Wireless Robot Control Using Wrist Movements from Surface Electromyogram Signal

Nowadays, the rates of disability are increasing in part due to aging populations and an increase in chronic health conditions. Human-Computer Interaction (HCI) system become vital for helping the temporary and permanently disabled patients. In this framework, sEMG signals of five motions which are neutral, left, right, up and down will be performed and recorded from subjects by using ADInstruments PowerLab 4/25T device. Then, the signal will be filtered with 4th order band-pass Butterworth filter and an 50 Hz notch filter used to remove the noise of the power line. Empirical Mode Decomposition (EMD) method and wavelet packet decomposition method were implemented for extracting the 11 features. Classifiers used are probabilistic neural network (PNN) and generalized regression neural network (GRNN) in order to analyze the performance of controlling the wireless patient remote system prototype via wrist sEMG signal. From the results obtained, PNN is the best classifier with the highest recognition rate of 87.0% by using the 10% of testing data and 90% of training data compared with GRNN which is 86.25%. After that, the graphical user interface (GUI) will be connected with the Blynk app to show the IoT platforms that allows patients for controlling the robotic system.