A Novel Sliding Window Approach for Offline Handwritten Character Recognition

The Handwritten character recognition (HCR) is one of the most promising tasks in the field of optical character recognition (OCR). This paper implements a character recognition system from word images in an offline manner. In an offline system, pen stroke information is not available to the recognition system. The approach presented here performs several steps in accordance to get the final recognized characters. A unique feature representation technique is adapted for handwritten segmented characters. These features are generated by vertically scanning the segmented characters using a sliding window. Experiments are performed on a standard benchmark data-set and finally, the class prediction accuracy in terms of similarity measure is computed and compared to evaluate the performance of the system.