Motion Gesture Detection and Tracking Algorithm Based on Compressive Sensing

Abstract

Real-time gesture detection and tracking algorithm is proposed to solve the problems of detection and tracking of gesture under the complex background. Firstly, an Adaboost cascade classifier is used to track by the feature model and classifier, which are constructed by real-time compressive tracking algorithm. The negative factors from the gesture posture, shade, skin color etc. are eliminated to improve the performance of the gesture detection and tracking by fusing the responses of classifier of real-time compressive tracking algorithm and the results of gesture detection based on Adaboost algorithms. Paper proposed algorithms can self-recovery when the tracking object is missed, so continuous recognition and tracking is guaranteed. Finally, several experiments are given to verify the developed algorithm and to demonstrate its practicality and effectiveness.