Journal of Pattern Recognition Research

Compound character is a special feature of Marathi script, derived from Devanagari. It joins two or more characters in various ways forming a new character. The complexity of compound characters makes its recognition a challenging task for the researchers. The frequency of occurrence of compound characters in Marathi language is more compared to other languages derived from Devanagari. This paper presents a novel approach for recognition of handwritten Marathi compound characters using a multi-stage multi-feature classifier. In the first stage, the compound characters are classified by two stage structural classification based on various structural parameters. In the second stage, various features like pixel density features, Euclidean distance features and modified wavelet approximation features are extracted from the structurally classified and normalized characters. The three features are applied to three different neural networks. The final recognition output is selected based upon majority voting and if all the results differ then final output is confirmed by the neural network with modified wavelet approximation features. The recognition accuracy given by the proposed system is 97.95%.