Emotion Recognition Based on EEG Using Generative Adversarial Nets and Convolutional Neural Network

Table 6

The recognition accuracy (%) of the two models in two classification tasks before and after data augmentation.

Subject	Case 1				Case 2
	Valence		Arousal		Valence		Arousal
	FBCCNN	FBSCNN	FBCCNN	FBSCNN	FBCCNN	FBSCNN	FBCCNN	FBSCNN
S01	86.04	82.67	86.43	83.16	90.83	89.25	92.25	89.09
S02	77.34	63.15	77.73	63.64	81.36	71.08	82.78	70.92
S04	80.44	62.44	80.83	62.94	85.13	72.39	86.55	72.24
S06	76.65	63.15	77.04	63.64	80.33	70.77	80.75	70.61
S07	90.10	70.05	88.49	70.54	95.10	80.19	95.52	80.03
S08	83.55	75.56	83.94	76.05	87.5	82.11	88.92	81.95
S09	84.58	73.84	84.97	74.33	89.86	79.94	90.28	79.78
S10	86.31	77.29	86.70	77.78	93.67	83.41	94.09	83.25
S11	80.10	66.60	80.49	67.09	86.16	75.89	87.58	75.74
S17	81.82	73.84	82.21	74.33	85.81	79.55	88.23	79.39
S18	79.41	66.94	79.80	67.43	86.05	73.86	87.47	73.70
S22	69.01	66.42	69.40	64.91	76.61	72.10	79.03	71.94
Average recognition accuracy results across subjects on “valence” and “arousal”
Across subjects	82.40	78.65	83.55	77.15	88.90	83.64	90.26	80.55