Controlling Embedded Systems Remotely via Internet-of-Things Based on Emotional Recognition

Table 4

HOG and SVM optimized parameters for training and testing with SMO.

Exp. No.	HOG-CS	HOG-bs	Image size	Feature length/image	Training time	Testing time	Accuracy	ConMat
1	7	2	80	1440	93.8	2	88.70	803	116
								110	971
2	9	2	80	756	3.26	0.8	88.90	817	102
								120	961
3	11	2	80	432	2.83	0.5	88.25	801	118
								117	964
4	13	2	80	360	2.76	0.3	88.35	799	120
								113	968
5	15	2	80	144	3.28	0.1	88.30	800	119
								115	966