2D motion detection bounded hand 3D trajectory tracking and gesture recognition under complex background

In this paper, a 2D motion detection bounded hand 3D trajectory tracking and gesture recognition system is proposed for virtual reality interactions. First, the Bayes decision rule for classification of background and foreground is utilized to automatically locate the hand that bounded within a rectangle, and then the trajectory of the hand in 3D space is tracked by mean shift particle filter and stereo imaging. The skin color feature is exploited for image matting that effectively segment the hand contour in video sequence automatically. Finally the hand gesture is recognized by the connected component analysis and line approximation. The proposed technique works without any markers or constraints, overcomes the disturbance of arms and faces in the scene, and can recognize multiple hands with different gestures under complex background.