论文部分内容阅读
要提高手写字符的识别率,关键在于提高相似字符的识别率。主曲线是一种新的基于非线性变换的特征抽取方法,它是通过数据分布“中间”并满足“自相合”的光滑曲线,较好地反映了数据分布的结构特征。本文尝试用主曲线这种新的方法来提取手写字符的结构特征,并基于这些特征来对相似字符进行模糊分类。所提方法在CE-DAR和OCRD手写体字符数据库上的实验结果表明:该方法不但是可行的,而且能有效提高相似字符的识别率,它为字符识别的研究提供了一条新途径。
To improve the recognition rate of handwritten characters, the key is to improve the recognition rate of similar characters. The main curve is a new feature extraction method based on nonlinear transformation. It reflects the structural characteristics of data distribution through the data distribution “middle ” and satisfies the smooth curve of “self-coincidence ”. This paper attempts to use the new method of the main curve to extract the structural characteristics of handwritten characters, and based on these characteristics to similar characters fuzzy classification. The experimental results on CE-DAR and OCRD handwritten character database show that this method is not only feasible, but also can effectively improve the recognition rate of similar characters, which provides a new way for the study of character recognition.