Fourier–Bessel Domain Adaptive Wavelet Transform-Based Method for Emotion Identification From EEG Signals

The letter presents a novel approach for analyzing a multi-sensor EEG signal with the aim of accurately identifying the emotional state of individuals. Identifying multiple classes of emotion using nonstationary EEG signals with good accuracy and efficiency is still an issue to address. The Fourier–Bessel domain adaptive wavelet transform is used to decompose EEG signals into various modes or components. To analyze the dynamics of modes Lyapunov exponent-based features are extracted from each mode. To classify feature values among different emotional classes namely, happy, sad, fear, and neutral, machine learning models have been used. To evaluate the performance of the proposed framework, EEG signals are recorded using ten distinct scalp sensors. EEG signals of 39 (20 males and 19 females) subjects were recorded. The proposed framework achieves an average classification accuracy of 96.91%. By incorporating emotion identification, human–system interaction can greatly enhance the user experience, as it improves engagement and contextual relevance.