【作者】 张聪；

【导师】 卢焕章；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2009， 博士

【摘要】 波达方向(DOA)估计技术已广泛应用于雷达、声纳、通信、生物医学等多种军事和民用领域。常规的空间谱估计算法虽然具有超分辨的性能,但是对来波信号本身的时域信息利用并不充分,且在多信号处理能力、低信噪比适应能力等方面通常存在局限。随着研究的深入,人们发现许多人工信号和自然界信号都具有循环平稳特性,而利用信号的这种时域特性能够获得常规DOA估计方法无法达到的性能优势,如选择性测向能力、较强的干扰及噪声抑制能力和突破阵元数限制的多信号处理能力。因此,基于信号循环平稳特性的DOA估计技术逐渐成为阵列测向研究中的一个热点。现有基于循环平稳的DOA估计研究主要集中于一维问题,对实际中有广泛需求的二维DOA估计技术的研究较少,而且已有的二维DOA估计算法较普遍地存在由时延选择导致的稳健性问题,对非均匀阵列二维DOA估计的研究也基本属于空白。针对上述这些问题并同时考虑相干源、宽带源等实际环境和应用背景,本文对利用信号循环平稳特性的DOA估计技术进行了深入研究,主要包括以下内容:针对现有二维DOA估计技术存在由时延选择带来的算法稳健性问题,结合多时延采样及孔径扩展的空时等效阵列构造思想,分别提出了基于双平行均匀线阵的多时延共轭循环波达方向矩阵(ML-CCDM)算法和基于简化阵列的多时延共轭循环波达方向矩阵(SML-CCDM)算法。ML-CCDM算法充分利用了各阵元输出共轭循环相关函数的多时延采样数据,不仅避免了最优时延选择问题和谱峰搜索操作,还构造出了两倍于真实阵列孔径的虚拟阵列,相对于同样无需谱峰搜索的CCDM算法而言,获得了更高的DOA估计精度、更好的低信噪比适应能力以及更强的多信号处理能力。SML-CCDM算法在ML-CCDM算法的基础上,摆脱了双平行阵列的限制,并利用较少阵元获得了与ML-CCDM算法几乎相当的性能。研究了非均匀阵列二维DOA估计技术,将最小冗余线阵与基于多时延采样的空时等效阵列构造方法相结合,提出了两种不同结构的稀疏阵列以及分别基于这两种阵列的二维DOA估计算法,将最小冗余线阵的应用拓展到共轭循环平稳信号二维DOA估计领域。这两种基于稀疏阵列的算法不仅分别继承了ML-CCDM算法和SML-CCDM算法的优点,在阵元数相同的条件下还具有更大的阵列孔径,因而能获得更高的DOA估计精度。此外,相对于均匀阵列而言,稀疏阵列的阵元间距更大,因此在工程实现时能够降低阵元间互耦等问题的影响。研究了相干循环平稳信号DOA估计技术,通过构造与真实阵列阵元具有一一对应关系的空时等效虚拟阵列,并结合现有去相干预处理技术,提出了基于虚拟阵列前后向空间平滑(VA-FBSS)的算法,避免了前后向空间平滑循环MUSIC算法面临的最优时延选择问题并获得了更高的DOA估计精度。同时,与循环前后向线性预测算法相比,VA-FBSS算法在来波方向接近的低信噪比环境下具有更好的DOA估计性能。在此基础上,针对VA-FBSS算法只适用于一维DOA估计的局限,采用双平行均匀线阵将基于虚拟阵列空间平滑的方法推广到了相干循环平稳信号二维DOA估计领域。研究了宽带循环平稳信号二维DOA估计技术,在现有基于阵元输出循环自相关的算法基础上,归纳总结出一种基于阵元输出循环互相关的信号处理模型,并提出了相应的宽带信号二维DOA估计算法,进一步提高了DOA估计精度。更多还原

【Abstract】 Direction of arrival (DOA) estimation techniques have been widely used in many military and civil areas, such as radar, sonar, communication, biological engineering and so on. Although the ordinary spatial spectrum estimation techniques break the Rayleigh Resolution Limit, they do not sufficiently exploit the signals’time domain information. And they usually have some drawbacks in multiple signal processing and low SNR signal processing. With the development of researches, people found that many man-made and natural signals possess the character of cyclostationarity. And, by exploiting the cyclostationarity of the signals, better DOA estimation performances in signal selective direction finding, interference or noise suppression and multiple signal processing can be achieved. Thus, the cyclostationarity-based DOA estimation techniques have been gradually becoming one of the hot research topics in array signal direction finding areas.Most of the cyclostationarity-based DOA estimation researches are focused on the one-dimensional (1-D) case. As for the two-dimensional (2-D) case, which is much more required, the existing researches are inadequate, and they commonly suffered from the lag selecting problem. Besides, few researches on the non-uniform array-based 2-D DOA estimation can be found. Considering the above limitations along with the practical environments with coherent signals and wideband signals, the following topics on the cyclostationarity-based DOA estimation techniques are investigated in this dissertation.To improve the robustness of the DOA estimation, which is threatened by the lag selecting problem, the ML-CCDM and the SML-CCDM are proposed. Both methods are based on the idea of combing the multiple lag sampling and the aperture extension oriented spatial-time equivalent array constructing. The ML-CCDM utilizes two parallel uniform linear arrays. By exploiting the conjugate cyclic correlations of the sensor outputs at multiple lags, this method not only avoids the optimal lag selecting problem or the 2-D spectrum peak searching, but also obtains a space-time equivalent array with twice the aperture of the real array. Compared with the CCDM, the ML-CCDM achieves better performance in estimation accuracy, noise suppressing and multiple signal processing. Based on the framework of the ML-CCDM, the SML-CCDM not only gets rid of the drawbacks of two parallel geometry arrays, but also obtains a performance which is almost the same as that of the ML-CCDM with fewer sensors.By combining the minimum redundancy linear array (MRLA) with the multiple lag sampling-based spatial-time equivalent arrays constructing idea, two sparse arrays along with the corresponding 2-D DOA estimation methods are proposed. These methods extend the application of MRLA to the 2-D DOA estimation of conjugate cyclostationary signals. The two sparse array-based methods not only inherit the merits of ML-CCDM and SML-CCDM separately, but also achieve larger array aperture with the same number of sensors, which also leads to an improved accuracy in DOA estimation. Besides, when compared with the uniform array, the sparse array possesses larger sensor distances. Therefore, the effects of mutual coupling in real-world implementations can be reduced by the above sparse array-based methods.By combing the spatial-time equivalent array with the existing decorrelating methods, a VA-FBSS method for DOA estimation of coherent signals is proposed. The VA-FBSS avoids the optimal lag selection problem faced by the FBSS-CMUSIC, and obtains a better performance in accuracy. Besides, when compared with the CFBLP, remarkable improvements in performance can be achieved on the condition of low SNR and relatively closer impinging signals. Considering the VA-FBSS is a 1-D method, the virtual array-based spatial smoothing technique is also extended to the 2-D case.Based on the existing cyclic autocorrelation-based method, a cyclic crosscorrelation-based signal processing model along with the corresponding 2-D DOA estimation method for cyclostationary signals is induced. This method improves the accuracy of the DOA estimation.更多还原