Abstract
In this paper, we proposes an face recognition method with an interactive recognition modules model. We model the recognition to two layers, lower layer is a Gabor wavelet transform system, higher is multi-agent systems. In the lower layer, 5 output images is extracted for whole face and 4 parts of face, left and right eyes, nose and mouth, from an input image by Gabor wavelet transform, An output image is for the distribution of frequency features for each parts or whole face. In the hight layer, the 5 outputs for whole face and 4 parts are 5 agents inputs, relatively, each agent recognizes the coresponded part with adding another agent recognition output. We confirmed by the experiment, that the proposed method is effective.
