OFDM系统中基于面向判决的信道估计研究

发布时间：2018-08-15 13:26

【摘要】：正交频分复用(Orthogonal Frequency Division Multiplexing, OFDM)系统在对抗多径效应时具有较高的可靠性,这使得OFDM在现代无线通信系统中得到了广泛的应用,例如基于长期演进(Long Term Evolution, LTE)的第四代移动通信中正是采用了OFDM作为物理层的关键技术。在OFDM系统中,频率选择性信道被转换为一组并行的频率平坦信道,这组频率平坦信道的信道响可以用一组具有单抽头系数的均衡器来补偿。这使得OFDM系统可以极大的简化均衡器设计同时维持较高的数据速率。正因如此,OFDM已经在许多商业无线系统中被看作是一种标准的物理层技术。为了进行相干解调,信道估计是OFDM接收机中必不可少的一个环节。传统的信道估计技术采用基于导频符号(Pilot)的方式进行。为了改进信道估计质量,可以采用面向判决(Decision-Directed, DD)技术,将判决后的数据符号反馈给信道估计器。假设判决结果是正确无误的,这相当于提高了导频符号的密度,从而改善信道估计的性能。传统的面向判决信道估计只能得到数据符号的硬判决结果,这种情况下的反馈方式也是唯一的,即将硬判决结果直接反馈给信道估计器。随着Turbo接收机技术的采用,软判决方式得到了广泛应用并显著改善了接收机的性能。软判决方式并不给出单一的判决结果,而是在获得接收信号的条件下,给出星座图中每个星座点的后验概率(A-posteriori Probability, AP)。相应地,在进行面向判决的信道估计时,信道估计器从判决器获得的也不再是硬判决结果,而是每个星座点的后验概率。与基于硬判决的面向判决不同,由于同时拥有每个星座点的出现概率,信道估计器可以有多种方法来利用这些后验概率进行面向判决的信道估计。因此,反馈的方式不再唯一,而是有多种可能性。所以,需要研究最优的反馈方式以达到最佳的信道估计性能,也就是说,需要解决软判决方式下信道估计的最优面向判决问题。在一般的中低速移动场景下,信道在一个OFDM符号内产生的变化通常可以忽略,因此,传统的OFDM系统中通常假设信道在一个OFDM符号持续时间内不会发生改变。然而,随着终端移动性提高,多普勒频移也相应增加。在这种情况下,信道变化十分快速,在一个OFDM符号内随时间的变化已经较为明显。若仍然假设信道在一个OFDM符号内不变,会带来较大的估计误差。因此,为了改进快变信道条件下信道参数的估计性能,需要将信道在一个OFDM符号内的变化考虑进去。传统的信道估计技术并不适用于这种情况,所以,需要研究快速时变信道条件下的信道估计问题。提出了一种基于最大后验概率准则(Maximum A-posteriori Probability, MAP)的迭代信道估计技术,用以解决单发送天线、单个OFDM符号系统中的最优面向判决问题。所得到的结果表明,应该采用调和平均的方法来构造信道估计时所用到的软数据符号。虽然基于MAP准则得到的方程没有解析解,但可以通过不动点迭代(Fixed Point Iteration, FPI)的方法迭代求解。进一步发现所提出的基于不动点迭代的MAP估计与期望-最大(Expection-Maximization, EM)算法具有紧密的联系。这种联系可以用来证明所提出的迭代信道估计算法的单调收敛特性。同时,还证明了所提出的方法在大信噪比(Signal-to-noise ratio, SNR)条件下具有单步收敛特性。仿真结果表明,所提出的算法在大信噪比条件下可以达到Bayesian Cramer-Rao界(CRB)。将基于MAP准则的迭代信道估计技术一般化,解决了多发送天线、多OFDM符号系统中的最优面向判决信道估计问题。有多根发送天线时,期望发送天线上应该采用调和平均的方式来构造软数据符号,而干扰发送天线上则应该采用算术平均的方式来构造软数据符号。同时,当考虑到时域上连续的多个OFDM符号后,信道参数在多个OFDM符号间的时变特性以及相关性也可以用来改进信道估计的性能。与单发送天线、单OFDM符号的情况类似,所提出的迭代信道估计算法是基于MAP准则推导得到的,并应用不动点迭代的方法来迭代求解。所提出的方法同样可以保证收敛特性,同时,当信噪比足够大时,该算法仍然具有一步收敛特性。研究了快速时变信道条件下OFDM系统中的信道估计技术。时变信道可以采用基展开模型(Basis Expansion Model, BEM)的方法来表示。对于BEM模型,基函数是已知的,只需要估计BEM模型的展开系数。本文中,考虑了连续的OFDM符号间的BEM模型系数的相关性,并基于最小均方误差准则(Minimum Means-Square-Error, MMSE)推导出了最优的BEM系数估计器。采用本文所提出的方法,当前OFDM符号的BEM系数可以通过对相邻OFDM符号的BEM系数作线性组合的方式得到,因此不再需要为每个OFDM符号都插入导频。本文中还考虑了实际的接收机,包括基于限脉冲响应滤波器的实现方案以及一种面向判决的迭代接收机结构。
[Abstract]:Orthogonal Frequency Division Multiplexing (OFDM) systems have high reliability against multipath effects, which makes OFDM widely used in modern wireless communication systems. For example, in the fourth generation mobile communications based on Long Term Evolution (LTE), OFDM is used as physics. Key technologies of the layer. In OFDM systems, frequency selective channels are converted into a set of parallel frequency-flat channels whose channel responses can be compensated by a set of Equalizers with single-tap coefficients. This allows OFDM systems to greatly simplify equalizer design while maintaining high data rates. OFDM has been regarded as a standard physical layer technology in many commercial wireless systems.
In order to perform coherent demodulation, channel estimation is an indispensable part of OFDM receiver. Traditional channel estimation techniques are based on Pilot symbols. In order to improve the quality of channel estimation, decision-directed (DD) technique can be used to feedback the decision data symbols to the channel estimator. If the decision result is correct, it is equivalent to increasing the density of pilot symbols and improving the performance of channel estimation. Traditional decision-oriented channel estimation can only get the hard decision result of data symbols. In this case, the feedback method is also unique, that is, the hard decision result is directly fed back to the channel estimator. With the adoption of receiver technology, soft decision method is widely used and the performance of receiver is greatly improved. Soft decision method does not give a single decision result, but gives a posteriori probability (AP) of each constellation point in the constellation diagram under the condition of receiving the received signal. In channel estimation, the channel estimator obtains a posterior probability of each constellation point instead of a hard decision result. Unlike the hard decision-based decision-oriented channel estimator, which has the probability of each constellation point appearing simultaneously, the channel estimator can use these posterior probabilities for decision-oriented channel estimation in many ways. Therefore, the feedback method is no longer unique, but there are many possibilities. Therefore, it is necessary to study the optimal feedback method to achieve the best channel estimation performance, that is, to solve the problem of optimal decision-oriented channel estimation under soft decision mode.
In general low-speed mobile scenarios, the channel changes within an OFDM symbol are usually negligible. Therefore, in traditional OFDM systems, it is usually assumed that the channel does not change within the duration of an OFDM symbol. In order to improve the estimation performance of channel parameters under fast-varying channel conditions, it is necessary to take into account the channel changes within an OFDM symbol. The technique is not suitable for this case, so it is necessary to study the channel estimation problem under fast time-varying channel conditions.
An iterative channel estimation technique based on the Maximum A-posteriori Probability (MAP) criterion is proposed to solve the optimal decision-oriented problem for a single transmit antenna and a single OFDM symbol system. Although there is no analytical solution to the equation based on MAP criterion, it can be solved iteratively by Fixed Point Iteration (FPI). It is further found that the proposed MAP estimation based on fixed point iteration is closely related to Expection-Maximization (EM) algorithm. The monotone convergence property of the proposed iterative channel estimation algorithm is also proved. The proposed algorithm has one-step convergence property under the condition of large signal-to-noise ratio (SNR). Simulation results show that the proposed algorithm can reach the Bayesian Cramer-Rao bound (CRB) under the condition of large signal-to-noise ratio (SNR).
The iterative channel estimation technique based on MAP criterion is generalized to solve the problem of optimal decision-oriented channel estimation in multi-transmit antenna and multi-OFDM symbol systems. Similar to the case of single transmit antenna and single OFDM symbol, the proposed iterative channel estimation algorithm is based on MAP benchmark. The proposed method can also guarantee the convergence property. At the same time, the algorithm still has one-step convergence property when the signal-to-noise ratio is large enough.
Channel estimation techniques for OFDM systems with fast time-varying channels are studied. Time-varying channels can be represented by the Basis Expansion Model (BEM). For BEM models, the basis functions are known and only the expansion coefficients of BEM models are estimated. In this paper, the BEM model coefficients between consecutive OFDM symbols are considered. Based on the Minimum Means-Square-Error (MMSE), the optimal BEM coefficients estimator is derived. Using the proposed method, the BEM coefficients of current OFDM symbols can be obtained by linear combination of BEM coefficients of adjacent OFDM symbols, so it is no longer necessary to insert derivatives into each OFDM symbol. In this paper, we also consider the actual receiver, including the implementation scheme based on limited impulse response filter and a decision-oriented iterative receiver structure.
【学位授予单位】：北京交通大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN929.53

【共引文献】