Emotion recognition using EEG signal and deep learning approach

Abstract

Emotion is a mental state, which originates in the brain and is closely related to the nervous system. Emotion can be defined as a feeling expressed through, or detectable by voice intonation, facial expression body language, as response from one’s mood relationship with others and most importantly the circumstance they are in. Although, Brain Computer Interface (BCI) are being developed to find a better human-machine interaction system using brain activity and it is frequently implemented by Electroencephalogram (EEG) signals. EEG is a well established approach to measure the brain activities which can be analyzed and processed to distinguish different emotions. In this thesis, we present an approach to classify human emotions using EEG signal by Convolutional Neural Network(CNN). In our model, we use the Dataset for Emotion Analysis using Physiological signals (DEAP) dataset, a benchmark for emotion classification research, to transform the EEG signal from time domain to frequency domain and extract the features to classify the emotions. Emotion can be classified based on the two dimensions of valence and arousal. Previous researches have used fewer channels and participants. Our approach which was carried out on 32 participants, has achieved an accuracy of 94.75% for the valence and 95.75% on the arousal detection, which is quite competitive with other methods of emotion recognition.