Recognition and Segmentation of 3-D Human Action Using HMM and Multi-class AdaBoost