基于主成分分析与核独立成分分析的降维方法

Dimensionality reduction method based on PCA and KICA

根据主成分分析(principal component analysis,PCA)法的降维去噪技术和核独立成分分析(kernelindependent component analysis,KICA)法的盲源分离技术,提出了一种关于两者的融合方法,即PCA-KICA方法。将该方法应用于线性和非线性高维混合信号的降维处理中,以相关系数和Amari误差为标准,同主成分分析与独立成分分析(principal component analysis-independent component analysis,PCA-ICA)融合方法进行比较。仿真结果标明,PCA-KICA方法与PCA-ICA方法相比,在处理复杂非线性高维混合信号时效果相当,但在处理线性高维混合信号时的效果较好。

According to the dimensionality reduction technology of principal component analysis （PCA） method and the blind source separation technology of kernel independent component analysis （KICA） method, a combined method, the PCA-KICA method, is presented. It is applied to dealing with some linear and nonlinear multidimensional mixing signal processing. Meanwhile, it is compared with the PCA-independent component analysis （PCA-ICA） method by correlation coefficient and Amari error. Simulation results indicate that, compared with the PCA-ICA method, the proposed method achieves a proximate effect when dealing with complicated non-linear muhidimensional mixing signals, but can achieve a better result when dealing with linear multidimensional mixing signals.