This work presents original developed methods, belonging to the soft computing area, that have been applied to solve complex signal processing problems arising in communication systems with ?adverse? conditions. The adverse conditions are given by the fact that the considered physical channel is non- linear, non stationary, and the noise is coloured impulsive noise. Modified vector quantization algorithms have been developed for adaptive channel estimation and equalization inside a digital transmission. An unsupervised hierarchical clustering method based on fuzzy partitions has been proposed and application to identification of different noise sources is discussed. The presented soft computing techniques have been developed within the area of digital communication, nevertheless they posses general applicability.