Wrist Angle Estimation Under Different Loads Condition Based on Multi-Layer Perceptron Neural Network and Surface Electromyography Signals

This research focuses on the impact of load on wrist angle prediction using surface electromyographic (
$s$
EMG) signals from forearm muscles, specifically considering the flexion-extension of the wrist The goal is to develop a model capable of accurately predicting wrist angles under various conditions. A multilayer perceptron neural network (MLPNN) has been constructed and tested, processing features extracted from the mean absolute value (MAV) of
$s$
EMG signals. The experiments included a diverse simulation of load to examine the model's responsiveness to these forces. The findings indicate that the model can reliably predict wrist angles, evidenced by high R-squared values and low mean squared error (MSE) rates. This study suggests the potential for developing high-precision tools for human posture analysis and prediction that could be applied in a variety of situations, including research and potentially in physical therapy and prosthetic technologies in the future.