Open AccessClassification of Handwritten Tamil Characters using Variable Length Puzzle Pieces
Author(s) -
R. Deepa,
Rajeswara Rao R
Publication year - 2019
Publication title -
international journal of innovative technology and exploring engineering
Language(s) - English
Resource type - Journals
ISSN - 2278-3075
DOI - 10.35940/ijitee.l3685.1081219
Subject(s) - tamil , scripting language , character (mathematics) , computer science , feature (linguistics) , artificial intelligence , variable (mathematics) , pattern recognition (psychology) , optical character recognition , font , character encoding , natural language processing , speech recognition , image (mathematics) , mathematics , linguistics , mathematical analysis , philosophy , geometry , operating system
offline handwritten character recognition system has been a challenge for Indian scripts, especially for South Indian languages. Huge number of characters of local languages including alphabets, consonants and composite characters make the recognition system more complicated. A good recognition system for subset of Tamil script, a famous South Indian script, is proposed in this work. Variable length feature vector is extracted from the thinned character image. This extracted feature is given to a novel simple classification algorithm which works based on probability. A subset of Tamil script, 20 character classes, is considered for experiment. The samples were taken from HP Labs dataset for Tamil language and a recognition accuracy of 88.15% has been produced.