Hands Detection Based on Statistical Learning

This paper proposes a hand detection methodbased on statistical learning training way. Using Microsoft´s Kinect sensor, to get the depth information. Through the analysis of the characetristics of hands, put out a kind of new features for statistical learning which approximate with Harr-like feature. The new feature is good at describing complex hand shape degeneration. With the help of Adaboost statistical learning, gets the training model. Experiment results demonstrate that using the new features with Adaboost algorithm can achieve more rapid and robust hands detection system.