A Dynamic Hand Gesture Recognition Algorithm Using Codebook Model and Spatial Moments

Nowadays, with the rapid development of intelligent hardware, the hand gesture interaction is becoming more and more popular in the field of Human-Computer Interaction (HCI). In this paper, we present a dynamic hand gesture recognition algorithm by using our improved Codebook (CB) modeling method and spatial moments. In the pre-processing stage, we train our CB background model with a video of short sequence of frame. Hand region is segmented by background subtraction with skin color detection in YCbCr color space. To extract hand features, we first find palm center and its major axis by calculating spatial moments of hand contour. Open fingers are then tracked by an improved curvature-based algorithm. In experiment, we design an analysis of performance on hand gesture recognition and the results show that our hand gesture recognition is satisfactory with 95.5%.