Hidden-Markov-Models-Based Dynamic Hand Gesture Recognition

This paper is concerned with the recognition of dynamic handgestures. A method based on Hidden Markov Models (HMMs) is presentedfor dynamic gesture trajectory modeling and recognition. Adaboostalgorithm is used to detect the user's hand and a contour-based handtracker is formed combining condensation and partitioned sampling. Cubic B-spline is adopted to approximately fit the trajectory pointsinto a curve. Invariant curve moments as global features andorientation as local features are computed to represent thetrajectory of hand gesture. The proposed method can achieveautomatic hand gesture online recognition and can successfullyreject atypical gestures. The experimental results show that theproposed algorithm can reach better recognition results than thetraditional hand recognition method.