An efficient feature extraction method for segmented cursive characters recognition

Handwritten document analysis is a persistent research area nowadays. It has large applications in various image processing domain as human computer interaction, machine translation and automation of reading various scanned images for handwritten fields in forms, postal address on envelopes and amounts in banks checks. The main factor which influence the performance of handwritten text recognition is the selection of an appropriate set of features for representing input samples. In this paper, we propose a efficient feature vector for handwritten character recognition, using a hybrid of the statistical and structural properties of a character to represent the particular character class. Experiments have been performed on standard database of handwritten digits and letters. The recognition accuracy is tested on a Neural Network classifier with different parameters. The results have been compared with existing features extraction algorithms. The comparative results shows the effectiveness of our approach.