Hand Gesture Detection Using Convexity Hull and Convolutional Neural Network

Abstract

Since the beginning of mankind, communication has always played a key role in everyday life and at the same time has pushed mankind to achieve targets for more convenient modes of communication. From the invention of the telephone to us having access to touch screen smartphones - we have come a long way. Unfortunately, we haven't been able to catch up at the same pace for people with speech impairment. While multiple studies on sign language recognition have been performed, there are still multiple constraints and limitations which are to be tackled. In this paper, application of convexity hull algorithm on hand gesture recognition has been proposed, where we further discuss the role of convexity defect and application of cosine rule in convexity hull. However, different skin tones and hand signs with similar gestures can cause false prediction thus to combat these limitations, CNN model has been proposed through which accuracy of 90% was achieved.