Nonlinear functions for blind separation and equalization
H Mathis · Repository for Publications and Research Data (ETH Zurich) · 2001
Nonlinear functions are an important part of blind adaptive algorithms solving filtering problems such as blind separation and blind equalization.Roughly speaking, they take over the role of a proper training reference signal, which is not available, hence the term "blind".On the other hand, many simple algorithms for blind deconvolution, such as Sato's algorithm and the constant-modulus algorithm (CMA) can be extended to work for a wider class of distributions by adding a simple coefficient norm factor in the update equation.