Edge Detection-Based Feature Extraction for the Systems of Activity Recognition

Table 8

Accuracy of classification for the proposed activity recognition system with Haralick texture features (without employing the proposed methodology) against depth dataset.

Activities	BN	JC	PLJ	RNN	SIM	SKP	WLK	OW1	OW2	JP	CLP	BXG	SUD
BN	88	0	2	1	1	2	0	1	2	1	0	2	0
JC	1	90	1	2	0	2	1	0	1	0	1	0	1
PLJ	0	2	85	0	2	0	2	1	2	1	1	2	2
RNN	2	2	1	79	1	2	1	2	1	1	4	2	2
SIM	1	3	2	1	77	4	2	1	3	2	1	2	1
SKP	2	1	1	2	2	81	1	2	1	2	1	3	1
WLK	1	2	2	1	0	1	83	1	2	1	2	1	2
OW1	2	1	2	2	1	2	1	76	2	2	1	2	6
OW2	1	2	1	1	2	1	0	2	86	1	2	0	1
JP	0	1	1	2	0	0	2	1	0	92	0	1	0
CLP	2	0	2	0	2	1	1	0	1	2	87	0	2
BXG	1	2	0	1	1	0	1	2	0	1	1	89	1
SUD	0	1	2	0	1	1	0	0	1	0	2	1	91
Average	84.9%

BN for bending, JC for jacking, PLJ for place-jumping, RNN for running, SIM for side movement, SKP for skipping, WLK for walking, OW1 for one-hand waving, OW2 for two-hand waving, JP for jumping, CLP for clapping, BXG for boxing, and SUD for sitting up and down.