Human Hand Gesture Recognition System Using Body Sensor Network

Wearable technology has gained massive popularity for applications in healthcare in recent years. Body sensor network (BSN) includes wearable sensors such as Myoware muscle sensor that can capture electromyography (EMG) signal from human muscles during a movement. This study involves the control of a 3d prosthetic hand using EMG data of basic hand gestures from healthy subjects and operating the prosthetic hand with this classified data. Artificial Neural Networks (ANN) are biologically inspired computational networks which help to build predictive models from large databases. Backpropagation combined with feedforward ANN is used to classify data here. Classification results showed an accuracy of about 91% for four-hand gestures - hand closed, hand open, pointing index finger, and pinch motion. The BSN-based prosthetic hand can facilitate rehabilitation of amputees, perform surgery using recorded EMG data from a surgeon’s hands and reduce industrial manpower in the future.