Table 4.

Details of music emotion recognition algorithms based deep learning method.

Method	Dataset	Features	Classifier/regressor	Performance
Keelawat et al. (2019)	12 recruited listened to 16 songs selected from MIDI	Segmented EEG	CNN	Accuracy of 78.36 and 83.67% in binary classification of arousal and valence, respectively.
Er et al. (2021)	Nine recruited listened to 16 audio tracks	Power spectrogram	Pretrained VGG16	Accuracy of 73.28% in quaternary classification.
Thammasan et al. (2016a)	15 recruited listened to 16 songs selected from MIDI	HFD, PSD, Discrete Wavelet Transform	Deep Belief Networks	Accuracy of 81.98% in binary classification of arousal and valence.
Rahman et al. (2020)	24 recruited listened to Twelve songs	DFA, Approximate Entropy, Fuzzy Entropy, Shannon's Entropy, Permutation Entropy, Hjorth Parameters, Hurst Exponent	Neuron Network	3 emotion scales (Depressing vs. Exciting and Sad vs. Happy and Irritating vs. Soothing).
Liu et al. (2022)	15 recruited listened to 13 music excerpts	Power spectrogram	Xception	Accuracy of 76.84% in HVHA vs. LVLA
Luo et al. (2022)	DEAP	PSD	RBF-SVM, LSTM	A SAM score of 6.17(high) and 4.76(low) in continuous valance scale, that is close to 6.98 and 4.36 evaluated in music database.
Hsu et al. (2018)	IADS	Segmented EEG	Neuron Network	MSE of 1.865 in 2D continuous SAM score.
Sheykhivand et al. (2020)	16 recruited listened to ten music excerpts	Segmented EEG	CNN, LSTM	Accuracy of 76.84% in HVHA vs. LVLA.
Li and Zheng (2021)	21 recruited listened to 15 music excerpts	Segmented EEG	Stacked Sparse Auto-Encoder	Accuracy of 59.5% and 66.8% in binary classification of arousal and valence, respectively.
Salama et al. (2018)	DEAP	Segmented EEG	3D CNN	Accuracy of 88.49% and 87.44% in binary classification of arousal and valence, respectively.