Face Recognition Using Eigenfaces

In this paper, we propose a PCA-based face recognition system implemented using the concept of neural networks. This system has three stages, viz. pre processing, PCA and face recognition. The first stage, preprocessing performs head orientation and normalization. The aspects that matter for the identification process are ploughed out using Principal Component Analysis (PCA). Using the initial set of facial images, we calculate the corresponding eigenfaces. Every new face is presented into the face space and is characterized by weighted-sum of corresponding eigenfaces that is used to recognize a face. To implement this face recognition system, we have created a database of faces with the help of neural networks and we have built one separate network per person. We obtain a descriptor by projecting a face as input on the eigenface space, then that descriptor is fed as input to the pre-trained network of each object. We select and report that which has the max output provided it passes the threshold already defined for the recognition system. Testing of the algorithm is done on ORL Database.