Thai Online Handwriting System Using Neural Networks and Fuzzy Logic

Abstract

In this paper, an on-line handwriting recognition system is proposed. Such a system combines the advantages of neural network and fuzzy logic algorithms. In this approach, Thai input-character is normalized in the preprocessing process. Then, the generic and specific features are extracted. Only a few generic features are fed to the neural network model so that the computational time of the training phase is minimized. After that, if the neural network yields a lower probability, then the fuzzy logic model is needed. In this case, the set of simple fuzzy rules are set up by means of the specific features. In this way, the experimental results illustrate that the proposed system can accurately recognize up to 86.28%.