There are many researches for head detection and tracking. Most of them are based on RGB images or RGB-D images. However, RGB based methods are sensitive to skin color and lightness. And RGB-D based approaches have a problem of calibration. Also for head tracking, a problem of occlusion by hands should be solved necessarily. To this end, we propose a depth-based head detection and tracking method. Also, using sub-block matching method we solve the problem of occlusion by hand.