Table 8.

Classification results for proposed method using different pre-trained CNNs and dDTF images on MAHNOB-HCI database in all frequency bands using the LOSO Cross-Validation method

Net	Bands	Accuracy (%)	Precision (%)	Recall (%)	F1-score (%)
AlexNet	Delta	76.55 ± 0.75	76.59 ± 3.43	76.50 ± 3.24	76.42 ± 3.22
	Theta	77.07 ± 0.74	77.12 ± 3.22	77.15 ± 3.52	77.10 ± 3.44
	Alpha	77.73 ± 0.73	77.77 ± 3.40	77.75 ± 3.53	77.64 ± 3.42
	Beta	77.28 ± 0.73	77.35 ± 3.43	77.26 ± 3.35	77.21 ± 3.41
	Gamma	77.40 ± 0.74	77.43 ± 3.32	77.32 ± 3.40	77.32 ± 3.45
VGG-19	Delta	84.36 ± 3.45	84.57 ± 3.32	84.32 ± 3.39	84.15 ± 3.34
	Theta	84.69 ± 3.49	84.83 ± 3.45	84.57 ± 3.35	84.52 ± 3.28
	Alpha	86.42 ± 3.25	86.66 ± 3.40	86.35 ± 3.36	86.34 ± 3.37
	Beta	85.30 ± 3.34	85.52 ± 3.42	85.24 ± 3.34	85.22 ± 3.34
	Gamma	85.42 ± 3.50	85.67 ± 3.38	85.36 ± 3.40	85.34 ± 3.32
ResNet-50	Delta	95.23 ± 3.27	95.73 ± 3.35	95.32 ± 3.37	95.13 ± 3.24
	Theta	96.65 ± 3.25	96.75 ± 3.34	96.65 ± 3.32	96.48 ± 3.25
	Alpha	97.78 ± 3.32	97.72 ± 3.33	97.71 ± 3.45	97.73 ± 3.24
	Beta	96.91 ± 3.24	96.93 ± 3.25	96.84 ± 3.24	96.75 ± 3.35
	Gamma	96.86 ± 3.21	96.90 ± 3.21	96.82 ± 3.40	96.73 ± 3.42
Inception-v3	Delta	94.26 ± 3.55	94.24 ± 3.52	94.33 ± 3.43	94.15 ± 3.22
	Theta	94.54 ± 3.52	94.60 ± 3.43	94.54 ± 3.25	94.37 ± 3.31
	Alpha	96.86 ± 3.54	96.83 ± 3.22	96.78 ± 3.44	96.70 ± 3.30
	Beta	95.25 ± 3.26	95.30 ± 3.45	95.32 ± 3.30	95.20 ± 3.32
	Gamma	95.20 ± 3.23	95.32 ± 3.35	95.24 ± 3.23	95.15 ± 3.34

Highest results are showed bold