Recognizing emotions from EEG subbands using wavelet analysis

Abstract

Objectively recognizing emotions is a particularly important task to ensure that patients with emotional symptoms are given the appropriate treatments. The aim of this study was to develop an emotion recognition system using Electroencephalogram (EEG) signals to identify four emotions including happy, sad, angry, and relaxed. We approached this objective by firstly investigating the relevant EEG frequency band followed by deciding the appropriate feature extraction method. Two features were considered namely: 1. Wavelet Energy, and 2. Wavelet Entropy. EEG Channels reduction was then implemented to reduce the complexity of the features. The ground truth emotional states of each subject were inferred using Russel's circumplex model of emotion, that is, by mapping the subjectively reported degrees of valence (pleasure) and arousal to the appropriate emotions - for example, an emotion with high valence and high arousal is equivalent to a `happy' emotional state, while low valence and low arousal is equivalent to a `sad' emotional state. The Support Vector Machine (SVM) classifier was then used for mapping each feature vector into corresponding discrete emotions. The results presented in this study indicated thatWavelet features extracted from alpha, beta and gamma bands seem to provide the necessary information for describing the aforementioned emotions. Using the DEAP (Dataset for Emotion Analysis using electroencephalogram, Physiological and Video Signals), our proposed method achieved an average sensitivity and specificity of 77.4% ± 14.1% and 69.1% ± 12.8%, respectively.

Multiple Transferable Recursive Feature Elimination Technique for Emotion Recognition Based on EEG Signals

“…By using a fast Fourier transformation, the frequency features (60 power features, 16 power difference features) were prepared. In each channel, the power features were computed on four frequency bands, i.e., theta (4-8 Hz), alpha (8-12 Hz), beta (12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30) and gamma (30)(31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45). Power difference features were employed to detect the variation in cerebral activity between the left and right cortical areas.…”

EEG emotion recognition using reduced channel wavelet entropy and average wavelet coefficient features with normal Mutual Information method

“…Following the recommendation in [4] only 3 subbands were employed utilizing only α, β, and γ bands to reduce the length of array which also an initial reduction of features complexity.…”

“…Using Ensemble Rapid Centroid Estimation (ERCE) [16], [17] the kernel radius R SVM is estimated from the training data with the estimation process as in [4]. Sequential Minimal Optimization (SMO) algorithm is implemented in the classification process to train the SVM [18].…”

“…• The wavelet transform (WT) is both a band-pass filter and a denoiser for decomposing and isolating EEG signals to obtain desired subbands, for example, extracting only alpha, beta, and gamma frequencies as they are more related to emotion elicitation [4].…”

“…Recently, Normalized Mutual Information (NMI) has been proposed for feature selection with the advantage of reduced complexity of features [9]. The implementation of NMI algorithm in EEG signal processing offers the possibilities to measure the relation between features in EEG channels and their represented emotions [4].…”