9.1 Introduction
| As noted in Chapter 1, mobile wireless communication systems are affected by propagation anomalies due to terrain or buildings which cause multipath reception , producing extreme variations in both amplitude and apparent frequency in the received signals, a phenomenon known as fading . Signal reception in such channels presents new challenges, and dealing with these is the main theme of this chapter. Channel estimation and data detection in various fading channels have been the subjects of intensive research over the past two decades. In what follows we provide a brief overview of the literature in this area. Single- User Receivers in Frequency-Flat Fading ChannelsNarrowband mobile communications for voice and data can be modeled as signaling over flat, i.e., frequency-nonselective, Rayleigh fading channels. Depending on the fading rate relative to the data rate, the fading process can be categorized as either slow ( time-nonselective ) fading , where the fading process is assumed to remain constant over one symbol interval and to vary from symbol to symbol; or fast ( time-selective ) fading , where the fading process is assumed to vary within the symbol interval. A considerable amount of recent research has addressed the problem of data detection in flat fading channels. Specifically, various techniques for maximum- likelihood sequence detection (MLSD) in slow-fading channels have been proposed. The optimal solutions under several fading models are studied in [167, 284, 308], and the exact implementations of these solutions involve very high dimensional filtering. Most suboptimal schemes employ a two-stage receiver structure, with a channel estimation stage followed by a sequence detection stage. Channel estimation is typically implemented by a Kalman filter or a linear predictor , and is facilitated by per- survivor processing [399, 525], decision feedback [167, 223, 280], pilot symbols [70, 94, 233, 337], or a combination of the above [215]. Other alternative solutions to MLSD in slow-fading channels include a method based on a combination of hidden Markov model and Kalman filtering [85, 86], and an approach based on the expectation-maximization (EM) algorithm [138]. Moreover, in [139, 181], turbo receiver techniques for joint demodulation and decoding in flat-fading channels are developed. Data detection over fast-fading channels has also been addressed in the recent literature. In [171, 524, 526], a linearly time-varying model is used to approximate the time variation within a symbol interval of a time-selective fading process, and several double-filtering receiver structures are developed. Another approach [360] that has been investigated is to sample the received signal at a multiple of the symbol rate, and to track the channel variation within a symbol interval using a nonlinear filter. Extensions of this method have been made to address the issue of tracking the random phase drift [249, 250] and the carrier frequency Doppler shift [359] in fast-fading channels. Single-User Receivers in Frequency-Selective Fading ChannelsMultipath effects over fading channels that cause time-varying intersymbol interference (ISI) constitutes a severe impediment to high-speed wireless communications. Although equalization of time-invariant channels has been an active research area for almost four decades [175], equalization of time-varying fading channels presents substantial new challenges and has received significant attention only recently, due to its potential for widespread application in high-speed wireless data/multimedia applications. Maximum-likelihood sequence estimation receivers for time-varying ISI channels with known channel state information are studied in [49, 171], which are generalizations of the Ungerboeck receiver for time-invariant channels [502]. In [93, 314, 489, 599], several MLSD receiver structures are developed that are based on the known second-order statistics of the fading process, instead of the actual channel state. When the fading statistics are unknown, they are usually estimated from the data in a training-assisted mode or decision-directed mode [96, 172, 240, 246, 268, 503]. Furthermore, symbol-by-symbol maximum a posteriori probability (MAP) schemes for equalizing time-varying fading channels have also been studied [22 “24, 481], where channel estimation is facilitated by some ad hoc Kalman-type nonlinear estimators, which take as inputs the a posteriori probabilities of the ISI channel state and the received signal. Related to these methods are the Bayesian equalization techniques developed for time-invariant ISI channels [77, 149, 213, 245], which essentially model the channel coefficients as slowly time-varying processes. Moreover, orthogonal frequency-division modulation (OFDM) techniques convert a frequency-selective fading channel into a set of parallel frequency-flat fading channels. Channel estimation and data detection methods in OFDM systems are developed in [110, 259, 260, 262, 263]. Another approach to equalization of time-varying channels, found in the signal processing literature [271, 486, 488], is to model the time-varying channel impulse response function by a superposition of deterministic time-varying basis functions (e.g., complex exponentials) with time-invariant coefficients [147]. Such a model effectively converts the time-varying ISI channel into a time-invariant ISI channel. High-order statistic (HOS) “based and second-order statistic (SOS) “based equalization methods for time-invariant channels can then be employed to identify the channel coefficients and thus identify and equalize the time-varying channel. Multiuser Receivers in Fading ChannelsData detection in multiuser fading channels has been addressed from a number of perspectives. Derivation and analysis of the optimum multiuser detection schemes under various fading channels are found in [65, 434, 513 “516, 547, 615]. Suboptimal linear multiuser detection methods for fading channels are developed in [227, 434, 468, 547, 577, 616, 617]. Techniques for joint fading channel estimation and multiuser detection that are based on the EM algorithm are proposed in [91, 119]. Moreover, adaptive linear multiuser detection in fading channels has been studied in [26, 184, 188, 547, 613]. A few recent works have addressed the exploitation of coded signal structure in sequence estimation. In [597] the reduced-state sequence estimation [106, 113, 114, 560, 561] technique is integrated with an error-detection code for channel equalization. In this method, some subset of the set of all possible paths in the trellis is generated to satisfy the code constraints. Similar ideas have also been applied to joint channel and data estimation where the estimation procedure is forced to yield valid code-constrained path sequences [54, 55, 212]. In this chapter we present several advanced methods for signal reception in fast-fading channels. This presentation is organized as follows. In Section 9.2 we discuss statistical modeling of multipath fading channels. In Section 9.3 we present coherent receiver techniques for fading channels based on the expectation-maximization algorithm. In Section 9.4 we discuss decision-feedback -based low-complexity differential receiver techniques for fading channels. Finally, in Section 9.5 we present adaptive receiver techniques for fading channels that are based on the sequential Monte Carlo methodology described in Chapter 8. The following is a list of the algorithms appearing in this chapter.
|
EAN: 2147483647
Pages: 91