Enhancing Emotion Recognition in EEG Signals using Fractional Fourier Transform

Electroencephalogram (EEG) signals are a valuable tool for emotion recognition due to their effectiveness. However, their non-linear and non-stationary nature often leads to suboptimal outcomes when using traditional single-feature extraction methods. To address these challenges, this paper employs the fractional Fourier transform (FRFT) method, which is well-suited for processing non-stationary signals, for emotion recognition in EEG signals. First, the FRFT was applied to convert EEG signals into time-frequency representations at various fractional orders. Seven features were then extracted from the EEG data at each order: maximum value, mean value, width, energy, power, variance, and differential entropy. Finally, the SVM method was used for the identification of emotions based on the EEG signals. The approach was validated using open EEG datasets (DEAP and SEED). Experimental results demonstrate that the optimal emotion recognition performance is achieved when the fractional order is set to 0.4.