<![CDATA[Occupational musculoskeletal disorders (MSDs) often result from prolonged non-ergonomic postures, especially in educational and industrial bench work activities. This study presents an approach to classify ergonomic and non-ergonomic working postures using surface electromyography (sEMG) signals and machine learning. sEMG data were recorded from four upper limb muscles during simulated bench work conditions. Time-domain and frequency-domain features were extracted from segmented EMG signals using sliding windows. Dimensionality reduction was performed using Principal Component Analysis (PCA), and classification was carried out using logistic regression. The proposed system achieved an overall classification accuracy of 75% in distinguishing ergonomic and non-ergonomic postures. Visualization using PCA and Linear Discriminant Analysis (LDA) showed clear class separation, validating the discriminatory power of the extracted features. While the small sample size and class imbalance were identified as limitations, the study demonstrates that a simple and interpretable model like Logistic Regression, when combined with proper feature engineering, can yield promising results.This work contributes to the development of low-cost, efficient, and interpretable ergonomic assessment tools. It is particularly relevant for vocational and educational environments where real-time posture monitoring and early prevention of MSDs are essential. Future research should focus on expanding the dataset, exploring deep learning methods, and implementing real-time wearable systems.]]>
Copyrights © 2025
