| Palm-held computing is on the rise. A keyboard is too big for a palm-held computer so a stylus and tablet system for interaction requires a much smaller interface. Hence, there is need for handwriting recognition. Very little research has gone into handwriting recognition in Arabic/Persian (A/P) due to the difficulty of the task. As the Arab world becomes increasingly computerized and mobile, and technology becomes increasingly ubiquitous, the need for a natural interface becomes apparent.This research will show average character recognition rates above 95% for A/P characters. Uses of Discrete Cosine Transform (DCT) for feature extraction with Neural Networks and Hidden Markov Models (HMM) as classifiers have shown great performance with multiple unconstrained writers. Multi layer perceptron (MLP) networks are emphasized, since this provides a higher performance during both training and testing. Since the overall method represents low computational overhead, it would be suitable for hand held devices - the target application device. |
