4.1 Introduction


As we have seen in preceding chapters, the use of multiuser detection (or derivative signal processing techniques) can return performance in multiuser channels to that of corresponding single- user channels, or at least to a situation in which performance is no longer limited by the multiple-access interference (MAI). Thus far, our discussion of these problems has focused on the situation in which the ambient noise is additive white Gaussian noise (AWGN). This was an appropriate model in previous chapters, since the focus there was on the mitigation of the most severe noise source, the MAI. However, as increasingly practical techniques for multiuser detection become available, such as the methods discussed in Chapters 2 and 3, the situation in which practical multiple-access channels will be ambient-noise limited can be realistically envisioned .

In many physical channels, such as urban and indoor radio channels [44, 45, 319, 320, 323] and underwater acoustic channels [52, 321], the ambient noise is known through experimental measurements to be decidedly non-Gaussian, due to the impulsive nature of the human-made electromagnetic interference and of a great deal of natural noise as well. (For measurement results of impulsive noise in outdoor/indoor mobile and portable radio communications, see [44, 45] and the references therein.) It is widely known in the single-user context that non-Gaussian noise can be quite detrimental to the performance of conventional systems designed on the basis of a Gaussian noise assumption, whereas it can actually be beneficial to performance if appropriately modeled and ameliorated. Neither of these properties is surprising. The first is a result of the lack of robustness of linear and quadratic signal processing procedures to many types of non-Gaussian statistical behavior [226]. The second is a manifestation of the well-known least favorability of Gaussian channels [128].

In view of the lack of realism of an AWGN model for ambient noise arising in many practical channels in which multiuser detection techniques may be applied, natural questions arise concerning the applicability, robustness, and performance of multiuser detection techniques for non-Gaussian multiple-access channels. Although performance indices such as mean-square error (MSE) and signal-to-interference-plus-noise ratio (SINR) for linear multiuser detectors are not affected by the amplitude distribution of the noise (only the spectrum matters), the more crucial bit-error rate can depend heavily on the shape of the noise distribution. Results of an early study of error rates in non-Gaussian direct-sequence code-division multiple-access (DS-CDMA) channels are given in [1 “3], in which the performance of the conventional and modified conventional (linear matched filter) detectors is shown to depend significantly on the shape of the ambient noise distribution. In particular, impulsive noise can severely degrade the error probability for a given level of ambient noise variance. In the context of multiple-access capability, this implies that fewer users can be supported with conventional detection in an impulsive channel than in a Gaussian channel. However, since non-Gaussian noise can, in fact, be beneficial to system performance if treated properly, the problem of joint mitigation of structured interference and non-Gaussian ambient noise is of interest [378]. An approach to this problem for narrowband interference (NBI) suppression in spread-spectrum systems is described in [133]. A further study [383] has shown that the performance gains afforded by maximum- likelihood (ML) multiuser detection in impulsive noise can be substantial compared to optimum multiuser detection based on a Gaussian noise assumption. However, the computational complexity of ML detection is quite high (even more so with non-Gaussian ambient noise), and therefore effective near-optimal multiuser detection techniques in non-Gaussian noise are needed. In this chapter we address the MAI mitigation problem in DS-CDMA channels with non-Gaussian ambient noise.

In the past, considerable research has been conducted to model the non-Gaussian phenomena encountered in practice which are characterized by sharp spikes, occasional bursts, and heavy outliers, resulting in a variety of statistical models, the most common of which include the statistically and physically derived Middleton mixture models [319 “323], the empirical Gaussian mixtures, and other heavy-tailed distributions, such as the Weibull, the K, and the log-normal, as well as the stable models [357]. Particularly accurate are the Middleton models, which are based on a filtered-impulse mechanism and can be classified into three classes, A, B, and C. Interference in class A is coherent in narrowband receivers, causing a negligible number of transients. Interference in class B is impulsive, consisting of a large number of overlapping transients. Interference in class C is the sum of the other two types of interference. The Middleton model has been shown to describe actual impulsive interference phenomena with high fidelity; however, it is mathematically involved for signal processing applications. In this chapter we use the widely adopted two- term Gaussian mixture distribution (which gives a good approximation to the Middleton models) to model the non-Gaussian noise, and discuss various robust multiuser detection techniques based on such a model. In the end, we will show that these robust signal processing techniques are also very effective in ameliorating other types of non-Gaussian noise, such as symmetric stable noise.

This chapter is organized as follows . In Section 4.2 we discuss robust multiuser detection techniques based on M -regression. In Section 4.3 we present asymptotic performance analyses for the robust multiuser detectors. In Section 4.4 we discuss implementation issues arising in robust multiuser detection. In Section 4.5 we treat the topic of robust blind multiuser detection. In Section 4.6 we present improved versions of robust multiuser detectors based on local likelihood search. In Section 4.7 we discuss robust group -blind multiuser detection. In Section 4.8 we consider robust multiuser detection in multipath channels. Finally, in Section 4.9 we briefly introduce a -stable noise and illustrate the performance of various robust multiuser detectors in such noise. The proofs of some results in this chapter are appended in Section 4.10.

The following is a list of the algorithms appearing in this chapter.

  • Algorithm 4.1: Robust multiuser detector ”synchronous CDMA

  • Algorithm 4.2: Robust blind multiuser detector ”synchronous CDMA

  • Algorithm 4.3: Adaptive robust blind multiuser detector ”synchronous CDMA

  • Algorithm 4.4: Robust multiuser detector based on slowest-descent-search ”synchronous CDMA

  • Algorithm 4.5: Robust group-blind multiuser detector ”synchronous CDMA

  • Algorithm 4.6: Robust blind multiuser detector ”multipath CDMA

  • Algorithm 4.7: Robust group-blind multiuser detector ”multipath CDMA



Wireless Communication Systems
Wireless Communication Systems: Advanced Techniques for Signal Reception (paperback)
ISBN: 0137020805
EAN: 2147483647
Year: 2003
Pages: 91

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net