Abstract

Electromyography (EMG) signals provide critical insights into muscular and neurological functions, but their complex nature makes accurate classification and outlier detection challenging. Traditional signal processing approaches often fail to address the variability in EMG signals, leading to suboptimal data interpretation. The integration of advanced algorithmic innovations, such as K-Nearest Neighbors (KNN) kernel-based Support Vector Machine (SVM), offers a robust solution for enhancing EMG signal processing. In this study, EMG signals from 500 datasets, sampled at 2 kHz, were preprocessed using wavelet transform for noise reduction and feature extraction. A hybrid KNN-SVM model was employed to classify the data and identify outliers, achieving superior performance. Results indicate a classification accuracy of 97.8%, sensitivity of 96.5%, specificity of 98.3%, and an outlier detection precision of 95.2%. These findings underscore the potential of the KNN kernel-based SVM approach in improving EMG signal interpretation, enabling accurate diagnosis and monitoring in clinical and research settings. The proposed methodology demonstrates a significant advancement in EMG signal processing, ensuring reliable classification and precise outlier detection.