Blind Bayesian multiuser detection for impulse radio UWB systems

Abstract

In this thesis, we address the blind parameter estimation and multiuser de- tection problems for impulse radio ultra-wide band (UWB) systems under frequency selective fading. The frequency selective UWB channel, having a large number of taps combined with the noise effect, causes certain problems in an UWB system. This channel is often subject to non-Gaussian noise, and with frequency selectivity, there are catastrophic effects. An impulse radio UWB transmitter with orthogonal spreading instead of replication along frames is proposed. In this system, the mentioned blind parameter estimation and multiuser detection problems are considered for three different noise types, Gaussian, impulsive and Cauchy noise. One of the most popular Bayesian estimators, Gibbs sampler, is employed at the receiver part for calculating the Bayesian estimates of the unknown parameters. Because a Gibbs sampler is a soft-input soft-output module capable of exchanging probabilistic information, the proposed detector is also employed within a turbo multiuser detection structure for coded UWB systems. Using this iterative decoding mechanism, further performance improvement is obtained for Gaussian and impulsive noise scenarios, but the sampler fails to converge for the Cauchy case. The bit error rate results for both the Gaussian and impulsive noise cases are given in comparison. Also the tracking performance of the sampler for the unknown parameters are provided for both cases. The simulation results show that the Gibbs sampler is effective in estimating the system parameters and that the proposed receiver provides significant performance gains after a few detection / decoding iterations. It is also concluded that with a proper system design that avoids inter-frame collisions and although an amount of complexity is introduced, the system is suitable for tight bit error rate requirements.