A hidden markov model extension of a neural predictive system for on-line character recognition

We present a neural predictive system for on-line writerindependent character recognition. The data collection of each letter contains the pen trajectory information recorded by a digitizing tablet. Each letter is modeled by a fixed number of predictive Neural Networks (NN), so that difSerent multilayer NN model successive parts of a letter. The topology of each letter-model only permits transitions from ench NN to itself or to its right neighbors. In order to deal with the great variability proper to cursive handwriting in the omni-scriptor framework, we implement during both Learning and Recognition a holistic approach by performing adaptive segmentation. Also, the Recognition step implements interactive Recognition and Segmentation. Our approach compares Neural techniques combined with Dynamic Programming to its extension to the Hidden Markov Models (HMM) framework. Our first system gives quite good recognition rates on letter databases obtained from 10 diferent writers, and results improve considerably when we consider the extension of the first system to the durational HMM framework.