A self-organising nonlinear noise filtering scheme

In this paper we present a new adaptive algorithm for suppression of impulse noise. The algorithm is based on a special combination of impulse detection and nonlinear filtering where only a small number of parameters is required. In contrast to conventional approaches where parameters have to be trained first, we propose a new unsupervised learning method which is related to blind equalizers and self-organizing maps. Thereby, we dramatically reduce the necessary a-priori information as well as the computational complexity. Further, simulation results show that the performance of the new self-organizing algorithm is equivalent to that of a previously reported method with supervised training which is superior over many other existing techniques for impulse noise removal.