中文摘要：

研究了MIMO雷达在对称α稳定分布（SaS，Symmetrica—stable）冲击噪声背景下，基于子空间的多目标DOA估计问题，并分析了空间分集特性对DOA估计性能的改善。由于SaS噪声不存在二阶及以上矩，使得传统的基于二阶或高阶累计量的多目标DOA算法在SaS噪声中性能得到恶化。为此，首先在分析MIMO雷达接收数据FLOM（Fractional lower order moment）矩阵子空间的基础上，给出FLOM—MUSIC算法。考虑到FLOM—MUSIC算法需要冲击噪声特征指数的先验信息，为避免噪声特征指数估计，提出基于无穷范数对接收数据归一化处理的InfMUSIC（Infinity—normnor malization MUSIC）算法。理论分析表明，无穷范数归一化后的数据协方差矩阵有界，且能分解成噪声子空间和信号子空间。计算机仿真验证了上述两种算法的有效性。仿真结果还表明在冲击噪声背景下，MIMO雷达的空间分集特性也能改善DOA估计的精度。

英文摘要：

This paper investigates the subspace-based multi-targets DOA estimation for MIMO radar amid Symmetricot -stable （ Sots ） impulsive noise environments, and analyzes the improvement of DOA estimation via spatial diversity. The conventional subspace-based DOA estimation algorithms are inapplicable to impulsive noise because array covariance matrix, a second-order entry, may not be defined. To settle this problem, we firstly extend the FLOM-MUSIC （fractional lower order moment MUSIC） algorithm to MIMO radars. We analyze the structure of the FLOM matrix to find that the FLOM- MUSIC can work if the fractional lower order is less than the characteristic exponent of the noise. In practice, the application of FLOM-MUSIC requires estimation of the noise characteristic exponent, which may be computationally burdensome and possible erroneous. To deal with this problem, we then propose a infinity-norm normalization MUSIC algorithm. This new algorithm transforms impulsive noise into a normalized-noise with zero-mean and finite variance. Theoretical analysis shows that covariance matrix formed from the infinity-norm normalized data is finite, and the covariance matrix also preserves the zero cross-correlation between the signal-subspace and the noise-space. Computer simulations verify the effectiveness of the two proposed subspace-based algorithms. The results also show that spatial diversity can improve the precision of DOA estimation in impulsive noise.