Semisupervised Deep Features of Time-Frequency Maps for Multimodal Emotion Recognition

Table 8

Classification accuracy and kappa score of the four-class scenario for different CNNs and classifiers.

Classifier	CNN
	AlexNet		VGG19		ResNet18		Inception-V3		EfficientNet-b0
	Acc	Kappa	Acc	Kappa	Acc	Kappa	Acc	Kappa	Acc	Kappa
SVM	0.876	0.752	0.881	0.762	0.901	0.816	0.928	0.856	0.904	0.808
ANN	0.875	0.750	0.886	0.774	0.885	0.802	0.901	0.801	0.898	0.796
kNN	0.864	0.728	0.866	0.732	0.875	0.782	0.891	0.783	0.884	0.768
Random forest	0.847	0.694	0.875	0.751	0.901	0.778	0.889	0.779	0.882	0.764
Decision tree	0.847	0.694	0.853	0.701	0.871	0.764	0.882	0.764	0.889	0.778

The bold values represent the highest accuracies.