Blind Separation and Equalization Using Novel Hill-Climbing Optimization

In this paper, we construct a maximum-likelihood-equivalent metric or auxiliary function, which can result in a novel expectation-maximization Hill-Climbing (EM-HC) optimization procedure; it can be easily implemented for the estimation, detection and clustering applications since it is based on the simple auxiliary function. In this paper, one major application of our new EM-HC method, namely the blind separation and blind channel equalization, is presented and an efficient Iterative weighted least-mean squared (IWLMS) algorithm is derived thereupon. The new IWLMS algorithm derived from the EM-HC techniques greatly outperforms the prevalent blind equalization algorithm based on the constant-modulus criteria according to simulations.