HMM-Based Recognition Method Research On Online Handwirting Uyghur Word

Uyghur is the character which Uighur use in their daily life. Modern uyghurbase on Arabic letters and are written from right to left write. At present, it is still inits early stage about online Uyghur handwriting recognition. Through analysingexisting material,main steps of the recognition including preprocessing, featureextraction and classification. Preprocessing is primarily responsible fordenoising.Feature extraction is responsible for finding out hidden information fromhandwriting stroke. Classification is the core in handwriting recognition.It isresponsible for classifying handwriting script according to the Uyghur letters’different writing characteristics.This paper mainly introduces an Uyghur handwriting recognition method basedon HMM.In other words,when we recognize Uyghur,we used HMM asclassifier.Firstly,we preprocess online handwriting data and extract features fromthese data. Secondly,we built HMM model for each Uyghur letter and train thesemodels to estimate their parameters.Finally,we use HMM models to recognizeunknown stroke sequence and find out which word they stand for.In the process ofhandwring recognition, it is difficult to deal with delay stroke. Because online dataformat is a timing sequence,it makes letter’s main stroke and delay strokesuncontinuous in sequence. That greatly increases the difficulty of processing delaystokes and recognizing handwriting word.To solve this uncontinuous problem,wepropose a delay stroke processing method which contains seeking and projection ofdelay strokes.Result shows this method is robust and recognition accuracy rate is upto93.71%.