This chapter focuses on recent research in Randomized Decision

Tree (RDT) algorithm; in particular, how to classify human face

and body using pixelwise classification of depth image. Similar to

popular machine learning algorithms, training of RDT is also com-

putation intensive. This chapter shows a more efficient technique

to reduce the training time of the RDT algorithm. Hence, it is

suitable for power-constrained devices. Besides, two applications

are presented in this chapter to show the efficiency of the technique:

(i) a fall detection system that monitor human fall down; (ii) a

human-computer interface system that enable human to use nose

and mouth to control computer mouse. The applications end with

experimental results and performance evaluations.