【作者】 罗柏文；

【导师】 于宏毅；

【作者基本信息】 解放军信息工程大学 ， 通信与信息系统， 2013， 博士

【摘要】 如何保障微弱信号的高效可靠接收一直是现代通信中的一个难点。目前由于技术水平的限制，天线口径不可能无限制提高，接收机噪声也很难进一步降低，一种有效的解决方法是利用多个天线对同一信号进行联合接收，将接收信号进行合成以提高信号质量。通过多天线信号的有效合成，增大了天线的等效口径，从而能够在较低的信噪比条件下实现高速可靠的数据通信。随机布设多天线信号合成具有应用灵活、成本较低以及性能稳健等优点，在众多领域具有广泛应用前景。目前该技术的研究工作主要针对深空探测这一特殊应用领域开展。在更为一般的场景中，利用多个天线所接收的信号进行合成，需要解决无先验信息条件下的信号间时延差异估计、频率相位差异估计及合成权值估计等问题。解决这些问题是该技术获得进一步推广应用的关键所在。本文主要围绕随机布设多天线信号波形合成中的关键技术开展研究，针对多天线信号的最佳合成权值估计、时延差异估计补偿、频差估计补偿以及多径信道下的合成问题进行讨论分析。论文的主要创新点可以概括为以下几个方面：1、针对基于最大输出信噪比和最大输出功率准则下的经典合成权值估计方法中存在的噪声方差难以准确估计、估计量有偏等问题，本文首先从理论上证明了以自相关系数为目标函数与以信噪比为目标函数在进行信号合成权值估计中的等价关系，基于上述等价关系，分别提出了针对两路信号和多路信号合成的最佳权值估计算法。传统的以合成信号信噪比作为目标函数的算法不能用于噪声方差难以准确估计的场合。而当各路信号的噪声方差不相等时，以合成信号功率为目标函数的算法为有偏估计。为了解决这些问题，本文通过理论推导证明了合成信号的自相关系数作为信号合成目标函数的正确性。根据该目标函数，提出了一种针对两路信号合成的ACE算法。针对多路信号合成权值估计问题，本文用特征值分解的方法提出了AC EIGEN算法。并利用一组信号相关矩阵的线性组合代替AC EIGEN算法中的信号相关矩阵，给出了MAC EIGEN算法，进一步提升了低信噪比条件下的合成性能。上述算法无需定时同步，与信号调制方式、信号带宽无关，无需估计噪声方差，适用于噪声方差不等、存在相关噪声的情景，具有更好的通用性。仿真结果表明，AC EIGEN算法和MAC EIGEN算法在低信噪比条件下的合成性能优于传统方法。2、证明了以归一化峰度为目标函数和以信噪比为目标函数进行最佳合成权值估计的等价关系，提出了一种以归一化峰度为目标函数的合成权值迭代搜索算法。仿真结果表明，通过迭代可以获得比ACE算法更低的方差性能。通过理论推导发现以合成信号的归一化峰度、信号瞬时功率归一化均方差等统计量在作为合成权值估计的目标函数，与以合成信号信噪比为目标函数是等价的。根据该等价关系，基于信号分量和噪声分量统计特性不同的假设，提出了以合成信号归一化峰度为目标函数的合成权值迭代搜索算法。该迭代搜索算法无需估计噪声方差，对信号调制参数透明，也无需定时同步。通过仿真发现，该算法可用于相关噪声条件下合成权值估计，并且通过迭代可以获得比ACE算法更低的方差性能。3、针对传统的自适应时延估计算法在多天线信号时延对齐中性能较差的问题，提出了以合成信号为参考的多信号时延自适应联合估计算法。从理论上证明了该算法的收敛性和估计量的渐进无偏性，并给出了算法估计量方差性能的理论分析结果。理论分析结果与仿真表明，与基于单路信号为参考的传统自适应时延估计算法相比，该算法降低了估计的方差，提高了多天线信号自适应时延对准性能。传统的自适应信号时延差异估计算法只针对两路信号之间的时差进行估计和补偿，而在例如多天线信号合成、目标定位等应用中，往往需要估计多路信号的时延差异。本文提出的自适应时延联合估计算法，以合成输出信号为公共的参考信号，各路信号分别根据与参考信号的误差自适应调整滤波器参数，最终通过迭代实现时延差异的估计与信号的对齐合成。本文对算法的收敛性给予了证明，并且理论推导了算法的均值特性和方差特性。仿真结果证明了理论推导的正确性。4、将对多天线信号之间频差问题转化为线性时变相位问题，本文首先提出了一种自适应相位差异估计补偿算法，证明了算法的收敛性和渐进无偏性。将该算法推广应用于信号间频差估计与补偿，提出了一种方差性能接近CRLB的无偏估计算法和一种方差性能良好且偏差可控的有偏估计算法。针对多天线信号之间的线性时变相位差异，按照相位差信号维纳解的形式，提出了渐进无偏的约束自适应相位差异估计补偿算法。基于该算法的相位差估计结果，本文分别利用时间平均的方法和线性拟合的方法提出了两种频差估计算法。理论推导和仿真结果显示基于时间平均的频差估计算法为有偏估计。同时仿真结果还显示，基于时间平均的频差估计算法的方差低于无偏估计的CRLB，而基于线性拟合的频差估计算法是一种性能接近CRLB的无偏估计算法。5、针对在无精确信道估计条件下，时域波形合成很难使得信号带宽范围内所有频点的信噪比最大化的问题，根据多径衰落信道下最佳合成架构，提出了基于余弦调制滤波器组的子带合成算法。仿真结果显示，该算法合成输出信号的解调质量优于最大比合并分集信号。根据宽带衰落信号窄带化处理的思想，提出了一种子带合成算法。该算法采用余弦调制分析综合滤波器组将接收信号分为若干带宽很小的子带信号，并且分别将各路对应的子带信号进行最大比合并。该算法对信号调制方式透明，无需进行定时同步，也无需估计信道响应。仿真显示该子带合成算法性能优于最大比合并的分集算法。更多还原

【Abstract】 How to efficiently and reliably receive the weak signal is one of major problems andchallenges for communication technology. Due to the limitations of the technical level, it’s verydifficult to further increase the antenna aperture or reduce the receiver noise. An effectivesolution is to use multiple antennas to joint receive the source signal. Received signals arecombined in order to improve the signal quality. The multi-antenna signal may be combined toimprove the effective aperture of the antenna. This is an effective method to solve thecontradiction between the low signal-to-noise ratio and high-speed communications needs.Multiple signals combining technology of random distribution antennas has the advantages offlexible, low cost and robust performance. It can be applied in many fields. Currently, thetechnology is mainly researched and used in the special environment of deep space exploration.Promotion and application of the technology in a more general application environment is facingmany difficulties such as irregular distribution of antennas, the individual differences of thereceiver and fading channel problems. To solve the problems under more general applicationenvironments is the key to further promote the application of the technology that must beaddressed.The researching works of this paper around the key technologies of signal waveformcombining on randomly distribution antennas. The time delay estimation and compensation, thefrequency difference estimation and compensation, the optimal combining weight estimation andcombining technology of multipath fading channel are discussed and analyzed in this paper. Themain work and innovations of the thesis can be summarized as following aspects.1. The classical combining algorithms under the criteria of maximum outputsignal-to-noise ratio and the maximum output power have the problems of estimating noisevariance and the biased ness of estimation. This paper presents and proves a new objectfunction, the combined signal autocorrelation coefficient, which is equivalent to thesignal-to-noise ratio in combining. The criterion of maximum combined signalautocorrelation coefficient is proposed for optimal signal combining. This paper alsoproposed combining weight estimation algorithms for two signals and multiple signalsrespectively according to the new criterion. For conventional combining weight estimation,algorithms using the combined signal SNR as the object function fail in some applicationscenarios which are very different to estimate the noise variance. The estimating algorithmsusing the combined signal power as the object function is biased with not uniform noisevariances. To solve these problems, this paper proposed and proved that the combined signalautocorrelation coefficient can be used as the object function in signal combining, which is equivalent to the combined signal SNR. Using this object function, the paper proposed a simplealgorithm for2signals combining. Using the eigenvalue decomposition method, the paperproposed the AC EIGEN algorithm for multi-signal combining weight estimation. Using a linearcombination of a set of signal correlation matrices instead the correlation matrix, the MACEIGEN algorithm is proposed. And the combining performance under low SNR environment isfurther enhanced. These algorithms do not need timing synchronization, has nothing to do withthe signal modulation, the signal bandwidth, no need to estimate the noise variance of signal, canalso be applied to scenarios of non-uniform noise power and related noise. Algorithms are withexcellent performance and high universality. Simulation results show that proposed algorithmsoutperform classical algorithms under low SNR environment.2. The equivalence between combined signal normalized kurtosis and SNR is proved.Using the new criterion, an iterative algorithm is proposed. The simulation results showthat iteration can get a lower variance than ACE algorithm. This paper found that statisticssuch as the combined signal normalized kurtosis, normalized mean square error of signalinstantaneous power are equivalent to the combined signal SNR in signal combining as theobject function. According to the difference of statistical properties of signal and noise, aniterative algorithm is proposed based on it. These algorithms do not need timing synchronization,has nothing to do with the signal modulation, the signal bandwidth, no need to estimate the noisevariance of signal, can also be applied to scenarios of non-uniform noise power and related noise.Simulation results show that proposed algorithms outperform the ACE algorithms with iteration.3. The paper proposed a joint adaptive estimation and compensation algorithm basedon signal combining for poor performance of traditional adaptive time delay estimationalgorithm in multi-antenna signal delay alignment. The paper theoretically proves theconvergence of the algorithm and the progressive unbiasedness of the estimation. Thepaper gives the theoretical analysis of variance performance. Theoretical analysis andsimulation results show that the new algorithm effectively reduces the variance of theadaptive time delay estimation, and it improves the alignment performance ofmulti-antenna signals compared with the conventional adaptive algorithm. The traditionaladaptive time delay estimation algorithms estimate and compensate difference of time delaybetween two signals. There are greater than or equal to3signal delay differences needs to beestimated in applications such as multi-antenna signal combining, targeting locating. Theproposed algorithm uses the combined signal as the common reference signal for joint adaptivetime delay estimate. The algorithm adaptive adjusts the filtering parameters of each signalrespectively based on the error with the reference signal. Ultimately, the algorithm achieves theestimate of time delays and the alignment of all signals through the iterative. Performance derivation and numerical simulation results show that the algorithm is effective to reduce thetime delay estimation variance, and to improve signal alignment performance.4. Frequency offset between the multi-antenna signals is transformed into lineartime-varying phase shift. This paper derives an asymptotic unbiased constraint adaptivephase shift estimation and compensation algorithm. The algorithm is extended to thefrequency difference estimation and compensation. Two frequency difference estimationalgorithms with variance performance close to the CRLB are proposed. According to theWiener solution in the form of phase shift signals, the paper proposes an asymptotic unbiasedconstraint adaptive phase shift estimation and compensation algorithm. Based on estimationresults of the phase shift estimation, this paper proposed two frequency difference estimationalgorithms using the time-averaged method and linear fitting method. The simulation resultsshow that the frequency difference estimation algorithm based on the average method is anasymptotic unbiased algorithm, with the variance lower than the CRLB. And the algorithm basedon the linear fit method is an unbiased algorithm with the variance close to the CRLB.5. Maximum ratio combining diversity algorithm in multipath fading channels can notmaximize the full-band signal-to-noise ratio of the combined signal. According to theoptimal combining theory of multipath fading signals, this paper proposed a sub-bandcombining algorithm using the classical cosine-modulated filter banks. Simulation resultsindicate that the demodulation performance of sub-band combined signal outperforms themaximum ratio combining diversity signal. According to the optimal combining model of thefading channel, the cosine modulated filter banks were introduced into the signal combining.And a sub-band combining algorithm is proposed. The classical cosine modulated analysis filterbanks are used to narrow band the received signals. And the corresponding sub-band signals arecombined with the criterion of maximum ratio combining. The algorithm is transparent to thesignal modulation method, without timing synchronization, makes the combined signal tomaximize the signal-to-noise ratio, and improves the quality of signal demodulation. Simulationresults indicate that the demodulation performance of sub-band combined algorithm outperformsthe maximum ratio combining diversity algorithm.更多还原