汉字识别研究中若干核心问题的新开拓

The global information times present new requirmements and challenges to character recognition research ceaselessly. This project made innovations in character recognition research field, achieved important research results on the theories and methods of low resolution gray image character recognition, HMM(Hidden Markov Model)-based character recognition, multilingual printed character recognition, document digitalization with full information, form recognition with low quality images, and made meaningful explorations in video character detection and recognition, writer identification etc.. The research results are industrialized into a series of software products successfully. Asian character recognition system succeeded in international competition; the software product is widely used to produce large amount of high-fidelity electronic publications, and become the most important tool in digital library' content making; form recognition application system is used as the kernel of the "Value-added Tax Invoice Recognition and Verification System" in the national "Golden Tax Engineering" . It achieved one scientific appraisal, the experts' comment is "the integral performance is at the world's leading level", and got one Beijing Advanced Science and Technology Award (2nd class).

高识别率和鲁棒性仍是汉字识别的重要研究课题，直接从灰阶图象进行文本版面分析、行字切分和识别，可以减少二值化造成的信息损失，提高识别鲁棒性。隐含马尔可夫模型对联机手写识别、文本整体系统识别和脱机手写识别有重要意义，将结构方法和统计方法有机结合是解决汉字识别的希望之途，联机手写汉字的统计空间和时间统一模型理论实践有重要意义。...