Super-exponential equaliser-a modified eigenvector algorithm (mEVA)

Blind equalisation may be seen as a particular case of independent component analysis and methods based on a entropy maximisation can be applied. Although implementation of these methods have not been computationally efficient yet, some of the recently proposed methods reach the level of practical applicability. Despite the common fundamental principle, i.e., the iterative elimination of second and higher order correlations, the algorithms perform differently. As a matter of fact, the algorithms possess many desirable properties compared to conventional methods, among these are their super-exponential convergence and their modesty regarding the number of required samples. In this paper another algorithm is proposed, which requires less computations than competitive algorithms by exhibiting a similar or better overall performance.