Fusion of Complementary Online and Offline Strategies for Recognition of Handwritten Kannada Characters Rakesh Rampalli (Indian Institute of Science, Bangalore, India rrakesh100@gmail.com) Angarai Ganesan Ramakrishnan (Medical Intelligence and Language Engineering Laboratory, Department of Electrical Engineering, Indian Institute of Science Bangalore, India 560012. ramkiag@ee.iisc.ernet.in) Abstract: This work describes an online handwritten character recognition system working in combination with an offline recognition system. The online input data is also converted into an offline image, and parallely recognized by both online and offline strategies. Features are proposed for offline recognition and a disambiguation step is employed in the offline system for the samples for which the confidence level of the classifier is low. The outputs are then combined probabilistically resulting in a classi- fier out-performing both individual systems. Experiments are performed for Kannada, a South Indian Language, over a database of 295 classes. The accuracy of the online recognizer improves by 11% when the combination with offline system is used. Key Words: Online handwriting recognition, Offline handwriting recognition, Classi- fier fusion, Kannada script, Re-sampling, Pen direction angle, Support vector machine, Spline curve, Directional distance distribution, Nearest stroke pixel, Transition count, Projection profiles, Principal component analysis, Mahalanobis distance. Categories: J.6, I.2.1, I.4.9, I.5.4 1 Introduction Online handwriting recognition systems employ an active tablet, on which the user writes the text. The tablet captures the movement of pen tip on its screen in terms of PEN DOWN, PEN UP information and x,y co-ordinates sampled at equal intervals of time. This transducer is connected to a computer online and the data captured by the tablet is sent to the recognition system. Like any other pattern recognition system, the conventional handwriting recognition system fundamentally consists of basic building blocks like segmentation, pre- processing, feature extraction and classification. On-line handwriting recognition methods using stroke-information are generally expected to achieve higher accu- racy than the off-line methods. However, the variations in the number of strokes,