A Review on the Computational Methods for Emotional State Estimation from the Human EEG

Table 2

Emotional state estimation model.

Author	Year	Stimulus; (duration)	#channels	Channel location	Emotional states	EEG features	Feature extraction methods	Classifier	Accuracy (%)/#classes	Online versus offline
Li and Lu [107]	2009	Picture (6 sec)	62	IS	Happiness, sadness	Gamma ERD	CSP	SVM	93.5/2	Offline
Murugappan et al. [81]	2010	Video (−)	62	IS	Happiness, surprise, fear, disgust, and neutral	Delta, theta, alpha, beta, and gamma	Wavelet transform	kNN, LDA	83.04/5 (62 channels); 79.17/5 (24 channels)	Offline
Lin et al. [111]	2010	Music (30 sec)	24	Fp1-Fp2, F7-F8, F3-F4, FT7-FT8, FC3-FC4, T7-T8, P7-P8, C3-C4, TP7-TP8, CP3-CP4, P3-P4, O1-O2 (bipolar)	Joy, anger, sadness, and pleasure	Delta, theta, alpha, beta, and gamma	Short-time Fourier transform	SVM	82.29/4	Offline
Petrantonakis and Hadjileontiadis [70]	2010	Ekman’s picture set (−)	3	Fp1, Fp2,; F3/F4 (bipolar)	Happiness, sadness, surprise, anger, fear, and disgust	ERD/ERS	Higher-order crossing	QDA; SVM	62.3/6 (QDA); 83.33/6 (SVM)	Offline
Hosseini et al. [112]	2010	IAPS (3 sec)	5	FP1, FP2, T3, T4, Pz	Calm-neutral, negative-excited	Delta, theta, alpha, beta, and gamma	Wavelet coefficients, Higuchi’s algorithm	Elman; ANN	82.7/2	Offline
Wang et al. [113]	2011	Video (4 min)	62	IS	Joy, relax, sad, and fear	Alpha, beta, and gamma	Minimum redundancy maximum relevance method	SVM	66.5/4	Offline
Brown et al. [24]	2011	IAPS (6 sec)	8	Fp1, Fp2, F3, F4, F7, F8, C3, C4	Positive, negative, and neutral	Alpha1 (6–8 Hz),; Alpha2 (8–10 Hz), Alpha3 (10–12 Hz)	Spectral power features	kNN	85.0/3	Online
Petrantonakis [69]	2011	IAPS (5 sec)	8	F3, F4,; C3, C4,; T3, T4,; P3, P4	Arousal, valence	Alpha and beta power	Asymmetry index	SVM	94.4/2	Offline
Stelios and Hadjidimitriou [114]	2012	Music (60 sec)	14	AF3, F7, F3, FC5, T7, P7, O1, O2, P8, T8, FC6, F4, F8, AF4	Arousal, neutral	Beta and gamma	Time-frequency (TF) analysis	SVM; KNN; QDA; MDA	86.52/2	Offline
Konstantinidis et al. [22]	2012	IAPS (2D emotional space)	19	IS	Arousal, valence	ERP (N100, N200)	Amplitude of ERP	SVM	81.3/4	Online

(i) SVM: support vector machine; kNN: k-nearest neighbor; LDA: linear discriminant analysis; QDA: quadratic discriminant analysis; MDA: Mahalanobis distance based discriminant analysis; SOM: self-organization map; ANN: artificial neural networks.
(ii) IS: channel configuration followed the 10/20 International System.