EEG-based human emotion recognition using entropy as a feature extraction measure

Brain Informatics

Table 1 Summary of studies conducted on EEG-based emotion recognition using entropy as a feature

Reference	No. of subjects	Emotions	Features	Database	Classifier	Accuracy
[30]	3 men, 3 women	Positive, negative	ES, DE, DASM, RASM	Private	SVM KNN	76.56% 84.22%
[77]	–	Positive, negative	Sample entropy	–	SVM weight classifier	85.11%
[51]	16 men, 16 women	Arousal, valence	Wavelet entropy	DEAP	SVM	65%
[29]	7 men, 8 women	Positive, neutral, negative	Dynamic sample entropy	SEED	SVM	64.15%
[50]	6 men, 7 women	Positive, neutral, negative	Power spectral entropy, correlation dimension	Private	SVM	79.58% 82.58%
[76]	5 men	Happy, neutral, disgust	RAQA; Shannon’s entropy and 5 others	eNTERFACE06_EMOBRAIN	Multilayer perception	36%
					Time-delay neural network	36%
					Probabilistic neural network	99.96%
[75]	5	Happy, sadness, fear	RAQA; entropy and 5 others	Private	SVM	92.24%
[3]	16 men, 16 women	Excitement, happiness, sadness, hatred	Shannon’s entropy and 3 others	DEAP	Multiclass SVM	94.097%
[33]	5 men, 5 women	Happy, calm, sad, fear	EMD approximate entropy	Private	Integration of deep belief network and SVM (DBN-SVM)	87.32%
[4]	16 men, 16 women	Excitement, happy, sadness, hatred	Approximate entropy, K-S entropy, permutation entropy, singular entropy, Shannon’s entropy	DEAP	SVM	59.8%
[4]	7 men, 8 women	Positive, neutral, negative	Spectral entropy and 12 other nonlinear entropy methods	SEED	SVM	83.33%
[71]	16 men, 16 women	2 and 3 level of labeling in arousal and valence space	Multiscale fuzzy entropy	DEAP	SVM	2-class 90.81% (A) 90.53% (V) 3-class 79.83% (A) 77.80% (V)
[74]	16 men, 16 women	HAHV, HALV, LAHV, LALV	EMD Sample entropy	DEAP	SVM	94.98% (binary class) 93.20% (multiclass)