Human Activity Recognition Using Gaussian Mixture Hidden Conditional Random Fields

Table 5

Classification results of the proposed system on Weizmann 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	Bend	Jack	Pjump	Run	Side	Skip	Walk	Wave 1	Wave 2
(A)
Bend	70	4	5	3	3	2	5	5	3
Jack	4	68	3	6	7	2	3	3	4
Pjump	2	4	75	6	3	2	2	2	4
Run	4	2	3	72	6	2	5	3	3
Side	5	3	5	4	65	6	4	6	2
Skip	4	6	4	3	5	67	3	2	6
Walk	4	2	4	7	3	4	70	3	3
Wave 1	2	1	3	3	4	5	7	71	4
Wave 2	2	5	3	6	4	4	3	5	68
Average	69.55
(B)
Bend	69	3	4	4	6	4	2	3	5
Jack	2	72	2	3	4	3	5	4	5
Pjump	1	4	75	2	4	5	4	2	3
Run	2	4	3	78	2	2	4	2	3
Side	2	4	5	3	70	4	3	5	4
Skip	2	1	3	2	4	80	3	3	2
Walk	2	0	3	4	3	2	82	1	3
Wave 1	2	2	3	4	3	2	3	77	4
Wave 2	1	2	1	2	3	1	3	4	83
Average	76.22
(C)
Bend	82	3	0	2	2	3	1	5	2
Jack	3	80	1	2	3	2	3	4	2
Pjump	3	4	85	3	0	0	1	2	2
Run	5	4	2	79	0	2	1	3	4
Side	0	1	5	4	81	3	1	2	3
Skip	3	1	2	2	3	88	0	0	1
Walk	0	2	3	2	1	2	83	3	4
Wave 1	1	3	2	2	4	2	3	78	5
Wave 2	1	2	2	2	2	3	1	0	87
Average	82.56
(D)
Bend	80	2	3	1	4	0	5	2	3
Jack	1	88	0	2	0	3	2	3	1
Pjump	0	2	90	1	0	3	0	2	2
Run	2	1	2	85	2	3	0	0	5
Side	4	1	2	3	80	4	1	2	3
Skip	1	4	0	5	1	84	0	3	2
Walk	2	1	0	0	1	2	89	2	3
Wave 1	3	0	1	2	0	2	0	91	1
Wave 2	4	1	3	0	2	3	0	2	85
Average	85.78