Frame-Based Facial Expression Recognition Using Geometrical Features

Table 4

The confusion matrix obtained by applying our approach on BU-4DFE database. The first row in each expression represents results using SVM classifier. The other row shows the results using KNN classifier.

	An	Di	Fe	Ha	Sa	Su	Ne
An	62.7	10.6	2.11	1.40	11.9	0.00	11.2
	42.25	7.042	2.816	2.816	16.90	0.00	28.16
Di	16.07	59.8	6.25	6.25	2.67	2.67	6.25
	19.64	46.42	1.785	12.50	3.571	1.785	14.28
Fe	7.62	3.38	55.9	11.8	11.8	0.85	8.47
	3.389	8.474	28.81	15.25	16.94	1.694	25.42
Ha	2.43	3.65	3.65	88.4	0.00	1.83	0.00
	0.00	7.317	1.219	85.36	3.658	2.43	0.00
Sa	17.4	1.58	11.1	0.00	53.4	0.00	16.34
	19.04	1.587	12.69	0.00	44.44	0.00	22.22
Su	0.00	3.45	1.14	0.57	1.14	93.7	0.00
	0.00	5.747	3.448	0.00	0.00	86.20	4.597
Ne	20.2	2.02	7.16	2.02	4.18	2.02	62.4
	6.00	0.00	12.00	2.00	6.00	2.00	72.00