ABSTRACT

This paper presents a new offline handwritten Devanagari word recognition system. Though Devanagari is the script for Hindi, which is the official language of India, its character and word recognition pose great challenges due to large variety of symbols and their proximity in appearance. In order to extract features which can distinguish similar appearing words, we employ Curvelet Transform. The resultant large dimensional feature space is handled by careful application of Principal Component Analysis (PCA). The Support Vector Machine (SVM) and k-NN classifiers were used with one-against-rest class model. Results of Curvelet feature extractor and classifiers have shown that Curvelet with k-NN gave overall better results than the SVM classifier and shown highest results (93.21%) accuracy on a Devanagari handwritten words set.