Human Activity Recognition Using Gaussian Mixture Hidden Conditional Random Fields

Table 6

Classification results of the proposed system on KTH action dataset (A) using ANN, (B) using SVM, (C) using HMM, and (D) using existing HCRF [30], while removing the proposed HCRF model (unit: %).

Activities	Walking	Jogging	Running	Boxing	Hand-wave	Handclap
(A)
Walking	79	5	6	4	3	3
Jogging	3	81	5	3	4	4
Running	6	4	77	5	5	3
Boxing	6	7	6	69	5	7
Hand-wave	4	7	5	5	73	6
Handclap	4	6	5	4	6	75
Average	75.66
(B)
Walking	82	2	3	5	4	4
Jogging	3	86	2	3	2	4
Running	5	3	80	4	5	3
Boxing	5	3	3	79	4	6
Hand-wave	1	4	3	3	89	0
Handclap	3	5	2	4	3	83
Average	83.17
(C)
Walking	86	3	2	4	2	3
Jogging	0	88	3	2	4	3
Running	0	3	90	0	4	3
Boxing	3	0	4	92	1	0
Hand-wave	1	3	2	2	91	1
Handclap	1	3	4	1	2	89
Average	89.33
(D)
Walking	90	3	0	3	4	0
Jogging	2	88	2	3	3	2
Running	4	2	92	0	0	2
Boxing	1	3	2	91	3	0
Hand-wave	0	1	3	2	93	1
Handclap	1	3	2	4	3	87
Average	90.17