Analysis of EEG Features in Normal and ASD Conditions Based on SWT Using Fisher Linear Discriminant Analysis

Background/Objectives: Autism Spectrum Disorder (ASD) diagnosis is challenging due to the complexity of Electroencephalogram (EEG) signals. This study aims to enhance the accuracy of ASD diagnosis by combining Fisher Linear Discriminant Analysis (FLDA) and Stationary Wavelet Transform (SWT) for effective classification of EEG signals. Methods: EEG signals were filtered and decomposed into frequency bands using SWT. Level 3 (gamma), level 4 (beta), and level 6 (theta) components were extracted. FLDA was used to lower dimensionality and enhance class separation between ASD and normal EEG. Classification accuracy was evaluated using a confusion matrix. Results: The combined approach of FLDA and SWT achieved a classification accuracy exceeding 90%, with 96% accuracy in classifying level 6 (theta) components in the normal class, demonstrating the method's potential to enhance early detection of ASD. Conclusions: Integrating FLDA and SWT in EEG analysis offers a novel and effective approach to enhancing ASD diagnosis accuracy and efficiency. The method offers a reliable tool for supporting early ASD diagnosis in clinical settings.