Gesture control by wrist surface electromyography

Surface electromyography (SEMG) systems are able to effectively sense muscle activity, irrespective of any apparent body motion, in a highly convenient and non-intrusive manner. These advantages make SEMG based systems highly attractive for use as a human computer interface. Despite such advantages, there are still a significant amount of challenges that should be resolved before such systems can be made viable. In this paper we focus on a wrist based SEMG system that is required to detect as well as recognize the gesture being made by the user. A major challenge in the detection of a gesture in an SEMG signal is the noise due to displacement of electrodes on the skin which does not belong to any of the well studied noise types. We use a bilateral filtering based approach to estimate such noise and then effectively detect the gesture signal. Next, we identify the gesture based on information contained in different frequency bands of the signal. Based on our experiments, we show that our system achieves an accuracy of 88.3% in identifying the correct gesture among rock, paper, and scissors gestures.