【作者】 张雪；

【导师】 刘帅；

【作者基本信息】 哈尔滨工业大学 ， 电子与通信工程（专业学位）， 2020， 硕士

【摘要】 波束形成技术作为阵列信号处理的重要分支,通过对权矢量的优化,实现期望信号增强并抑制干扰和噪声的目的,在雷达、声呐、通信等领域得到广泛应用。传统波束形成算法在理想条件下具有最大的输出信干噪比,但在工程应用中,训练数据中的期望信号成分、导向矢量失配、幅相误差等非理想因素的存在,使得算法性能严重下降,甚至失效。针对此问题,本文利用子空间投影和独立分量分析技术,研究基于矩阵重构的鲁棒波束形成(RAB)算法,为RAB算法工程化的应用提供理论支撑。本文的具体研究内容如下:首先,对干扰噪声协方差矩阵(IPNCM)重构的有效性进行深入研究,矩阵重构有效避免了由期望信号引起的性能下降,但在低信噪比时性能提升有限,且增加了计算复杂度。因此,本文提出了改善因子的概念,给出矩阵重构类算法有效性和必要性的信噪比门限参考值,为RAB算法工程应用及后续研究内容奠定理论基础。其次,基于子空间投影思想提高算法鲁棒性,提出了特征空间矩阵重构RAB算法。该方法利用Capon谱估计器在干扰源的离散角扇区上积分重构IPNCM,再将权矢量向信号子空间投影来修正期望信号失配误差,有效地改善了算法鲁棒性。同时针对该算法高复杂度积分运算问题,提出基于GaussLegendre积分的矩阵重构RAB算法。该算法通过对三阶Legendre正交多项式的零点直接求和,实现了免积分运算条件下的IPNCM重构,有效降低了矩阵重构类RAB算法的计算复杂度,仿真结果表明该算法在性能和计算效率上都具有明显优势。最后,基于独立分量分析思想提出了基变换矩阵重构超RAB算法。该算法通过独立分量分析技术得到角度相关基,并利用基变换在角度相关基中去除期望信号,直接从信号子空间中构造IPNCM,实现了对阵列误差的超鲁棒性。在此基础上,深入研究了独立分量分析的幅度和初始相位不确定性对算法的影响,提出自相关矩阵重构超RAB算法和线性变换矩阵重构超RAB算法。仿真结果表明,基于独立分量分析思想的RAB算法对阵列误差具有超鲁棒性。更多还原

【Abstract】 Beamforming technique is a great significance field of array signal processing,which can enhance the desired signal and suppress interference and noise by adjusting the weight vector.Therefore,it is widely used in radar,sonar and other fields.The traditional beamformer has the largest output signal-tointerference-plus-noise ratio under ideal conditions.However,in engineering applications,the existence of non-ideal factors such as desired signal components in the training data,steering vector mismatch,amplitude and phase errors make algorithm performance seriously degrade,even fail.To address the problem,the feature subspace and independent component analysis techniques are used to study the robust adaptive beamforming(RAB)algorithm based on matrix reconstruction,which provides theoretical support for the engineering applications of RAB algorithm.The main research contents of this thesis are as follows:Firstly,the effectiveness of interference-plus-noise covariance matrix(IPNCM)reconstruction is studied.Matrix reconstruction can effectively avoid the performance degradation caused by desired signal components.However,the performance improvement is limited and the computational complexity is increased when the input signal-to-noise-ratio is small.Therefore,the concept of improve factor is proposed.Then the signal-to-noise-ratio threshold reference value is obtained to evaluate the effectiveness and necessity of reconstructionbased RAB algorithms,which lay the theoretical foundation for later research.Secondly,the feature subspace technique is utilized to improve robustness.The eigenspace-based matrix reconstruction RAB algorithm is proposed.The proposed method performs the reconstruction-based IPNCM with respect to the Capon spectral estimator integrated over discrete angular sectors associated with the interferences.Subsequently,an improved version of steering vector mismatch correction method is introduced,which utilized the projection technique of weight vector to signal-plus-interference subspace.The proposed method effectively improves algorithm robustness,but requires the high-complexity integration operation.To address the problem,the matrix reconstruction RAB algorithm based on Gauss-Legendre quadrature is proposed.The IPNCM can be efficiently obtained by simple summation with respect to the zeros of the three-order Legendre orthogonal polynomial without integral,which effectively reduces the computational complexity of the reconstruction-based RAB algorithms.The simulation results show that the new algorithm has significant advantages in performance and computational efficiency.Finally,using independent component analysis technique,the basis transformation matrix reconstruction super RAB algorithm is proposed.We obtain the angle related bases by independent component analysis technique.Consequently,we construct the IPNCM directly from the signal-plus-interference subspace by eliminating the component of the desired signal from the angle related bases,which achieves super robustness to array errors.On this basis,the influence of independent component analysis amplitude and initial phase uncertainty on the algorithm is deeply studied.We propose the autocorrelation matrix reconstruction super RAB algorithm and the linear transformation matrix reconstruction super RAB algorithm.The simulation results show that the RAB algorithms based on independent component analysis achieve super robustness to array errors.更多还原