Abstract: To increase signal number distinguished by traditional MUSIC algorithm, this paper proposes a MUSIC algorithm based on three-dimensional smoothing by a coprime pair of two-dimensional sparse electromagnetic vector arrays. The coprime relationship between two arrays is used in the algorithm to generate a coprime co-array with more degrees of freedom. And the rank of its autocorrelation matrix is restored by three-dimensional (two-dimensional spatial domain and polarization domain) smoothing algorithm in order to apply to the traditional MUSIC algorithm. A major advantage of this method is that the freedom of the original array is also systematically increased by even using second-order statistics. Computer simulation results show that the proposed algorithm can estimate the number of signals more than the number of physical array elements and have high resolution.