EMG signal statistical features extraction combination performance benchmark using unsupervised neural network for arm rehab device

The paper presents important research to select the best features extraction for designing Arm Rehabilitation Device (ARD) for patient who had failure of their limb that highly beneficial towards rehab program. The device used to facilitate the tasks of the program is proved to improve the electrical activity in the motor unit and minimize the mental effort of the user. Electromyography (EMG) is the techniques to analyze the presence of electrical activity in musculoskeletal systems related to muscle movement. To prevent from the muscles paralyzed, it is becoming spasticity that the force of movements should minimize the mental efforts needed. To achieve this, the rehab device should analyze the surface EMG signal of normal people to be implemented to the rehab device. The EMG signal collected using non-invasive method is implemented to set the movements’ pattern of the arm rehab device. The signal are filtered and extracted for three time domain features of Standard Deviation (STD), Mean Absolute Value (MAV) and Root Mean Square (RMS). The features combinations are important to produce best classification result with reduced error. The best combination features for any movements, several trials of movements are used by determining the right combination using Self-Organizing Maps (SOM) for the classification process and this paper proved a proper combination will help to determine the best features combination in designing the best ARD