Online New Tai Lue Handwritten Character Recognition Based On The SVM

Character recognition is an important branch on pattern recognition domain. A typical character recognition system consist of four parts,the character acquisition,character track preprocessing,character feature extraction,classification and identification.According to the different methods of identification,the character recognition can be divided into online character recognition and off-line character recognition.And off-line character recognition can be divided into off-line printed character recognition and off-line handwritten character recognition.At present, some character recognition technology are very mature,such as Chinese,English etc.And there are a lot of corresponding commercial products in our life,but the research of New Tai Lue character recognition technology started later and got less achievement. At the same time,mobile device.such as mobile phone,for the Dai people become more and more in demand,and the handwriting input is an important interactive way,so it is necessary to do more work on studying on-line handwriting Tai Lue character recognition technology. In this context,this article studies the online handwritten character recognition of New Tai Lue.This paper introduces the development of on-line handwriting recognition technology and the main character recognition technology of Chinese and English firstly.Then,in this foundation,this paper studies the online handwritten New Tai Lue character recognition,the main work accomplished includes the following:1. This article introduces the background and study status of New Tai Lue character recognition,and analyzes the difficulty.Then proposes the New Tai Lue characteristic set for online New Tai Lue character recognition, contains direction code,the number of character stroke,the direction of start stroke and end stroke, etc.2. To collect the New Tai Lue character data by the mobile phone software. To preprocess the New Tai Lue character data,including data normalization,resampling,etc. According to the feature extraction method, extracting the characters of Tai Lue characters and forming the character feature vectors.3. An application of SVM algorithm in online New Tai Lue character recognition and the result would be the first candidate.On this basis,the Euclidean distance is used to choose the second to the tenth candidate, and getting the correct recognition rate of top1-rate of top 10.4. The detailed result and analysis of experiments are given. Finally, this paper summarizes the work of whole thesis, and put forward the study direction in future.8463 samples were used for experiment,7012 samples of them were training samples,1451 samples were testing samples.The experiment results showed that the rate of top 1 is 88.35%,rate of top 5 is 96.76%,rate of top 10 is 97.93%.