Enhancing ASD Cognitive Assessment with P300 Classification using EMD-based qEEG Wavelet Features

Individuals with Autism Spectrum Disorder (ASD) often face challenges in effective communication due to deficiencies in their joint attention skills. Brain-Computer Interface (BCI) systems based on the P300 Electroencephalogram (EEG) signal has shown promising results in neuro-rehabilitation training for individuals with Autism spectrum disorder(ASD), helping them overcome these communication barriers. However, detecting the P300 signal in ASD is particularly difficult due to its noisy nature, lower amplitude, and higher latency compared to other individuals. In this paper, we propose a novel application of the Empirical mode decomposition (EMD) using the Hilbert Huang transform to identify the P300 signal in a BCI system involving ASD subjects. Further, the optimized mode was selected and utilized and the qEEG fused time-frequency domain wavelet features were extracted for each selected mode of each individual channel. To address the class imbalance issue in the dataset, To handle imbalanced data, we applied SMOTE for data augmentation, and for optimized feature selection, we used PCA. We conducted a comparative classification analysis using five popular machine-learning classifiers. Among the classifiers, Random Forest outperformed the others, achieving 93% accuracy. These results demonstrate the effectiveness of our approach in enhancing P300 signal identification for joint attention-based cognitive task classification for individuals with ASD, surpassing state-of-the-art methods.