To date, many researchers have proposed various methods to recognize handwritten characters, which in one way or the other use image segmentation techniques. In this paper, image segmentation has been avoided successfully to encode the characters in developing an automated unconstrained handwritten character recognition system. In the offline mode, several training images corresponding to each known handwritten character are fused to generate patterns for each of them. The patterns are stored in 10×10 matrices, irrespective of the image sizes, thus saving storage space significantly. In real time, characters are extracted by sampling the handwritten scripts at different intervals determined heuristically. Extracted characters are matched with the stored patterns for recognition. The characters which do not match are classified as disjoint components and processed further for recognition. In order to develop a robust system, the characters are represented as fuzzy sets, and a fuzzy confidence value for each of them is computed for recognition. This method is applicable for verification of the recognized characters too. Based on the system feedback, more training images can be fused at any point of processing, which improves recognition rate. The system is tested with a large number of handwritten scripts, independent of the writing style to demonstrate its effectiveness.

Many jobs, like the postal system, require involvement of human beings to read handwritten scripts and take decisions accordingly. Recognition of handwritten characters also has wide applications in card-oriented activities. This technology, when used along with biometrics, could be utilized in a number of future security applications. Automation aims at reducing manual intervention, thus saving a considerable amount of time, money and energy.

Development of an Intelligent System for Extraction and Recognition of Handwritten Characters, recognition, characters, system, scripts, patterns, character, segmentation, components, considerable, demonstrate, disjoint, extracted, generate, heuristically, independent, involvement, intervention, methods, automated, processing, Recognition, recognized, researchers, biometrics