Machine Learning for Optical Character Recognition System

Summary

Optical character recognition involves identification, classification and in some applications, correction of optical symbols/patterns present in a digital image. Recognition can be focused on online printed text, offline and also on handwritten documents. Many applications such as in postal addresses, bank checks and vehicle number plate verification etc. require OCR systems to make the processing fast. Segmentation, Feature extraction, classification techniques play a vital role to perform character recognition. There are different phases in an OCR to efficiently process the text such as optical scanning, location segmentation, pre‐processing, segmentation, representation, feature extraction, training and recognition and post‐processing. In training phase ANN can be used to make system efficient to process huge data. Recognition of handwritten text is an active area of research. Various techniques involved in OCR and their limitations are discussed along with an overview of precision rate of ANN based approaches.