In the digital communication there are still many open problems. These topics can be very diversive covering different digital modulation schemes and their usage in multiple access communication, and the methods of mitigating the problems caused by the channel like fading and multipath propagation. Currently we are working on the multiuser detection methods of a CDMA system.

     In the communication problems we face with difficulties when there is an impulsive contamination in the received signals. A few examples for the impulsive disturbing signal are man-made noises, the carrier of a near-by transmitter, the sounds emitted by sea mammals in underwater communication, different types of athmospheric phenomena like electrical discharging, etc.. In such circumstances the conventional signal processing techniques which use second order statistics (SOS) like correlation fail since they tend to infinity. One solution is to use fractional lower order statistics (FLOS) of the data. In these scenarios the impulsive signals can be modelled with alpha-stable distributions. Alpha-stable distributions are suitable for modelling many impulsive phenomena. These distributions also cover the Cauchy and Gaussian distributions. According to the generalized central limit theorem (GCLT) the limit distribution of independent and identically distributed random variables is stable. The shape parameter alpha controls the impulsiveness of the distribution. Other than the location parameter and the dispersion parameter which corresponds to the variance of the Gaussian distributions, alpha-stable distributions have a skewness parameter which enables one to model non-symmetric distributions. With alpha-stable process modelling FLOS-based counter parts of the usual SOS-based methods can be developed. One possible application is the sinusoidal parameter estimation in impulsive noise environments with model-based estimators using FLOS of the data.

     To separate the individual signals in a linear mixture is a problem encountered in many disciplines, varying from the cocktail party problem where we are interested only in the speech signal of a specific person, or the problem of extracting the heart signal of the fetus contaminated by the noise and the signals of the mother or separating the components in the cosmic microwave background. When we know little about the statistical properties of the signals in the mixture we talk of blind source separation (BSS). An important technique which is used in source separation is independent component analysis (ICA) where the signals in the observed linear mixture are assumed to be independent. We are interested in the application of ICA technique to communication problems where in a multiuser detection scenario of a CDMA system the observed signal can be thought of a mixture of independent components.