Gesture and posture estimation by using locally linear regression

Many methods for measuring human motion have been proposed because they can be applied to various fields such as CG animation, virtual reality, and gesture recognition. This paper proposes a method for gesture and posture estimation which uses locally linear regression. We extract face and hand regions from stereo input images, and estimate their 3D positions by stereo matching. Performed gestures are estimated by hand position and velocity, and human posture is also estimated by using locally linear regression.