Robust Face Recognition Using Radial Basis Probabilistic Neural Network

In this paper, we present a human face recognition method based on radial basis probabilistic neural network (RBPNN). Before classification, discrete cosine transform (DCT) is used to extract textural features from frontal face images. Dimension reduction is achieved considering high frequency coefficients as redundant features, thereby omitting them. Subtractive clustering-based RBPNN is used for classification. Two databases - ORL database and LU database are used to evaluate the model. Experimental results demonstrate better classification efficiency of the model with respect to other popular artificial neural networks (ANN) classifiers. Recognition rate of 98% is achieved using this model. 3% and 6% change in recognition rate are observed in ORL and LU database respectively for applying different noise level. Training time remains below 2 sec for both databases.