In recent years, the use of Convolution Neural Network (CNN) with a huge amount of images in databases, has made
the deep learning technique very beneficial. Our objective is to improve the Face Recognition system using Deep Neural Network
because of the importance of this system in many applications such as security systems, mobile authentication, access control and
banking using ATM. We will use a convolution neural network to make the face recognition performance being analogous to
humans. CNN technique learns features discriminatively not handcrafted to improve recognition accuracy. The learned face
representations are very valuable for face recognition and are also capable of reconstructing face images in their frontal views. We
propose a deep neural network model which is 15-layer to learn discriminative representation, obtain and outperform the state-ofthe
art methods on ORL (Olivetti Research Laboratory face database) and YTF (YouTube Faces database). The comparison will
be done to CNN with Fuzzy Hidden Markov Models (FHMM) and Principle Component Analysis (PCA). For our presented CNN
method, we have obtained the best recognition accuracy of 99.69 %. The presented system based on deep neural network
transcends the state of the art methods in the field of face recognition.