An Analytical Approach of EEG Analysis for Emotion Recognition

Emotion is the fundamental behavioral attributes of humans. To identify emotional variations from Electroencephalogram signals have currently expanded consideration amid BCI researchers. In this work, emotion recognition from EEG is performed using 21channel EEG acquisition device employing 10–20 method of electrode placement. The experiment being performed on issues of the peer group of 20–25 years of 16 university students (eight females and eight males). Audio-visual stimuli are used for bringing four dissimilar emotions (Happy, Sad, Fear and Relaxed) and corresponding signals are processed for emotion classification. At first EEG signals are filtered using Butterworth 4th order filter which is band limited by 0.5-60 Hz after that smoothened with the help of Surface Laplacian filter. Filtered EEG signals are feature extracted using Power Spectral Density, Wavelet Decomposition, Hjorth Parameter and AR parameter. After that Linear SVM classifier is used. Support Vector Machine classifier generates the best result when used with Wavelet coefficient feature extraction technique (96.81%). The experimental result also shows the diminutive interval EEG can be used for sensing the emotional thought variations effectively. We found that the EEG signals contained adequate information to separate four different emotion classes.