| In2002, the US frequency regulator Federal Communications Commission allows theentrance of UWB technology to the commercial application. Since that time, industrial, academicand governmental research in that area has abounded. Moreover, US cable news networks remarkthe UWB technology as the top one in the ten most popular technologies in2004. At present, UWBwireless communication has been one of the popular physical layer technologies in the shortdistance and high-rate wireless networks. However, due to non-ideal communication environment,the frequency selectivity and fading of wireless channel restrains bit error rate performance,throughput, capacity and network flexibility of UWB systems. In case of knowing the givenchannel condition or obtaining accurate channel estimates, in general, the receiver can compensatethe distortion effects incurred by the transmission channel. Although the incoherent detectionadopted by the traditional communication systems can simplify the construction of the receiverthrough avoiding the complicated channel estimation, for AWGN channels, the penalty for the useof incoherent reception is only about3dB whereas this number increases significantly as the delayspread increases just like UWB channels. Similar to the existing communication systems, researchon channel estimation is a key issue and an important subject for UWB communications, i.e. howto retrieve the channel state information aimed at channel impulse response or channel frequencyresponse. Hence, we focus on the research of channel estimation theory and methods for UWBsystems. Several channel estimation methods have been proposed in this dissertation, whereestimation performance is analyzed in detail and simulations are also provided. The main contentsand contributions of this paper are listed as follows.(1) Presented is the research on blind identification based on SOS for the SIMO channels withthe cross relations (CR). Since the conventional batched algorithms concern singular valuedecomposition (SVD) or eigenvalue decomposition (EVD), they require a substantialcomputational complexity and are difficult to implement in a real time. Hence, adaptive blindmethods, i.e. MCN and MCLMS, are presented recently. Both schemes can avoid the matrixdecomposition but have an error floor of estimation performance. The MCN method converges at ahigher rate, at a cost of high complexity whereas the MCLMS method owns a low complexity butconverges slowly. In order to overcome these drawbacks, the paper proposes an iterative subspacemethod with a more faster convergence rate based on the inverse power (CRSI_INV). However,there is a contradiction between the convergence rate and the numerical stability due to the matrixinversion in the CRSI_INV. Then, another alternative approach is proposed based on perturbationanalysis and Taylor series expansion. Moreover, conducted is a first order perturbation analysis on mean square error performance of the proposed estimators. After that, an iterative identificationmethod based on CR subspace is proposed by employing power algorithm. Finally, simulationsdemonstrate that all proposed methods overcome the error floor effect of estimation performance;the two inverse iterative methods have better performance over the existing adaptive algorithms,and require less data samples; there exits a high consistency between the theoretical values basedon perturbation analysis and the experimental results; the performance of the proposed iterativemethod with a lower complexity is the same as the pre-proposed two methods and even exceedsthat of the multichannel Newton method.(2)“Structured” channel estimation methods based on the training sequence are investigated,where multipath delays and multipath gains are estimated jointly. Although the classical maximumlikelihood (ML) approach produces excellent results, it is impractical since the number ofparameters to estimate in a realistic UWB channel is very high and the non-linear search over ahigh dimensional parameter space is computationally prohibitive. The sliding window (SW) and thesuccessive cancellation (SC), the two suboptimal estimation algorithms for ML criterion areproposed recently. However, the simple SW algorithm causes significantly performance loss andthe SC algorithm introduces a large computational processing delay since channel parameters areestimated sequentially instead of in a parallel manner. In other words, a contradiction between theaccuracy and the processing time for channel estimation exhibits in UWB systems. To solve thisproblem, based on multiple iterations of least mean square (LMS) algorithm, the paper proposesthree iterative channel estimation schemes by refining the delays estimate, the path gains estimateor both of the output of the regular SW estimator in a parallel manner. As a result, any of threeschemes is more flexible due to the tradeoff between the processing time and the estimationaccuracy. Notably, the performance of the hybrid scheme, one of three schemes, can even approachthat of the SC method with iterations and gain a lower complexity, since it can tune both the delaysestimate and the path gains estimate iteratively.(3) We carry out research on an “unstructured” channel estimation method with the aid of thenoise varaince. The MMSE channel estimator has a significant performance gain by exploiting thecorrelation property of channel, but at the cost of high complexity. Although the complexity of LSmethod is small, its mean square error is relatively high. An alternative approach is time-domain MLestimation methods with a low complexity whereas the knowledge of the finite delay spread of thechannel should be acquired before channel estimation. A low complexity channel estimationscheme without requiring channel information is ideal in practical environment. Moreover, noisevariance is an important parameter since it can be exploited to assist in channel estimation andequalization. Thereby, the object of paper is to design a low complexity channel estimation schemeusing the noise variance but not the relevant channel information. Nevertheless, the conventional method of noise variance estimation cannot work in case of using the estimated channel coefficientsby LS criteria, whereas the LS channel estimator is easy to be implemented. Hence, in order torealize this object, we firstly propose a novel noise variance estimator using the LS channelcoefficient estimates with the aid of the Hermitian property of UWB channels. After that, a low-complexity channel estimation scheme based on threshold filtering is proposed where anyknowledge of channel statistics is not required. As a result, in case of optimal pilot sequence, theproposed scheme can almost attain the best estimation results while it can provide a tradeoff betweenestimation performance and computational complexity in case of non-ideal pilot sequence.(4) Aimed at a novel UWB physical layer transmission scheme, investigates the channelestimation and SNR estimation methods based on ML criteria. MIMO techniques are firstly appliedto SC-FDE UWB systems to obtain both frequency diversities and spatial diversities, for very highand reliable data transmission. Then, presented is a theoretical performance analysis result onconditional bit error rate (BER) by approximating the ISI as Gaussian noise. Furthermore, based onmaximum-likelihood principle, channel estimator and SNR estimator are presented to compute theMMSE equalizer’s coefficients. Conclusively, the proposed scheme significantly outperforms theSISO scheme on account of exploiting both spatial diversities and frequency diversities. Thetheoretical BER values can match well the simulated values over the highly frequency-selectiveUWB channels. The designed channel estimation scheme has an ideal performance to help theequalization of received signal.(5) Presented is rerearch on "unstructured" channel estimation robust to the NBI signal. In themeanwhile, robust noise variance estimation and NBI detection in the LS channel estimates are alsoinvestigated. Typically, two kinds of approaches are adopted for mitigating the effect of jamming onchannel estimation. In the first, they are obtained by modifying or redesigning the conventionallinear channel estimators, while in the second, the nonlinear channel estimators based on robuststatistics concept. Note that when the information on noise variance and NBI cannot be derivedaccurately, the first kind of existing estimators can degrade significantly. As far as the availablerobust estimators based on robust concepts are concerned, the reasonable choice of relevantperformance parameters based on robust concepts is a little difficult problem since there is acontradiction between the estimation efficiency and the robustness. Importantly, they have a littlehigher complexity. In order to overcome these drawbacks of the regular methods, we first constructinstantaneous noise samples so as to model the estimation of noise variance as parameter estimationof an exponential distribution and to turn the NBI detection in the LS channel estimates into ageneral constant false alarm rate problem. Then, a robust noise variance estimator based on orderstatistic is presented into avoiding the impact of NBI signal. Next, a robust channel estimationscheme using conditional median filtering is proposed on the basis of noise variance and NBI information. The proposed channel estimation method has advantages of performance andcomplexity over the conventional methods. Moreover, the proposed noise variance estimator andNBI detector can be applied into the other channel estimator and channel equalizer. |
PDF Full Text Request