SMPCA: A new Hybrid Approach for Face Recognition

Face recognition attracts much attention during recent years due to its many applications in different fields such as security systems, entertainment, criminal identification andetc. Face recognition is a complicated area since the human faces change depending on their age, expressions and etc.In this paper we propose a new hybrid approach for face recognition which is used MPCA (Modular Principle Component Analysis) and a SOM (Self-Organization Map)artificial neural network. In this algorithm, each image is divided to some sub-images, and then a weight is assigned toeach of them by MPCA method. Afterwards, SOM is employed for classifying all sub-images according to these weights. Theproposed algorithm is evaluated using well-known ORL database and its performance compare with MPCA algorithm. Experimental results show that applying SOM neural networkleads to considerable improvements and also the recognition rate doesn't alter tremendously when the illumination of the images is varied or they are corrupted by additive noise

کلیدواژه ها:

Face recognition ، MPCA ، SOM ، Artificial Neural network