希尔伯特—黄变换手写体汉字识别方法  

New HHT-based handwritten Chinese character recognition method

作　　者：宋万干[1] 杨扬[2] 栗欣[2] 

机构地区：[1]淮北煤炭师范学院计算机科学与技术学院,安徽淮北235000 [2]北京科技大学信息工程学院,北京100083

出　　处：《计算机应用》2009年第12期3363-3365,共3页journal of Computer Applications

基　　金：国家自然科学基金资助项目(60673160)

摘　　要：针对脱机手写体汉字的非平稳特性,提出了一种基于希尔伯特—黄变换的手写体汉字识别方法。该方法根据汉字图像的特点将一维HHT扩展为二维HHT,用以获得汉字的瞬时频率特征,并将局部的瞬时频率特征和全局的Zernike矩特征组合作为最终的识别特征。实验结果表明,该方法是有效的。Concerning the non-stationary characteristic of off-line handwritten Chinese characters, a new recognition method by Hilbert-Huang Transform （HHT） was presented. According to the characteristic of images of Chinese characters in this method, two-dimensional HHT for handwritten Chinese characters was developed based on the one-dimensional one to get the frequency feature of Chinese characters. Then the partial frequency feature and the global Zernike moments feature of Chinese characters were combined as a final recognition characteristic. Experimental result shows that the .method is effective.

关 键 词：脱机手写体汉字识别 希尔伯特-黄变换 特征提取 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]