An Adaptive Independent Component Analysis Method
International Symposium on Computers and Informatics (ISCI), 2015-1-17 ~ 2015-1-18, pp 1097-1104, 2015
According to the existing problem of the convention methods, an adaptive independent component analysis method is proposed. First, the signals are divided into the heavy tailed and light tailed signals according to the kurtosis. For the heavy tailed signal, the method off-line computes the score function and establishes the lookup table of the standard alpha stable distribution, and then compute the score function of the mixture signals. For the light tailed signal, the score function is estimated by the general Gaussian model. Simulated results show that, the proposed algorithm has a well performance and a lower computational complexity.
Independent Component Analysis; Blind Source Separation; Alpha Stable Distribution; General Gaussian Model