ICA算法的研究可分为基于信息论准则的迭代估计方法和基于统计学的代数方法两大类，从原理上来说，它们都是利用了源信号的独立性和非高斯性。一般情况下，所获得的数据都具有相关性，所以通常都要求对数据进行初步的白化或球化处理，因为白化处理可去除各观测信号之间的相关性，从而简化了后续独立分量的提取过程,然后再用基于负熵最大的FastICA算法，即可对图像及信号进行解混。-ICA algorithm research can be divided into iterative estimation method based on information criterion and based on statistical algebra two categories, in principle, they are the use of the independence of the source signal and the non-Gaussian. In general, the data obtained are relevant, so usually requires preliminary data whitening treatment or ball, because whitening can remove the correlation between each of the observed signal, thereby simplifying the subsequent extraction of independent components process, and then use the negative entropy largest FastICA algorithm based on image and signal to de-mix.