Research On Recognition Of On-Line Handwriting Documents

In the process of text input with traditional handwriting input method, users are in need of checking whether the recognition result is correct and selecting the true recognition results from the list of identified candidates. Such operations will interrupt the users'ideas. Writing process is difficult to write with ease and grace. The efficiency of the handwriting input will be reduced greatly. As recognition based on local information can't take full advantage of context information, the recognition rate is difficult to improve further.With the problems for the traditional handwriting input method, this article proposes a method that supports users the freedom to edit the written handwriting document. Full use of the results of handwriting document similarity analysis and context, this approach which compromises the process of text editing and handwriting recognition composes natural and efficient method of human-computer interaction.Based on summarizing research status on handwriting character recognition algorithm and handwriting similarity analysis algorithm, the main contributions of this thesis are as follows:(1)Propose an improved dynamic time warping algorithm, and implement the handwriting text document similarity analysis using by handwriting character's similarity calculation.(2)Propose handwriting document recognition algorithm and adaptive method of recognition candidate based on the results of handwriting document similarity analysis and context.(3)Design and implement the main functions of various parts of handwriting document online recognition system, and evaluate the algorithm's validity and effectiveness.Experiments show that recognition accuracy using by this article recognition method promotes nearly 4.7% on the first candidate and nearly 2.5% on the first five candidates, compared with handwriting character recognition algorithm. The function of intelligent modify of this article can help users to modify the 18% of the recognition errors automatically. This article implement handwriting document online recognition system which supports users the freedom to edit the written handwriting document. This system meets users'requirements of continuous input and quickly recognition.