Gabor Representation Based Probabilistic Subspace Analysis for Face Recognition
一种基于Gabor描述的概率子空间人脸识别方法

In this paper, a GPSA face recognition approach by probabilistic subspace analysis in Gabor wavelet domain is presented. First, an improved Gabor representation scheme for face images is given, then a Gabor based probabilistic subspace model is built, so recognition is performed in a manner of probabilistic matching. The discriminatory information yielded from both probabUistic subspace analysis and Gabor representation is exploited altogether in GPSA method, and hence the robustness of face recognition system is enhanced effectively. The experimental results on a mixture face database including 190 individuals show that the proposed GPSA method outperforms the existing PSA method.