EMG based classification of thumb posture using portable thumb training system

Loss of human limbs that are caused by traumatic accidents, vascular diseases and diabetes that lead to amputation has great impact to the well-being of the affected segment of societies. They usually require prostheses to restore the original functionality of the missing limbs so to be able to assist them in the activities of daily living. Extensive works have been reported in developing prostheses that not only could work as close as the natural limbs but also look alike the original ones. Many of these prostheses are based on the myoelectric control which requires the electromyography (EMG) signal to be measured and analyzed from the remaining nerves or muscles left after amputation. It is still a challenge however to detect, process, classify and apply the signal appropriately. In this paper, a study on the relationship of thumb-tip force related to EMG signals using a well-designed, portable thumb training system is conducted. The EMG signal is classified by using machine learning techniques when the thumb is flexed at different angles and exerts different magnitude of forces at its tip. The thumb training system is developed with compliances to fit different shape and size of human hands. The result from the analysis done using WEKA software shows that the EMG signal can be used to estimate the posture of the thumb at different flexed angles and thumb tip forces