Fusion of Fuzzy FFL-KLT and PCNN Features on the Face Recognition Problem

The problems related to face recognition has been investigated from different points of view. Artificial Neural Networks, ANN, is not the exception. This paper presents a novel approach for face recognition based on three features, first the Pulse Coupled Neural Network Features; second, the Face Feature Lines, FFL, and third, the Karhunen Loève transformation, KLT. The facial features are inputs used in a neurofuzzy system that involves their fuzzyfication. These fuzzy inputs are fed to a RBF neural network with a variable architecture on the first layer. The membership functions used to fuzzify the facial features are created according to the features automatically. The system performs well with ORL and YALE face databases, reaching recognition rates comparable with the current face recognition systems.