Table 1.

Classification results for proposed method using different pre-trained CNNs and PDC images on MAHNOB-HCI database in all frequency bands while using tenfold Cross-Validation method

Net	Bands	Accuracy (%)	Precision (%)	Recall (%)	F1- score (%)
AlexNet	Delta	75.76 ± 0.49	75.74 ± 0.56	75.78 ± 0.51	76.72 ± 0.55
	Theta	76.28 ± 0.52	76.25 ± 0.55	76.24 ± 0.63	76.24 ± 0.50
	Alpha	78.32 ± 0.63	78.30 ± 0.53	78.35 ± 0.50	78.26 ± 0.67
	Beta	77.35 ± 0.65	77.34 ± 0.56	77.36 ± 0.55	77.32 ± 0.56
	Gamma	77.15 ± 0.55	77.12 ± 0.50	77.16 ± 0.56	77.13 ± 0.50
VGG-19	Delta	85.25 ± 0.54	85.22 ± 0.50	85.26 ± 0.50	85.20 ± 0.52
	Theta	85.45 ± 0.58	85.75 ± 0.55	85.78 ± 0.54	85.42 ± 0.55
	Alpha	87.38 ± 0.46	87.33 ± 0.50	87.42 ± 0.52	87.32 ± 0.52
	Beta	86.05 ± 0.52	86.03 ± 0.53	86.10 ± 0.52	86.02 ± 0.53
	Gamma	85.75 ± 0.50	85.72 ± 0.40	85.78 ± 0.65	85.72 ± 0.50
ResNet-50	Delta	93.24 ± 0.53	93.23 ± 0.54	93.25 ± 0.50	93.22 ± 0.52
	Theta	93.64 ± 0.54	93.63 ± 0.55	93.66 ± 0.54	93.61 ± 0.58
	Alpha	96.26 ± 0.53	96.23 ± 0.56	96.28 ± 0.55	96.23 ± 0.54
	Beta	94.53 ± 0.52	94.52 ± 0.57	94.55 ± 0.52	94.50 ± 0.55
	Gamma	94.22 ± 0.51	94.16 ± 0.53	94.24 ± 0.55	94.20 ± 0.50
Inception-v3	Delta	92.10 ± 0.54	92.08 ± 0.53	92.14 ± 0.55	92.07 ± 0.52
	Theta	92.39 ± 0.58	92.37 ± 0.50	92.42 ± 0.50	92.35 ± 0.50
	Alpha	95.85 ± 0.56	95.83 ± 0.54	95.87 ± 0.52	95.83 ± 0.60
	Beta	93.29 ± 0.50	93.25 ± 0.59	93.34 ± 0.56	93.26 ± 0.57
	Gamma	93.07 ± 0.53	93.05 ± 0.56	93.12 ± 0.55	93.05 ± 0.55

Highest results are showed bold