earticle

영어

The feature extraction and classification method(s) used to recognize handwritten characters play an important role in Handwritten Character Recognition applications. A suitable feature extractor and a good classifier play a very important role in achieving high recognition rate for a recognition system. If we want to develop a new feature extractor for a script, it will help us if we have the knowledge of the recognition ability of the existing feature extractor. Kannada is a major south Indian script spoken by about 50 million people. This paper examines a variety of feature extraction approaches and classification methods which have been used in various Optical Character Recognition applications which are designed to recognize handwritten numerals of Kannada script. The study has been conducted using 8 different features computed from zonal extraction, image fusion, radon transform, fan beam projections, directional chain code, discrete fourier transform, run length count and curvelet transform along with ten different classifiers like Euclidean distance, Chebyshev distance, Manhattan (city block) distance, Cosine distance, K-NN, K-means, K-medoids, Linear classifier, Artificial Immune system and Classifier fusion are considered.