Divide and conquer technique in online handwritten Kannada character recognition

• Authors: M. Mahadeva Prasad;M. Sukumar;A. G. Ramakrishnan
• Venue: Proceedings of the International Workshop on Multilingual OCR
• Year: 2009

The peculiar nature in which one or more consonants combine with vowels to produce a compound character in Kannada language results in a huge number of character combinations, running to tens of thousands or more. The aim of the work is therefore, to reduce the number of character combinations by employing a divide and conquer technique. In the first level of the technique, the structural and the dynamic features of online handwritten Kannada characters are exploited to segment the compound Kannada characters into 282 distinct symbols. This reduction in the number of classes overcomes the huge data collection problem and also reduces the computational complexity. In the second level, these 282 symbols are further divided into three distinct sets of stroke groups, thus further reducing the search space for the recognition engine. One or more of these stroke groups can combine to form any of the thousands of Kannada compound characters. Since the focus of this paper is the above strategy, a simple classifier has been used to validate the effectiveness of the proposed scheme in handling the difficult task of recognizing all possible character combinations of Kannada. The features extracted from the segmented stroke groups are mapped to lower dimensional space using PCA. The subspace features of distinct stroke groups are fed to the respective classifiers in an order and the output of these classifiers are combined to output the Unicode of the recognized akshara. The proposed work is an attempt made for the first time in Kannada language which considers all possible combinations of symbols, including Kannada numerals.