3D Hand Tracking and Motion Analysis with a Combination Approach of Statistical and Syntactic Analysis

This paper describes a new approach to solve the problem of real-time 3D hand tracking and motion analysis with a combination approach of statistical and syntactic analysis. The fundamental idea is to divide the problem into two levels according to the hierarchical property of hand gestures. The lower level of the approach implements posture detection and tracking with a statistical method based on Haar-like features and the AdaBoost learning algorithm. With this method, a group of hand postures can be detected and tracked in 3D mode in real-time. The robustness against cluttered backgrounds is achieved by background subtraction and smoothing. The higher level of the approach implements hand motion analysis using the syntactic analysis based on stochastic context-free grammars. Two structured gestures are analyzed in our experiment. Given an input string, based on the stochastic parsing, the corresponding structured gesture can be identified by looking for the SCFG that has the higher probability to generate the input string.