An Efficient Feature Selection Method for Video-Based Activity Recognition Systems

Table 6

Accuracy of the designed approach with greedy backward selection on depth database.

Activities	BD	JK	PJU	RUN	SMO	SKI	WAK	1HW	2HW	JU	CL	BOX	SS
BD	81	2	1	2	3	2	1	2	2	1	0	1	2
JK	1	88	2	1	0	2	1	2	1	0	1	1	0
PJU	0	1	90	2	1	0	2	1	0	1	1	0	1
RUN	2	2	0	84	2	1	1	0	2	1	2	2	1
SMO	1	2	2	1	83	2	1	2	1	2	1	0	2
SKI	2	0	1	2	2	85	1	2	1	0	1	2	1
WAK	0	2	0	1	0	2	89	1	0	2	1	0	2
1HW	1	0	2	0	1	0	0	92	2	1	0	1	0
2HW	1	2	1	2	0	1	2	2	84	1	2	1	1
JU	2	2	1	1	2	2	1	1	4	80	1	2	1
CL	2	1	2	2	1	1	2	2	1	2	81	1	2
BOX	0	2	1	1	2	0	1	2	0	1	2	87	1
SS	1	0	2	0	1	2	0	0	2	0	1	0	91
Average	85.8%