中文摘要：

针对指数嵌入族（Exponentially Embedded Families,EEF）准则在快拍数小于阵元数情况下无法估计声源个数的问题,本文提出一种新的空间声源个数估计算法.首先通过球麦克风阵列采集空间声场高阶信息,建立球阵列信号模型,将声源个数估计扩展到三维空间.继而将观测信号空间分解为信号子空间和噪声子空间,利用最小均方差（Minimum Mean-Squared Error,MMSE）方法估计观测信号空间及噪声子空间的协方差矩阵,确保矩阵估计的一致性和准确性.在此基础上改进似然比函数,同时引入新的自由度计算,使得算法在快拍数小于阵元数的情况下能有效估计声源个数.仿真结果表明,在进行空间声源个数估计时,相对于EEF准则,新的算法不仅适用于快拍数小于阵元数情况,同时提高了估计准确率.

英文摘要：

The exponentially embedded families（ EEF） criterion fails to enumerate sources w hen the number of snapshots is smaller than that of array sensors. To solve this problem,a novel estimation algorithm is proposed based on the EEF criterion in this paper. First a spherical microphone array is used to sample high-order sound field information in 3D space and the array signal model is constructed to estimate the number of spatial acoustic sources. Then the observation space is divided into a signal subspace and a noise subspace. The covariance matrices of the observation space and the noise subspace are estimated by minimum mean-squared error（ M M SE） method. Based on the consistent and more accurate estimates,w e calculate a new likelihood ratio function and a parameter freedom to modify the conventional EEF criterion. The proposed method can enumerate sources effectively in the case of the deficient number of snapshots. Compared w ith the conventional EEF criterion,simulation results demonstrate that the proposed algorithm has better performance for source enumeration.