Robust decorrelating decision-feedback multiuser detection in non-Gaussian channels

Abstract

Multiuser detection research has been pursued under the Gaussian noise hypothesis. However, in many realistic channels where impulsive noise sources are ubiquitous, the Gaussian statistical model is hardly justifiable. There is, therefore, a strong motivation for the development of robust non-Gaussian signal processing techniques to safeguard against the influence of outliers from degrading detector performance in impulsive channels. This paper investigates a simple approach to robustify the decorrelating decision-feedback (DDF) multiuser detector. The proposed detector involves a chip-based nonlinear front-end for impulsive noise filtering followed by the classical DDF detection. The nonlinear front-end exploits knowledge of the users’ signal amplitudes to constrain the useful signals to fall within the linear region of the nonlinear clipping function. The performance of the proposed robust DDF detector is investigated through extensive computer simulations, and it is shown that substantial improvement in performance can be achieved by incorporating the nonlinear front-end when the channel noise follows heavy-tailed non-Gaussian distributions.