摘要
传统典型相关分析准则函数在界定同组特征矢量中各元素之间不相关时采用乘法运算,易出现无法保证两组特征矢量同时达到同组元素协方差意义下最小的问题。为解决这一问题,提出了一种改进的典型相关分析方法,对典型相关准则函数分母部分进行修正,将其中的乘法改成加法,对该方法进行推导得到投影矫正系数,该系数对两组特征矢量特征方程进行调整,保证两组特征矢量同时达到同组元素协方差意义下最小。该方法有效地改善了融合效果、提高了人脸识别的正确率。在ORL和Yale人脸数据库上的实验结果都验证了其有效性。
The traditional canonical correlation analysis criterion in defining function of the same group on the correlation between the elements use the multiplication operation, and it is an easy problem that it can't guarantee the two groups can be in smallest eovariance of the same group of elements at the same time. To solve this problem, an improved canonical correlation analysis was proposed which made an amendment in the denominator of the criteria, it would change the multiplication to the addition, it had a projection correction coefficient by derivation, the projection correction coefficient adjusted the characteristic equations of feature sets and guaranteed the two groups could be in smallest covariance of the same group of elements at the same time. The new method effectively improves the integration effect and the face recognition rate. In ORL and Yale face database, the experiments verify its effectiveness.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2010年第2期388-390,共3页
Journal of System Simulation
基金
国家自然科学基金(60875004)
江苏省自然科学基金(BK2009184)
关键词
特征融合
典型相关分析
准则函数改进
人脸识别
feature fusion
canonical correlation analysis
improved discrminant criterion
face recognition