Hand gesture recognition using codebook model and Pixel-Based Hierarchical-Feature Adaboosting

This paper presents an approach for hand gesture recognition that can be employed to enhance the capability of existing applications, especially in sign language communication. For practical use, the hand posture is taken at the back instead of the front and occurred under unexpected background environment. Unlike the front-hand, the back hand view image is less information than the front-viewed. Thus, the recognition among lack of information is the challenge of this task. Codebook-based foreground detection model is used to detect the hand region under an unexpected background environment. Moreover, the Pixel-Based Hierarchical Feature method is proposed to extract the importance features which are further classified by Adaboosting that yields a high recognition rate. For performance evaluation, we have applied perturbation recognition rate analysis of five alphabet patterns and the experimental results shows that the proposed method provides higher recognition accuracy than existing method.