Research On On-line Handwriting Identification Based On Information Fusion

Like some biological characteristics such as fingerprint and iris,handwriting which is a behavioral characteristics of a human,can be used to identify someone.With the rapid advancement of pattern recognition,online handwriting identification has been more and more enlightened.And because of the wider fields which the technology has applied to,including financial and self-examination,handwriting identification is badly needed.In addition,it has a wide range of application in areas of identifying people.As for handwriting identification algorithm,the traditional one has many methods.But a single algorithm can't guarantee the performance of identification and its stability.So the result is not very satisfying.During the over ten years,new achievements keep emerging constantly in the research concerning information fusion,causing a breakthrough of the technology for online writer identification.It plays an important role especially in the process of multi classifier comprehensive judgment,and increases the stability of the identification system,reduce the ambiguity of information,and improve the accuracy of handwriting identification.This article puts forward the method of on-line handwriting identification based on information fusion,and the concrete contents are as follows.It introduces the architecture of on-line handwriting fusion identification,and analyzes the methods of dynamic features and static features.The former adopts the two stage identification.The first stage uses the majority vote to screen out the handwriting samples roughly.And the second stage makes a detailed classification for these samples.While the later extracts the characteristics of the GLCM,and use the SVM identify the samples in different category.On the basis of the 180 pieces of handwriting written by 30 people,this article uses dynamic features and static features to identify a single character of handwriting,and the former achieved a higher precision.But the result was still not very satisfying.Finally,a synthetically study of the two different identification methods and the fusion of the results of the two multi-character handwriting not only makes use of their respective advantages but also makes up for shortcomings and a higher rate of correct identification is naturally achieved.The Algorithm of the handwriting identification fusion selects the weighted average method and the fuzzy integral method.The fusion of the former consists of the confidence between the accuracy of the identification and the confidence between the training samples and the test samples through the weighted average method.Then a good candidate can be elected according to a new confidence which is the samples under test.Confidence is the similarity between the test samples and the training samples when measuring.The more confidence indicates the greater degree of similarity between the two samples.The fuzzy Integral Method is another algorithm of fusion.First obscure the results of different methods of identification and measure them with the unifying confidence.Then elect the candidate of the test sample by getting the confidence after fusion through the fuzzy operation.Finally,in this article,a multi-character fusion experiment was made by the way of multi-character and multi-fusion about dynamic and static characteristics.The results show that after the fusion,the precision of handwriting identification has improved significantly from 85.67 percents up to 96.67 percents compared with the precision without the fusion.