NeurOCR: A Neural Network based Approach to Optical Character Recognition (OCR) Systems

The recognition of optical characters is known to be one of the earliest applications of Artificial Neural Networks, which partially emulate human thinking in the domain of Artificial Intelligence. The current paper focuses on the use of neural network in order to mitigate the problems of digital handwriting recognition by using Self-Organizing Map model for fast processing and less processing power consumption keeping its deployment on PDA in mind. The document is expected to serve as a resource for learners and amateur investigators in pattern recognition, neural networking and related disciplines. Technology Used is C#, .Net 3.5 Framework and Heaton Research Neural Network API has been used exhaustively for deploying the calibrated Self-Organising Model (SOM) approach. Characters are input when a user draws on a high-resolution box. Unfortunately, this resolution is too high to be directly presented to the neural network. To resolve this problem, we use the techniques of cropping and down sampling to transform the image into a second image that has a much lower resolution. It was observed that the use of Neural Networks for Recognising characters proves out to be more efficient and robust compared to other hard computing techniques. The Self-Organising Model network is proved to be the most prominent competitor for such an application providing precise outputs for recognised characters.