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

Table 7

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

Classifiers	CNN
	AlexNet		VGG19		ResNet18		Inception-v3		EfficientNet-B0
	Acc	Kappa	Acc	Kappa	Acc	Kappa	Acc	Kappa	Acc	Kappa
SVM	0.891	0.782	0.903	0.806	0.918	0.836	0.953	0.906	0.915	0.830
ANN	0.889	0.778	0.895	0.790	0.909	0.818	0.915	0.830	0.911	0.822
kNN	0.879	0.758	0.892	0.784	0.914	0.828	0.912	0.824	0.908	0.816
Random forest	0.881	0.762	0.897	0.794	0.901	0.802	0.909	0.818	0.908	0.816
Decision tree	0.874	0.748	0.867	0.734	0.899	0.798	0.901	0.802	0.901	0.802

The bold values represent the highest accuracies.