Blind channel identification using lower-order statistics

Non-Gaussian statistical signal processing is important when the signals deviate from the ideal Gaussian model. One of the most important non-Gaussian distributions is the family of the alpha-stable (0 less than (alpha) less than 2) distributions. They are proven to be effective in modeling the impulsive signal environments both in theory and in practice. This paper presents a brief introduction to the alpha-stable distributions, their applications to signal processing using fractional lower-order statistics, and the alpha-spectrum, a new spectral analysis tool for blind channel identification in impulsive environments.