东北大学学报（自然科学版）

东北大学学报(自然科学版) ›› 2008, Vol. 29 ›› Issue (9): 1322-1325.DOI: -

• 论著 • 上一篇    下一篇

基于主元分析的频谱整体识别方法

李允公;张金萍;吴宁祥;刘杰;   

  1. 东北大学机械工程与自动化学院;
  • 收稿日期:2013-06-22 修回日期:2013-06-22 出版日期:2008-09-15 发布日期:2013-06-22
  • 通讯作者: Li, Y.-G.
  • 作者简介:-
  • 基金资助:
    国家自然科学基金资助项目(50775029)

PCA-based integrative spectrum identification method

Li, Yun-Gong (1); Zhang, Jin-Ping (1); Wu, Ning-Xiang (1); Liu, Jie (1)   

  1. (1) School of Mechanical Engineering and Automation, Northeastern University, Shenyang 110004, China; (2) School of Mechanical Engineering, Shenyang Institute of Chemical Technology, Shenyang 110142, China
  • Received:2013-06-22 Revised:2013-06-22 Online:2008-09-15 Published:2013-06-22
  • Contact: Li, Y.-G.
  • About author:-
  • Supported by:
    -

RichHTML

0

PDF (PC)

618

摘要： 根据频谱的整体数据进行模式识别和分类时必须考虑如何降低识别过程中的计算量问题,提出了一种基于PCA的频谱整体识别方法.该方法将N点频谱视为N维空间中的点,首先利用已知频谱样本建立数据矩阵,继而进行PCA处理并确定满足信息保留率门限值的主元方向个数,实现高维数据的降维,并计算各类频谱在低维空间投影点的中心,得到数据模板.在方法的识别应用中以距离最小为判据准则进行识别归类.数值仿真和语音识别实验结果说明所提方法性能稳定,识别准确率较高,具有一定的实际应用价值.

关键词: 频谱, 主元分析, 频谱识别, 特征向量, 模式分类

Abstract: Reducing the computational complexity is indispensable for the pattern identification in according to the integrative spectrum data. A PCA-based integrative spectrum identification method is therefore proposed. It regards an N-point spectrum as a point in the N-dimension space and forms a data matrix by use of the known spectrum as samples. Then, after PCA, the number of directions of principal components satisfying the threshold values of information remaining to reduce the dimensions of high dimensional data. And the centers of projective points of various spectrum in low dimensional space are computed to obtain the corresponding data templates. In applications, the identification results are classified according to the criterion which implies the shortest distance. The results of numerical simulation and voice recognition reveal that the method proposed has stable performance with high accuracy of identification and may take effect in its applications.

中图分类号: