Feature Description by Improved Local Gabor Filters and ICA

A new face recognition algorithm using the RBF network is proposed based on the improved local Muti-channel Gabor Filters and fixed point ICA. The normalized face image is firstly muti-degree sampled and blocked, and then the blocked face image was filtered by multi-orientation Gabor filters with multi-scale to extract their corresponding Local Gabor Magnitude Map (LGMM), which were constructed to higher dimensional feature vectors. Next, the dimensionality of these vectors is reduced by means of principal component analysis. Finally, the independent components in the resulting vectors with dimensionality reduced were analyzed and extracted by using ICA recognition classification. Experimental results on ORL and YALE face show that the proposed algorithm, which achieves more recognition accuracy rate than other methods.