Offline Handwritten Sanskrit Simple and Compound Character Recognition Using Neural Network

The intricacy of Sanskrit simple and compound characters makes recognition a laborious task for the probe. For forming a new compound character, we join two or more characters in different ways. In early days of computer wisdom, Optical Character Recognition is one of the working contents of research. In the field of Optical Character Recognition, pattern recognition is an especial feature of Sanskrit script. The prevalence of compound characters in Sanskrit language is more assimilate to other languages scripts. This paper reports a Levenbug-Marquardit algorithm also known as damped least squares method for recognition of handwritten Sanskrit simple and compound characters with different feature sets. Data was collected from people of different age groups categorized as child, young and old. At the first step, we perform pre-processing, second the main notion of the propound technique is to produce the character image using classifier, which will be directly used to recognize the Sanskrit simple and compound characters. Average recognition rate of 78.4, 80.5, and 77.4 % is achieved from our simulated work on noise free images.