声矢量阵降维平行因子模型及方位估计方法

摘要

针对原有的声矢量阵三阶 PARAFAC（平行因子）模型维数高、参数求解过程运算量大的缺点，建立了一种降维的 PARAFAC 模型。将声矢量阵看作空间共点的声压传感器子阵和振速传感器子阵，计算各子阵输出数据的自协方差，并构造了三阶张量，最后证明该张量满足三阶 PARAFAC 模型并利用交替迭代算法估计声源参数。仿真和实测数据表明：该方法可以用于多目标方位估计且估计精度优于超分辨率的 ESPRIT 算法。%As the existing PARAFAC(Parallel Factor)model of vector hydrophone array(VHA)has a large dimension and needs a large amount of complex computation during the parameters estimation,a PARAFAC model with a smaller dimension is established.First,a single VHA is regarded as several parallel sub-arrays:a pressure sensor sub-array and two/three velocity sensor sub-arrays.Second,au-to-covariance matrix of each sub-array is calculated to form a third order tensor.The tensor is verified so as to meet a third order PARAFAC model,and the unknown parameters are obtained from TALS operations.Simulations and experiments show that the dimension-descending PARAFAC model can be used to find the DOAs of multiple sound sources and that it is better in accuracy than the ESPRIT algorithm.