Principal Component Analysis and Generalized Regression Neural Networks for Efficient Character Recognition

Low dimensional feature representation with enhanced discriminatory power is of paramount importance to any recognition systems. Principal component analysis (PCA) and Neural Network are commonly used techniques of image processing and for recognition purpose. In this paper, a new scheme of combining PCA and Neural Network is used for character recognition. PCA is a dimensionality reduction technique based on extracting the desired number of principal components of multidimensional data. Generalized regression neural network (GRNN), where it has redial basis layer and a special linear layer is used for subsequent classification purpose. Experiments on the character database (printed and handwritten) demonstrate the effectiveness and feasibility of the proposed method.