中文摘要：

基于反问题的正则化波束形成技术能以较高的计算效率得到稳健的声源识别结果。然而由于其正则化解中的正则化矩阵取决于低效的传统波束形成方法,使得基于反问题的正则化波束形成的声源识别结果精准度较低。为了在低信噪比环境下进一步提升其声源识别性能,基于Tikhonov正则化一般形式解提出一种双重迭代优化算法。该算法基于延时求和波束形成算法与互谱运算构造出新的正则化矩阵,并结合迭代方法对新正则化矩阵和波束输出进行优化,最终以较少的迭代步数经两次迭代运算有效提高了声源识别精度和稳定性。最后,通过数值仿真和实验算例,进一步验证了双重迭代优化算法的可行性和有效性。

英文摘要：

The inverse problem based regularized beamforming technique can lead to robust sound source location result with the high cal- culation efficiency. However, due to the regularization matrix of regularized solution which depends on the inefficient conventional beam- forming method, the sound source identification accuracy of the inverse problem based regularized beamforming technique is unsatisfacto- ry. To further enhance the sound source location performance of regularized beamforming in the low-SNR environment, a dually iterated regularized beamforming method based on the Tikhonov regularization is proposed. The modified algorithm first structures a new regulari- zation matrix based on the delay-and-sum beamforming and cross spectra, and iteration method is used to optimize the new regularization matrix and beamforming output. The numerical simulation and experimental example are implemented to verify the feasibility and validity of the modified algorithm proposed in this paper.