Abstract:

A novel fast human face-detecting algorithm is proposed to overcome the redundancy in the construction of human face detector by AdaBoost algorithm. In the investigation , AdaBoost algorithm is first used as a feature se-lectorto select the Haar-like wavelet features , namely AdaHaar features. Then , the selected low-dimension Ada-Haarfeatures are further selected to optimally remove the redundancy and to obtain the corresponding weights. Fi-nally, a human face detector is linearly aggregated by the selected features and the weights. Experimental results show that the proposed human face detector can accurately and fast locate face with low memory cost.

Key words: human face detection, human face recognition, AdaBoost algorithm, genetic algorithm, feature extractlOn, pattem recogmtlOn