麦克风阵列鲁棒频率不变波束形成算法
Robust Frequency-Invariant Beamforming Algorithm for Microphone Array
张正文 1张振平 1廖桂生 2巩朋成3
作者信息
- 1. 湖北工业大学电气与电子工程学院,湖北 武汉 430068
- 2. 西安电子科技大学雷达信号处理国家重点实验室,陕西 西安 710071
- 3. 武汉工程大学计算机科学与工程学院,湖北 武汉 430205
- 折叠
摘要
在实际应用中,频率不变波束形成器通常受到麦克风阵列失配误差的影响,因此提高频率不变波束形成器的鲁棒性具有重要意义.针对上述问题提出了一种约束优化模型,可以在保持频率不变波束形成的同时提高阵列的鲁棒性.首先设计目标波束图,考虑到差分麦克风阵列本身具有频率不变的波束图,选用传统二阶超心型差分麦克风波束图做为目标波束图.上述模型以麦克风阵列权矢量的二范数作为目标函数来最大化鲁棒性,在无失真约束,目标波束主瓣逼近约束以及旁瓣增益精准控制约束下实现频率不变.然后在交替方向乘子法算法框架下,将优化问题分解为多个优化子问题求解,然后对每个优化子问题分别求解,通过仿真验证了在交替方向乘子法算法下上述模型的可行性与有效性,最终达到了麦克风阵列鲁棒频率不变波束响应的效果.
Abstract
In practical applications,frequency invariant beamformer is often affected by microphone array mismatch errors,so improving the robustness of frequency invariant beamformer is of great significance.Aimed at a-chieving robust FI beamforming for the microphone arrays,this paper constructed a corresponding mathematical opti-mization model,which can improve the robustness of the array with FI beampattern.First,due to the property of fre-quency invariance for differential beamforming,the hypercardioid of conventional second-order differential microphone arrays(DMAs)was selected as the reference beampattern;The binomial of the microphone array weight vector was taken as the objective function to maximize the robustness of array;Under the constraint of signal without distortion,the beampattern was designed to match the desired beampattern in mainlobe region and controled below the predefined level in sidelobe region.Subsequently,in the framework of Alternating Direction Method of Multipliers(ADMM)algorithm,this optimization problem was decomposed into multiple optimization subproblems,and then each optimization subproblem was solved separately;Finally,the feasibility and effectiveness of the model under the ADMM algorithm were verified by simulation experiments,and the effect of robust frequency invariant beam response of the microphone array was achieved.
关键词
麦克风阵列/频率不变/波束形成/交替方向乘子法Key words
Microphone array/FI/Beamforming/ADMM引用本文复制引用
基金项目
国家自然科学基金(61601178)
湖北省自然科学基金(2018CFB545)
出版年
2024
NETL
NSTL
万方数据