The paper proposes a framework for recognizing handwritten Gurmukhi alphabets. Handwritten character recognition has been a fundamental problem in computer vision which is now dynamic and demanding research sphere in the domain of pattern recognition and image processing. It has manifold applications specially transformation of scene text image into structural text form. Character recognition is a technique of detection, segmentation and identification of characters from an image. A consequent goal of handwritten character recognition is to trigger the human reading potential and to enhance the reading, understanding and editing intelligence of computer like human do with text. However, very little work has been carried out on recognizing characters for the Gurmukhi script. The paper relies on the state of the art deep convolutional neural networks for extracting representations of characters. A new data-set comprising of 5960 images of 40 handwritten Gurmukhi alphabets was proposed for carrying out the various experiments. An overall accuracy of 91.76 % was achieved by the model that out-performed other state of the art classifiers.
Volume 12 | 03-Special Issue
Pages: 1-9
DOI: 10.5373/JARDCS/V12SP3/20201231