Comparative Study of Blind Source Separation (BSS) Techniques for ASD Classification in EEG Signals

This work compares preprocessing, feature extraction, and classification techniques applied to EEG signals from individuals with ASD and control subjects. The FastICA and SOBI algorithms, different windowing strategies, mRMR-based feature selection, and SVM-based classification were evaluated. The results indicated more evident differences in the delta, alpha, and gamma bands, and showed that methodological choices directly influence model performance.