Estimating human arm´s muscle force using Artificial Neural Network

Muscle force models have many applications in human-machine motion analysis, human-machine interfacing, rehabilitation and robotics. A Hill-type model was used to estimate muscle force. In this work, we are going to introduce a new model to estimate human muscle force. Our model, estimates human muscle force based on a rectified smoothed electromyography (RSEMG) signal using the back-propagation Artificial Neural Network (BPANN) method. BPANN is used to process raw EMG signals to estimate muscle force. Our results, demonstrate that the new model improves the accuracy the estimation. The proposed BPANN model can efficiently extract muscle force features from (EMG) signals. It is fast and easy to implement. Our results showed that the regression of our neural model exceeded 99%. We used the mean square error (MSE) to measure the performance of our network. The MSE result of our neural model was very small, as we expected. Our work proves that mathematical models can be ignored if BPANN algorithms are used.