An empirical approach on Hand Gesture Recognition based using Convolution Neural Network

Abstract

Hand Gesture Recognition is becoming one of the innovative approaches to work together with automated systems with high precision. The hand gesture mechanism helps to interact in a natural way with the machines to communicate & instruct them accordingly. In this paper, we have designed a dynamic and static hand gesture recognition approach using convolution neural networks. The gesture was recorded using a local webcam. Extracted frames were processed using gaussian thresholding then CNN was used with defined parameters. For this experimentation, 4 classes were considered that includes 3200 images (800 each). The performance measure of 84.8% accuracy was achieved in this approach for the defined four gesture i.e. shape C gesture, Bye action, Ok action and forward action.