矢量阵改进MUSIC算法

Improved MUSIC Algorithm for Vector Sensor Arrays

摘要:

针对传统多重子空间分类（Multiple Signal Classification, MUSIC）算法在矢量水听器（Acoustic Vector Sensor, AVS）阵列方位估计（Direction-of-Arrival, DOA）中存在的性能退化问题，提出了一种改进的MUSIC算法（Modified MUSIC, MMUSIC）。传统矢量阵MUSIC由于声压、振速通道噪声统计特性的不一致，信号子空间被扩展，噪声子空间受到污染，破坏了预期的正交性，进而导致空间谱估计出现虚假伪峰，甚至分辨率下降。为此，本文基于对矢量阵接收模型的深入分析，揭示了声压与振速通道噪声功率差异而引入“虚源”的机理，重构了噪声子空间的划分准则，利用特征向量与导向矩阵的正交性构建新的谱函数，精确识别并克服虚源影响。仿真结果表明，所提MMUSIC算法在低信噪比和小角度间隔等恶劣条件下，仍能有效分辨相干信号，提升了矢量阵的DOA性能与稳健性。

Abstract:

This paper proposes a Modified MUSIC (MMUSIC) algorithm to address the performance degradation of the conventional Multiple Signal Classification (MUSIC) algorithm in direction-of-arrival (DOA) estimation with acoustic vector sensor (AVS) arrays. Conventional MUSIC relies heavily on the strict orthogonality between the signal and noise subspaces. However, in AVS arrays, the statistical inconsistency between pressure and particle velocity channel noise leads to an expansion of the signal subspace and contamination of the noise subspace, thereby violating the expected orthogonality. This results in spurious spectral peaks and even degraded resolution. To overcome this issue, a detailed analysis of the AVS reception model is conducted, which reveals the mechanism by which noise power imbalance between pressure and velocity channels introduces “virtual sources.” Based on this analysis, the proposed algorithm reconstructs the criterion for noise subspace partitioning and formulates a new spectrum function by exploiting the orthogonality between eigenvectors and the steering matrix, enabling accurate identification and elimination of virtual source effects. Simulation results demonstrate that the proposed MMUSIC algorithm can effectively resolve coherent signals under challenging conditions such as low signal-to-noise ratio (SNR) and small angular separations, thereby significantly enhancing the resolution and robustness of AVS-based DOA estimation.