Research Of On-line Handwritten Chinese Characters Recognizing Based By Support Vector Machine

Text is one of important tools for human exchanging information and an important medium for information dissemination, exchanged and record. With the rapid development of science and technology and the popularity of computers, how to process and recognize Chinese characters with computer has become a very important field of research. In addition, the rapid popularization of intelligent terminals such as mobile digital devices also accelerate the pace of development of on-line handwritten Chinese character recognition technology, and make it become an important research topic of character recognition.Statistical pattern recognition method has been widely used among on-line handwritten Chinese character recognition technologies,becauce it is not sensitive to noise and distortion and its classification performance is relatively stable. However, traditional statistical pattern recognition method is used under the conditions of enough samples,that means the number of samples tends to infinity. In practical problems, the sample size is generally limited, so that traditional statistical pattern recognition methods are difficult to achieve ideal results. The statistical learning theory is designed specifically for small samples, so support vector machines based on statistical learning theory can well solve statistical pattern recognition problem under the conditions of limited samples. Therefore, on-line handwritten Chinese character recognition based on the support vector machine has important value.In this paper, the input, preprocessing, feature extraction and classification of online handwritten Chinese character recognition system are discussed and reseached in detail. Finally,online handwritten Chinese character recognition system based on support vector machines is implemented. Specific work as follows:(1) In this paper, the development and status of the online Chinese character recognition are outlined and the basic processeson of on-line handwritten Chinese character are analyzed and handwriting input, preprocessing, feature extraction, classification, post processing and result output are introduced respectively.(2) After analyzing the principle and advantages of the support vector machine, it is applied to the online handwritten Chinese character recognition. On the basis of analyzing support vector machine multi-class classification algorithm, multi-class classifier of on-line handwritten Chinese character recognition is improved.(3) The penalty factor C that is a key parameter in the support vector machine algorithm is studied by matlab experiment using the libsvm testing tool and finally determine the optimal value of penalty factor that is used in this paper. After obtaining optimal parameter, the improved Multi-class classification algorithm was tested. Experimental data verify the improved multi-class classification algorithm to be available.(4) On-line handwritten Chinese character recognition system based on support vector machine is implemented by C++ language. The system includes handwriting input module, preprocessing module, feature extraction module, training module, classification decision-making module and output module. The system can output in real-time by writting on the "Writing area" using mouse,and its recognition rate of handwritten Chinese characters of regular script can reach 96.7%.