Skip to Main Content
To the goal of multiple degree of freedom (M-DOF) prosthetic hand control by characterizing events in surface electromyograms (sEMG), a system of sEMG acquisition, detection and recognition is built. It can classify commonly used eighteen kinds of hand gestures with six channels. The dynamic cumulative sum (DCS) is applied for on-line detection. Energy changes are detected with this approach, corresponding to finding the beginning and end of movements. In classification strategy, the sample means value (s.m.v) is used for extracting signal features, and three parallel back propagation(BP) neural networks are used for pattern recognition. For the system to be efficient and simple, we don't classify eighteen kinds of hand gestures directly, instead, to reasonably plan these movements according to fingers conditions. Then hand gestures are obtained according to fingers states. This method makes the recognition dimension reduced from eighteen to three and decreases recognition complexity. The results show that, in average, over 96% of the movements are correctly detected and classified. The proposed system may facilitate the development of prosthetic hand control strategies using sEMG.