Table 9. Comparison of SI experiments with existing methods.

Database	Reference	Feature	Accuracy(%)
RAVDESS	Proposed approach	AlexNet + FS + MLP	82.75
RAVDESS	Proposed approach	AlexNet + FS + SVM	80.94
SAVEE	Sun & Wen (2017)	Ensemble soft-MarginSoftmax (EM-Softmax)	51.50
SAVEE	Haider et al. (2021)	eGeMAPs and emobase	42.40
SAVEE	Proposed approach	AlexNet + FS + MLP	75.38
SAVEE	Proposed approach	AlexNet + FS + SVM	70.06
Emo-DB	Badshah et al. (2017)	DCNN + DTPM	87.31
Emo-DB	Sun & Wen (2017)	Ensemble soft-MarginSoftmax (EM-Softmax)	82.40
Emo-DB	Yi & Mak (2019)	OpenSmile features + ADAN	83.74
Emo-DB	Guo et al. (2019)	Statistical features and empirical features + KELM	84.49
Emo-DB	Meng et al. (2019)	Dilated CNN + BiLSTM	85.39
Emo-DB	Haider et al. (2021)	eGeMAPs and emobase	76.90
Emo-DB	Lech et al. (2020)	AlexNet	82.00
Emo-DB	Mustaqeem, Sajjad & Kwon (2020)	Radial basis function network( RBFN) + Deep BiLSTM	85.57
Emo-DB	Proposed approach	AlexNet + FS + MLP	92.65
Emo-DB	Proposed approach	AlexNet + FS + SVM	90.78
IEMOCAP	Xia & Liu (2017)	SP + CNN	64.00
IEMOCAP	Chen et al. (2018)	Dilated CNN + BiLSTM	69.32
IEMOCAP	Guo et al. (2019)	Statistical features and empirical features + KELM	57.10
IEMOCAP	Yi & Mak (2019)	OpenSmile features + ADAN	65.01
IEMOCAP	Daneshfar, Kabudian & Neekabadi (2020)	IS10 + DBN	64.50
IEMOCAP	Mustaqeem, Sajjad & Kwon (2020)	Radial basis function network( RBFN) + Deep BiLSTM	72.2
IEMOCAP	Proposed approach	AlexNet + FS + MLP	89.12
IEMOCAP	Proposed approach	AlexNet + FS + RF	86.23