Subject-Independent Emotion Recognition System from EEG Signals Using Continuous Wavelet Transform and Alexnet

Abstract

This study proposes a deep learning-based method for recognizing human emotions from Electroencephalogram (EEG) signals. To enhance the representation of EEG data, Continuous Wavelet Transform (CWT) has been employed to convert EEG signals into time-frequency images (scalograms). Alexnet model has been utilized for building subject-independent emotion classification system. The proposed approach was built as 2-stage recognition system; one for arousal and the other for valence. Data were imported from EEG AMIGOS dataset which includes four emotions; Calm, Fear, Happy, and Sad. The evaluation results have demonstrated a superior performance compared to the state-of-the-art methods in subject-Independent emotion recognition system from EEG signals, achieving average accuracy of 65.75% and 67.75% for arousal and valence respectively.