Depth Static Hand Gesture Recognition

Abstract

Finger-spelling recognition is the popular way to realize Human-Computer Interaction (HCI). The sign language datasets for finger-spelling recognition are mostly collected by the color camera and the depth camera. The depth camera directly captures the 3D information between the performer and lens, which can extract the precise hand gesture texture information without being influenced by the complex background and illumination. This work develops a hand gesture recognition system using depth image. This system employs a VGGNet to learn the feature extraction and classification. The experimental results show that the proposed hand gesture recognition system can correctly recognize the hand gesture using only depth image. This may reduce computational cost of the hand gesture recognition system.