Table 4.

Average accuracies of bianry and 4-class classification using different approaches with several input types on the DEAP dataset.

Input Type	Method	CNN	Accuracy
			Valence	Arousal	4-Class
handcrafted features	PSD [19]	2D	0.9446	0.9552	0.9326
		3D	0.9543	0.9578	0.9454
raw data	concatenation [32]	2D	0.9048	0.9166	0.9142
		3D	0.8883	0.8942	0.9056
	3D EEG stream	3D	0.9874	0.9928	0.9932