In order to obtain flexibility of gesture recognition system, a sequence of range image has been used for both erasing background noise and distinguishing gestures of which categories are depending on depth features. The essential processing was to spatially divide the space in front of the person into, for example, 3 by 3 by 3 voxels and calculate the ratio of hand area in each voxels. This paper newly introduces a tracking method for a walking person who is also moving.