中文摘要：

MUSIC方法是阵列高分辨测向子空间方法中的经典算法，该算法的一个缺点就是需要进行源个数估计，源个数估计不正确将会导致虚假峰和漏峰。提出一种不需要源个数估计的MUSIC方法，该方法首先将快拍分组，对每一组的自相关进行特征值分解，以最小特征值对应的特征向量作为噪声向量，然后将搜索向量在每一组噪声向量上的投影构成新的矩阵，DOA对应的新矩阵将缺秩，根据新矩阵的缺秩程度来估计DOA。仿真结果证明了该方法的有效性。

英文摘要：

MUSIC is a classic algorithm in the subspace method of direction finding. One disadvantage of this algorithm is the need of sources number, A new MUSIC algorithm without sources number estimation is brought up, The new method first groups the sampling data,then a noise vector corresponding to the smallest eigenvalue is calculated in every group, The projection of the searching vector onto the noise vector is calculated and forms a new matrix. If the search angle equals DOA,the new matrix would become rand deficient. The smallest singular value is used to detect to which degree the matrix is rand deficient, Simulation results demonstrate the effectiveness of the method.