Real time elbow joint angle estimation using sEMG signals

The Surface Electromyographic (sEMG) signal is convenient for prosthetic device control due to its non-invasive acquisition and its intrinsic relation to the user's intention. This thesis presents an algorithm for estimation of the real time elbow joint angle from sEMG signals acquired from the muscles of biceps and triceps. The algorithm developed in the thesis uses time-domain feature extraction methods such as mean absolute value (MAV), root mean square (RMS) and waveform length (WL). Estimation of the joint angle using extracted sEMG features is performed by Artificial Neural Networks (ANN), specifically a Multilayer Perceptron (MLP) and a general regression neural network (GRNN). The overall system is implemented and tested in a real time hardware setup. The results indicated that developed method could be successfully used for a prosthetic arm device posture control.