摘要
介绍一种基于核典型相关分析(KCCA)的字符识别方法。首先选取核函数将低维数据映射到高维空间,再利用典型相关分析(CCA)的思想对数据进行降维,最后利用分类器对降维后的数据进行分类识别。通过对MINST手写数字字符库的实验结果表明,利用KCCA对特征数据进行降维后,能够在新的特征空间中寻找到较好的线性模型,即新特征能够被较好地分类识别。
A method of handwritten numeral recognition based on kemel canonical correlation analysis(KCCA) is introduced in this paper. First,kernel function is chosen to map data in lower dimension into higher dimension. Then the dimension of input data is reduced by means of the principle of canonical correlation analysis(CCA). At last, the method classifies reduced data via classifier. This algorithm is tested on the MINST character database. The results demonstrate that the algorithm can find a better linear model in the feature space after using KCCA to reduce feature data,and new feature can be classified and recognized well.
出处
《光电子.激光》
EI
CAS
CSCD
北大核心
2008年第4期558-561,共4页
Journal of Optoelectronics·Laser