ON - LINE UNCONSTRAINED HANDWRITINGRECOGNITIONBASED ON PROBABILISTIC TECHNIQUESHomayoon

This paper discusses a probabilistic on-line handwriting recognition scheme, based on Hidden Markov Models (HMM's), and its implementation for recognizing handwritten words captured from a tablet. Statistical methods, such as HMM's have been used successfully for speech recognition. These methods have recently been applied to the problem of handwriting recognition as well. This paper, discusses a general recognition system for large vocabulary, writer-independent, unconstrained handwritten text. A key characteristic of the recognition system described here is its real-time performance on Intel 80486 class PC platforms without the large memory requirements of traditional HMM-based systems. A word-error rate of 18.9% is achieved for a writer-independent 21,000 word vocabulary task in the absence of any language model. This recognizer is modiied to handle digits written in Persian or Arabic, producing error rates less than 7%. The grounds for developing a full-scale Persian recognition system are also set.