A new brain-inspired robust face recognition through elimination of variation features

Independent perceptual feature extraction and modeling the interactions among them are important issues in brain-inspired pattern recognition models. In the face recognition task, person code and different variation codes can be considered as these features. Here, besides extracting the elementary features, perceptual feature modules and their relatedness are modeled. This feature extraction method is a very powerful preprocessing tool in dealing with variation and noise. It would be shown that recognition accuracy for noisy and varied data is highly improved if this classification is implemented in the perceptual feature space instead of the elementary feature space.