<![CDATA[Purpose – This study aims to analyze and compare the performance of a Multi-Layer Perceptron (MLP), a One-Dimensional Convolutional Neural Network (1D CNN), and a hybrid model for classifying three elbow joint angle movements using surface electromyography (sEMG) signals. Design/methods/approach – This dataset consists of sEMG signals collected from 15 healthy participants performing three elbow joint angle movements at 45°, 90°, and 135° with 50 repetitions, for a total of 2,250 data points. The proposed MLP model uses time-domain features extracted using Root Mean Square (RMS) and a Kalman filter, while a 1D CNN learns features from the raw, segmented signals from Shield EMG. A hybrid model combines both features. Model performance is evaluated using accuracy, precision, recall, F1 score, and confusion matrix. Findings - The results show that the 1D CNN model achieved score accuracy 0.78, outperforming the MLP model with accuracy of 0.65 , indicating superior feature learning from the raw sEMG signal. The hybrid model achieved accuracy of 0.82 and was more stable in discriminating intermediate elbow joint angles, indicating that feature fusion improves classification reliability. Research implications/limitations – This study was limited by the relatively small number of participants and the lack of an external validation dataset, which may impact the generalizability of the results. Future research should include larger and more diverse populations and explore more advanced architectures. Originality/value – This study provides a comparison between MLP, 1D CNN, and hybrid model approaches for sEMG-based elbow joint angle classification, highlighting the strengths of each method and offering insights for the development of robust rehabilitation technologies.]]>
Copyrights © 2026
