3607	137	12	abnormal	abnormal	ADJ
ajst-3607	137	13	behavior	behavior	NOUN
ajst-3607	137	14	detection	detection	NOUN
ajst-3607	137	15	and	and	CCONJ
ajst-3607	137	16	recognition	recognition	NOUN
ajst-3607	137	17	method	method	NOUN
ajst-3607	137	18	based	base	VERB
ajst-3607	137	19	on	on	ADP
ajst-3607	137	20	improved	improve	VERB
ajst-3607	137	21	resnet	resnet	NOUN
ajst-3607	137	22	model	model	NOUN
ajst-3607	137	23	[	[	X
ajst-3607	137	24	j].computers	j].computer	NOUN
ajst-3607	137	25	,	,	PUNCT
ajst-3607	137	26	materials	material	NOUN
ajst-3607	137	27	and	and	CCONJ
ajst-3607	137	28	continua	continuum	NOUN
ajst-3607	137	29	,	,	PUNCT
ajst-3607	137	30	2020	2020	NUM
ajst-3607	137	31	,	,	PUNCT
ajst-3607	137	32	65	65	NUM
ajst-3607	137	33	(	(	PUNCT
ajst-3607	137	34	3):2153	3):2153	NUM
ajst-3607	137	35	-	-	SYM
ajst-3607	137	36	2167	2167	NUM
ajst-3607	137	37	.	.	PUNCT
ajst-3607	138	1	[	[	X
ajst-3607	138	2	14	14	NUM
ajst-3607	138	3	]	]	X
ajst-3607	138	4	zhang	zhang	PROPN
ajst-3607	138	5	x	x	PROPN
ajst-3607	138	6	,	,	PUNCT
ajst-3607	138	7	cui	cui	PROPN
ajst-3607	138	8	j	j	PROPN
ajst-3607	138	9	,	,	PUNCT
ajst-3607	138	10	tian	tian	PROPN
ajst-3607	138	11	l	l	PROPN
ajst-3607	138	12	,	,	PUNCT
ajst-3607	138	13	et	et	PROPN
ajst-3607	138	14	al	al	PROPN
ajst-3607	138	15	.	.	PUNCT
ajst-3607	138	16	local	local	ADJ
ajst-3607	138	17	spatiotemporal	spatiotemporal	ADJ
ajst-3607	138	18	feature	feature	NOUN
ajst-3607	138	19	based	base	VERB
ajst-3607	138	20	voting	voting	NOUN
ajst-3607	138	21	framework	framework	NOUN
ajst-3607	138	22	for	for	ADP
ajst-3607	138	23	complex	complex	ADJ
ajst-3607	138	24	human	human	ADJ
ajst-3607	138	25	activity	activity	NOUN
ajst-3607	138	26	detection	detection	NOUN
ajst-3607	138	27	and	and	CCONJ
ajst-3607	138	28	localization	localization	NOUN
ajst-3607	138	29	[	[	X
ajst-3607	138	30	c]//	c]//	ADJ
ajst-3607	138	31	proceedings	proceeding	NOUN
ajst-3607	138	32	of	of	ADP
ajst-3607	138	33	the	the	DET
ajst-3607	138	34	first	first	ADJ
ajst-3607	138	35	asian	asian	ADJ
ajst-3607	138	36	conference	conference	NOUN
ajst-3607	138	37	on	on	ADP
ajst-3607	138	38	pattern	pattern	NOUN
ajst-3607	138	39	recognition	recognition	NOUN
ajst-3607	138	40	.	.	PUNCT
ajst-3607	139	1	piscataway	piscataway	PROPN
ajst-3607	139	2	:	:	PUNCT
ajst-3607	139	3	ieee	ieee	NOUN
ajst-3607	139	4	,	,	PUNCT
ajst-3607	139	5	2011	2011	NUM
ajst-3607	139	6	:	:	PUNCT
ajst-3607	139	7	12	12	NUM
ajst-3607	139	8	-	-	SYM
ajst-3607	139	9	16	16	NUM
ajst-3607	139	10	.	.	PUNCT
ajst-3607	140	1	[	[	X
ajst-3607	140	2	15	15	NUM
ajst-3607	140	3	]	]	X
ajst-3607	140	4	yu	yu	PROPN
ajst-3607	140	5	t	t	PROPN
ajst-3607	140	6	h	h	PROPN
ajst-3607	140	7	,	,	PUNCT
ajst-3607	140	8	kim	kim	PROPN
ajst-3607	140	9	t	t	PROPN
ajst-3607	140	10	k	k	PROPN
ajst-3607	140	11	,	,	PUNCT
ajst-3607	140	12	cipolla	cipolla	PROPN
ajst-3607	140	13	r.	r.	PROPN
ajst-3607	140	14	real	real	ADJ
ajst-3607	140	15	time	time	NOUN
ajst-3607	140	16	action	action	NOUN
ajst-3607	140	17	recognition	recognition	NOUN
ajst-3607	140	18	by	by	ADP
ajst-3607	140	19	spatiotemporal	spatiotemporal	ADJ
ajst-3607	140	20	semantic	semantic	ADJ
ajst-3607	140	21	and	and	CCONJ
ajst-3607	140	22	structural	structural	ADJ
ajst-3607	140	23	forests[c]//	forests[c]//	PROPN
ajst-3607	140	24	proceedings	proceeding	NOUN
ajst-3607	140	25	of	of	ADP
ajst-3607	140	26	the	the	DET
ajst-3607	140	27	21st	21st	ADJ
ajst-3607	140	28	british	british	ADJ
ajst-3607	140	29	machine	machine	NOUN
ajst-3607	140	30	vision	vision	PROPN
ajst-3607	140	31	conference	conference	PROPN
ajst-3607	140	32	.	.	PUNCT
ajst-3607	141	1	bristol	bristol	PROPN
ajst-3607	141	2	:	:	PUNCT
ajst-3607	141	3	bmvc,2010:1	bmvc,2010:1	PROPN
ajst-3607	141	4	-	-	PUNCT
ajst-3607	141	5	12	12	NUM
ajst-3607	141	6	.	.	PUNCT