Pattern Recognition of Electromyography Signal in Upper Limb Movement Using Supervised Machine Learning: Ipsilateral Hair Combing Activity

Abstract

Surface electromyography (sEMG) provides valuable insights into muscle activation and force generation, especially in tasks involving upper limb movement. However, identifying distinct signal patterns that reliably represent muscular force remains a significant challenge due to variability in physiological signals and motion complexity. This study aims to investigate the relationship between EMG signal features and muscle force exertion during a functional arm movement, specifically, ipsilateral hair combing (IHC) activity, among healthy subjects. EMG data were acquired from deltoid, biceps brachii and palmaris longus during IHC activity. There were 16 healthy participants (male: 5; female: 11; average age: 39.1 years; weight: 72.9 kg; height: 160.7 cm) performed IHC activity while EMG signals were recorded using the EMG data acquisition system (Delsys Bagnoli-8). IHC activity were recorded using the motion capture system (Vicon-Nexus 1.8.1). Four time-domain features such as mean absolute value (MAV), root mean square (RMS), variance (VAR) and maximum amplitude (MAX) were extracted and used as input for classification. Four supervised machine learning models such as Naïve Bayes (NB), Support Vector Machine (SVM), Random Forest (RF) and k-Nearest Neighbors (k-NN) were applied. Model performance was evaluated using confusion matrix, accuracy and F1 score. The pattern recognition framework involved four stages: signal acquisition, feature extraction, model training and testing. The analysis demonstrated a clear correlation between changes in EMG signals and the corresponding muscular force exerted during arm movement phases. Among the classifiers, SVM achieved the highest recognition accuracy which is 83.5% and F1 score of 0.817, indicating its superior capability in distinguishing muscle activation patterns. This study establishes a promising approach for muscle force classification using EMG signals during upper limb movement specifically in IHC activity. The results offer practical implications for assistive rehabilitation systems, especially in monitoring and supporting patients recovering from neurological injuries.