Analysis of EEG Signals for Emotion Recognition Using Different Computational Intelligence Techniques

Abstract

Emotion is a very essential aspect in day-to-day life. Emotion can be analyzed by using facial expression, gesture, verbally, and many other ways. But there are some demerits in this technique, so Electroencephalography (EEG) signal is used for recognition of emotion. The most important role of wavelet transform is to remove the noise from the biomedical signals. Analysis of EEG signals using computational intelligence technique like discrete wavelet transform and bionic wavelet transform is presented in this paper. A new modified wavelet transform called Bionic Wavelet Transform (BWT) has been applied here for analysis of biomedical signals. By adapting value of scales, T-function of bionic wavelet transform is varied and its effects on the value of the threshold are noticed. This is called the BWT which is used for emotion recognition using EEG signals. For classification purposes, different classifiers, i.e., Artificial neural network (ANN), k-nearest neighbor (K-NN), Naïve Bayes, and support vector machine (SVM) are presented in this paper. From the proposed algorithm, i.e., with BWT, it is observed that the emotion is better classified than with WT. EEG data are taken from enterface06_emobrain dataset in which there is having the dataset of two subjects which are applied to evaluate the performance of the proposed classifier. In order to find the best method for denoising, signal-to-noise ratio is calculated for different emotions of EEG signal and it is observed that BWT removes the noise better from EEG signal than WT.