A Robust Handwritten Digit Recognition System Based on Sliding window with Edit distance

Handwritten Number or digit recognition plays an important role in different platforms and applications. Here, a novel technique based on a string edit distance algorithm is proposed to recognize offline handwritten digit images. The recognition system tries to build up each digit’s string and then predicts the digit class by comparing the test string against the existing string from the training set. The lower the edit distance among the two digit’s string, the greater the chances of similarity among them. Some digit databases have evaluated this system and a variety of mixed datasets have been developed to validate the system’s robustness. So that a dataset can train the system and another dataset can test it. Four types of datasets are used as evidence for the excellence of the robustness of the projected system. In some cases, the recognition system is trained with very less training samples from a specific dataset, but a large number of samples from a different dataset are tested. The outcomes of these experiments are shown by tabular as well as a pictographic version.