Malayalam online handwriting recognition system: A simplified fuzzy ARTMAP approach

On-line handwriting recognition is the automatic conversion of handwritten text into letter codes as it is written on a special digitizer or PDA which is usable within computer and text-processing applications. Some of the commonly used techniques for online handwritten characters include Hidden Markov Model (HMM), elastic matching, structural analysis, Time delay neural networks, and combination of multiple classifiers. This paper describes Malayalam online handwriting recognition system using SFAM artificial neural network technique. The structural and directional information is extracted/collected from each character/stroke. The extracted features, which form the feature vector, are passed as input to a Simplified Fuzzy ARTMAP (SFAM) artificial neural network (ANN) classifier. The SFAM classifier compares the input data with the trained data and finds the nearest prototype from the database that `resonates´ with the input pattern. The labels or recognized characters are assigned their corresponding Unicode code points and displayed using appropriate fonts. Using this technique we have obtained an average accuracy of 98.26%.