Two stage Classification for Online Telugu Handwritten Characters
Abstract
Real time applications of human being is closely in association with the use of advanced and highly reliable electronic gadgets. Online handwritten character recognition system attracts researchers due to its usage in general applications. Many handwritten character recognition applications use English as the input language. But, recognition of native language like Telugu, which is being spoken by lakhs of people in South India, is important as India is a multilingual country. This work describes an online handwritten character recognition system working in combination with an offline recognition system. Online handwritten numeral recognition system is developed using Local features like(x,y)co-ordinates,(Δx,Δy) and (Δ2x,Δ2y) and the global features like tan() are considered and ANN modellingused for Classification of 52 Telugu vowels and consonants. Offline handwritten numeral recognition system is developed using zonal discrete cosine transform (DCT) as feature and Neural Network technique as the modelling technique. Our method produced a good recognition rate of 97% accuracy for 56 classes including vowels and consonants of online Telugu handwritten character dataset.
Keywords: handwritten character; recognition; feature extraction; classification