A system for identifying writer from Thai handwriting image

Many techniques have been reported on handwriting-based writer identification. Most of such techniques assume that the written text is fixed (e.g. in signature verification). A neural network based technique in which the written text is not fixed to identify a writer from Thai handwritten image, is presented. Without fixed text, it is much more difficult for a fake writer to sneak through the system test. Each writer has to write training set of 63 Thai characters and 25 testing sentences. The results of 99.43% accuracy on the classification of 250 test documents from 10 writers are very promising.