Investigation of Human Emotion Pattern Based on EEG Signal Using Wavelet Families and Correlation Feature Selection

Abstract

Human emotion is transcendence for human beings given by God compared to other living creatures. Emotion has a significant role in human life. There were many studies done to recognize human emotion using physiological measurement, Electroencephalograph is one of the methods. Yet, the previous studies have not been discussed the wavelet families which has the best performance and optimal channel in human emotion recognition. This research used the power feature from some wavelet family, i.e. Daubechies, symlet, and coiflet with CFS feature as the selection method to choose the superlative feature among alpha, beta, gamma, and tetha frequencies. Based on this research results, the coiflet method has the most accurate value on recognizing the emotion among the wavelet families. The use of CFS feature selection method could increase the accuracy from 81% to 93%, also obtained five most dominant channel, i.e. T8, T7, C5, CP5, and TP7 on power feature of alpha and gamma frequencies so the temporal part of the left brain is more dominant in recognizing human emotion.