Face detection based on empirical mode decomposition and matching pursuit algorithm

A novel method combining the empirical mode decomposition, matching pursuit algorithm and Bayes classifier for frontal face detection is presented. Firstly, the method get 27 intrinsic mode functions (IMFs) by using empirical mode decomposition for 3 mean faces which are composed of 1000 Asian faces, 800 faces with glasses and 800 American faces, respectively. Secondly, the method uses the matching pursuit algorithm with a dictionary of these 27 IMFs for face feature extraction. Finally, the Bayes classifier is trained to detect frontal face in an image. The experimental results show that the proposed method has a detection rite of 98.45% in a subset of the FERET face database containing 2000 face images, and a detection rate of 92.11% in the BiolD face database containing 1521 face images.