Table 6.

Comparison of classification accuracy of multiple binary classifiers.

Kernel amount	Kernel size	Round of training	Happy	Surprise	Sadness	Anger	Fear	Disgust	Average Accuracy
1,632	3 × 3 1 × 1	1	0.4191	0.5126	0.5704	0.8600	0.7513	0.7811	0.6491
		2	0.4512	0.5901	0.5991	0.7741	0.7632	0.4842	0.6102
		3	0.5101	0.5609	0.6004	0.8292	0.7339	0.7925	0.6711
	5 × 5 3 × 3	1	0.5298	0.5618	0.5698	0.7801	0.7189	0.6970	0.6429
		2	0.5312	0.6298	0.5710	0.8212	0.7331	0.7335	0.6698
		3	0.5201	0.5400	0.5192	0.8301	0.7231	0.5941	0.6210
	7 × 7 5 × 5	1	0.5231	0.5602	0.6154	0.8614	0.7616	0.7112	0.6721
		2	0.5491	0.5301	0.6716	0.8413	0.7531	0.7291	0.6790
		3	0.5613	0.5491	0.6271	0.8413	0.7531	0.8043	0.6893

The bold values are the best values obtained in the experiment.