Improving the Accuracy of Upper Limb Exoskeleton Movements Using Machine Learning Based on EMG for Stroke Patient Rehabilitation

Post-stroke rehabilitation often focuses on restoring upper limb mobility, which is critical for regaining independence in daily activities. Upper limb exoskeletons are increasingly used as assistive devices for rehabilitation, enabling controlled, repetitive movements to aid recovery. However, accurate control of these devices remains challenging. This study explores the application of machine learning (ML) to improve the accuracy of upper limb exoskeleton movements by utilizing electromyography (EMG) signals. The system integrates Raspberry Pi Zero 2W, Muscle Sensor V3, and MPU6050 to capture and process EMG signals, extracting features such as Root Mean Square (RMS), Mean Absolute Value (MAV), and Variance. These features are then used to train a Random Forest Regression model to predict joint angles, enhancing exoskeleton movement accuracy. The results demonstrate that the combination of the MAV feature and Random Forest Regression achieved the highest accuracy, with an RMSE of 12.197 and an R² of 91.6%. The exoskeleton system successfully predicts and follows the patient’s intended movements, providing real-time adjustments based on EMG data. This innovation is expected to improve the rehabilitation process for stroke patients by offering personalized, precise movement control. With further development, this system has the potential to significantly enhance the effectiveness of upper limb rehabilitation.