一种基于置信度的动态特征提取算法

Dynamic Algorithm of Feature Extraction Based on Believing-Degree

结合实际工程中的某些要求 ,将统计学中的置信度概念引入特征提取中 ,提出了一种按照精度要求进行有效的特征提取的算法 ,该算法能够实现在抑制漏警错误或者虚警错误的同时降低特征维数的功能。最后介绍了此算法成功地应用于现钞识别系统的系统设计细节以及实验结果 。

Combined with the requirement of real project,the statistic concept of believing degree is introduced into feature extraction and a dynamic algorithm for effective feature extraction is proposed,which is based on accuracy requirement and capable of constringing the feature dimension while reducing leaked or pretend alarm probability.The design detail of a note identification system,which adopts the algorithm,and its test results are introduced.The real applications have verified its effectiveness.