Adaboost Human Detection Based on a DSP Platform

Human detection is a task met in many applications such as surveillance & security, safe driving system, intelligent vehicles, etc.. It is more complicated compared to the detection of car and other targets, due to the variety of poses and external appearance of human bodies. This paper presents an adaboost human detection algorithm and its implementation on a DSP platform. The detector uses haar-like features as classifiers. A cascade of boosted classifiers is obtained after extensive training of hundreds of positive and negative samples. Experimental results with the adaboost human detection algorithm are presented in the paper.