交叉验证多路径匹配追踪水声矢量阵稀疏方位估计

水下运动目标的高分辨DOA估计和目标的左右舷分辨问题一直是水声阵列信号处理中的一个核心问题。矢量阵相比于声压阵具有天然的左右舷分辨能力和更高的处理增益,近年来得到了广泛关注。Capon等一些传统高分辨处理方法存在不能解相干源、需要多快拍处理以及对阵列流行误差敏感等多种问题。针对水声阵列信号处理领域面临的以上问题,利用声呐工作场景中空间目标的稀疏性,本文提出了一种基于交叉验证技术的多路径匹配追踪（Multiplepath Matching Pursuit with Cross Validation,CV-MMP）声矢量阵稀疏DOA估计算法。该算法采用交叉验证技术可以在未知场景中目标个数的条件下实现稀疏DOA的估计,相比于常规的声矢量阵Capon算法而言,可以在小快拍数甚至单快拍数条件下实现多目标的稀疏DOA估计以及高分辨能力。仿真和海试试验数据处理验证了提出的算法的有效性。 

Multipath matching pursuit using a cross-validation technique for sparse direction-of-arrival estimation with an acoustic vector array

High-resolution Direction-of-Arrival（DOA） estimations and the starboard ambiguity of moving underwater targets have always been key issues in underwater acoustic array signal processing.Compared with sound pressure arrays,vector arrays have natural advantages with respect to solving the starboard ambiguity problem and obtaining higher processing gains.Traditional high-resolution DOA estimation methods such as Capon have disadvantages such as being unable to resolve coherent sources,requiring multiple snapshot processing,and being sensitive to array manifold errors.High-resolution DOA estimation and the starboard ambiguity of moving underwater targets have always been challenging research topics.On one hand,maneuvering underwater targets reduce the coherence time of the received signals,which ultimately leads to poor performance when using high-resolution DOA estimation technologies based on the covariance matrix of the received signal.On the other hand,traditional DOA estimation technologies based on sound pressure arrays have the problem of port and starboard ambiguity,which can be solved by maneuvering the sonar platform.However,maneuvering the sonar platform can impair the coherence of the received signal,on which some algorithms rely.This approach greatly limits the combat effectiveness and performance of the platform.Given the aforementioned problems and taking advantage of the target sparsity,a cross-validation multipath matching pursuit technique based on the sparse DOA estimation of an acoustic vector array is proposed in this article for sonar observations.The proposed algorithm uses cross-validation technology to achieve a sparse DOA estimation with an unknown number of targets in a sonar observation scene.Compared with the conventional acoustic vector array-based Capon algorithm,the proposed algorithm can achieve a sparse DOA estimation and high-resolution capability with small numbers of snapshots or even single snapshots.The effectiveness of the proposed algorithm is verified via simulations and sea trial data processing.