Automatic face recognition for access control

Pattern recognition is one of the fields where neural networks (NN) have been most applied. These problems have, however, been dealt with using ad hoc architectures to date. In the past, the architectures used for such problems were too large and the number of connections was such that they could not be supported by low-level hardware. Here, we propose an intermediate solution, where we attempt to optimize the performance of a simple NN and show that, using suitable learning techniques, simple architectures can be made to deal with complex problems. The task presented is the problem of authenticating individuals from an image of their face for security purposes in access control to top security environments.