主分量分析和独立分量分析方法的比较研究

Research on comparison of principal component analysis with inde- pendent component analysis

中国科学院地质与地球物理研究所, 北京 100029

Institute of Geology & Geophysics, Chinese Academy of Sciences, Beijing 100029, China

统计特征方法在信号处理、模式识别等领域的应用越来越广泛,特别是独立分量分析(ICA)在理论研究和实际应用中备受关注。主分量分析(PCA)和独立分量分析总的方法和思路比较相似。在分析PCA和ICA的原理及特点的基础上,分别用PCA和ICA对模型和实际地震数据进行了验证,并比较和分析了两者在信号处理中的特征提取能力。结果表明:在实现信号分离时,PCA分解后的各个分量仍然保持一定的相关性,而ICA分解后的各个分量保持独立。

In recent years, statistical property methods have been applied extensively in signal processing and pattern recognition, with the independent component analysis (ICA) in particular because it has been improved in both theoretical research and actual application. The application of principal component analysis (PCA) and ICA are similar in signal processing. The principles and characteristics of PCA and ICA, PCA are analyzed and verified by theoretical model and real seismic data respectively. In addition, the extraction abilities of both methods in signal processing are compared and analyzed ; the result indicates that the PCA components are still related to each other while the ICA components keep independent.