An Investigation on the Performance of Hybrid Features for Feed Forward Neural Network Based English Handwritten Character Recognition System

Optical Characters Recognition (OCR) is one of the active subjects of research in the field of pattern recognition. The two main stages in the OCR system are feature extraction and classification. In this paper, a new hybrid feature extraction technique and a neural network classifier are proposed for off-line handwritten English character recognition system. The hybrid features are obtained by combining the features extracted using diagonal, directional, Principal Component Analysis (PCA) techniques along with statistical and geometry feature extraction technique. The hybrid features are used to train a feed forward back propagation neural network employed for performing classification tasks. The hybrid features derived from two hundred character sets of lowercase English alphabets (a to z) were used for training the network. The overall recognition system has been tested extensively and shown to perform better than individual feature extraction techniques. The hybrid technique suitably combines the salient features of the handwritten characters to enhance the recognition accuracy.