Authors : Mrs.Snehal S. Golait , Dr.L.G. Malik, Prof.A.Thomas
Abstract–Feature Extraction plays a vital role in any character recognition system. It involves measuring those features of the input pattern are relevant to classification. The characteristics of the feature extraction techniques have to be independent of the scalable font characteristics such as type, size, style, tilt, rotation and should be able to describe the complex, distorted, broken characters effectively. A feature vector should be simple, reliable, complete, and compact to recognize any input character with high accuracy similar to human perception. This paper provides a feature extraction for Handwritten Marathi Compound Character using structural and statistical methods. Aside from the similarity of Character, Complex shape researcher founds difficulty to find features of characters. The recognition is carried out using structural and statistical feature extraction and multistage classification scheme. The Proposed system used Edge map as a structural feature extraction method. DFT and DWT used as statistical features extraction method. We apply all techniques on segmented characters able to get more than 90% accuracy in recognition process.
