Hand Gesture Recognition Using Deep Network Through Trajectory-to-Contour Based Images

Vision-based hand gesture recognition involves visual analysis of hand shape, position and/or movement. Most of the previous approaches require complex gesture representation as well as selection of robust features for proper gesture recognition. In this paper, a model-based method for hand gesture recognition has been presented using convolutional neural network. The model is fed with trajectory-to-contour based images obtained from isolated trajectory gesture through segmentation and tracking the hand motion, thereby estimating the hand motion trajectory for recognition. Conventional methods can extract low-level features, while deep learning approaches learn image features hierarchically from local to global with multiple layers of abstraction from vast number of sample images. Feature learning capability of CNN architecture has given outstanding results on three different datasets.