A novel face recognition system inspired by computational neuroscience

The human brain operates superior than machines in most of the situations. Computational neuroscientists try to translate brain functions into the language of mathematics then modeling and implementing them into a machine. We can introduce novel methods and technologies inspired by the functions of the human brain. One of these technologies is behind the incredible power of human face recognition. Here we introduce a novel method for face recognition. In this method, we modeled the simple cells in the visual cortex (which are responsible for orientation selectivity) with the Directional Filter Banks. Then in order to normalize the illumination, we used the Single Scale Retinex. After that, for the other regions of the visual cortex (V2, V4, PIT...), we used the well known HMAX model of object recognition. After these feature extraction levels we need a classifier. The PFC area in the cortex performs like a classifier. In this stage we introduced a sparse representation based classifier by solving the L-1 regularized least square problem. Experimental results on facial image databases with varying illumination and pose shows that the performance of our model is comparable with well known methods in automatic face recognition.