Page 41 - 《应用声学》2024年第6期

第 43 卷 第 6 期 王晓楠等： 多项式结构恒定束宽波束形成器的高效稀疏化设计 1217

现有稀疏算法相比，所提稀疏算法在满足旁瓣响应 [6] Barfuss H, Bachmann M, Buerger M, et al. Design of ro-
约束要求的同时，表现出了更佳的期望方向信号无 bust two-dimensional polynomial beamformers as a con-
vex optimization problem with application to robot audi-
失真特性。
tion[C]//2017 IEEE Workshop on Applications of Signal
Processing to Audio and Acoustics. IEEE, 2017: 106–110.
6 结论 [7] 薛晗, 陈华伟, 张展, 等. 多项式结构宽带波束形成器的性能
分析及其改进 [J]. 声学学报, 2022, 47(3): 293–308.
本文提出了一种多项式结构恒定束宽波束形 Xue Han, Chen Huawei, Zhang Zhan, et al. Performance
analysis and improvement of polynomial broadband
成器的高效稀疏化设计方法，在现有稀疏算法的基
beamformers[J]. Acta Acustica, 2022, 47(3): 293–308.
础上，通过引入求解自由度更高的 MSRV 设计，改 [8] Wang T, Chen H. Robust design of Farrow-structure-
善了多项式结构波束形成器的恒定束宽性能，同时 based steerable broadband beamformers with sparse tap
weights via convex optimization[J]. EURASIP Journal on
利用迭代加权 ℓ 1 范数进行抽头稀疏优化，并采用
Audio, Speech, and Music Processing, 2015, 2015(1): 14.
ADMM 求解算法进一步提高了优化的效率。研究 [9] Frank A, Ben-Kish A, Cohen I. Constant-beamwidth
结果表明，相较于现有稀疏算法，所提稀疏算法在 linearly constrained minimum variance beam-
former[C]//2022 30th European Signal Processing Con-
不同指向角度下的主瓣频率不变特性更优，同时也
ference. IEEE, 2022: 50–54.
保持了良好的设计鲁棒性。另外，将具有更高求解 [10] 鄢社锋, 王文侠. 交替迭代多约束波束优化设计 [J]. 声学学
自由度的 MSRV设计与抽头稀疏优化问题结合，有 报, 2021, 46(6): 896–904.
Yan Shefeng, Wang Wenxia. Design of multiply con-
利于挖掘更多潜在的稀疏抽头。对比现有的利用经
strained beamformer via alternating direction method of
典 CVX 工具箱进行优化的稀疏设计，本文提出的 multipliers[J]. Acta Acustica, 2021, 46(6): 896–904.
ADMM 稀疏设计方法取得了更高的稀疏度，且优 [11] Xue H, Chen H, Wang X. Design of robust polynomial
beamformers using worst case performance optimization
化求解的耗时可以降低两个数量级。
via alternating direction method of multipliers[J]. IEEE
Sensors Journal, 2023, 23(7): 7690–7704.
[12] Wang W, Yan S, Mao L. Time-domain frequency-invariant
参 考 文 献
beampattern synthesis via alternating direction method
of multipliers[J]. The Journal of the Acoustical Society of
[1] Son P L. Irregular microphone array design for broadband America, 2020, 147(5): 3372–3375.
beamforming[J]. Signal Processing, 2022, 193: 108431. [13] Liu Y, Zhang L, Ye L, et al. Synthesis of sparse arrays
[2] 李涛, 蒋小勇, 周胜增. 基于宽带稳健 STMV 波束形成的相关 with frequency-invariant-focused beam patterns under ac-
检测方法 [J]. 声学技术, 2019, 38(5): 600–603. curate sidelobe control by iterative second-order cone pro-
Li Tao, Jiang Xiaoyong, Zhou Shengzeng. Correlation de- gramming[J]. IEEE Transactions on Antennas and Prop-
tection based on wideband robust STMV beamforming[J]. agation, 2015, 63(12): 5826–5832.
Technical Acoustics, 2019, 38(5): 600–603. [14] Yang J, Lin J, Shi Q, et al. An ADMM-based approach to
[3] Wang X, Cohen I, Chen J, et al. On robust and high robust array pattern synthesis[J]. IEEE Signal Processing
directive beamforming with small-spacing microphone ar- Letters, 2019, 26(6): 898–902.
rays for scattered sources[J]. IEEE/ACM Transactions on [15] Horn R A, Johnson C R. Matrix analysis[M]. Cambridge:
Audio, Speech, and Language Processing, 2019, 27(4): Cambridge University Press, 2012.
842–852. [16] Oliveira D E, Wolkowicz H, Xu Y. ADMM for the SDP re-
[4] 任维怡, 陈华伟, 鲍彧. 稳健宽带波束形成器设计的低阶统计 laxation of the QAP[J]. Mathematical Programming Com-
量法 [J]. 应用声学, 2015, 34(5): 413–424. putation, 2018, 10(4): 631–658.
Ren Weiyi, Chen Huawei, Bao Yu. Design of robust [17] Bala J, Lakhwani K. Single image desmogging using
broadband beamformers based on lower-order statistics oblique gradient profile prior and variational minimiza-
method[J]. Journal of Applied Acoustics, 2015, 34(5): tion[J]. Multidimensional Systems and Signal Processing,
413–424. 2020, 31(4): 1259–1275.
[5] Bao Y, Chen H. A chance-constrained programming ap- [18] Chen H, Ser W, Zhou J. Robust nearfield wideband beam-
proach to the design of robust broadband beamform- former design using worst case mean performance opti-
ers with microphone mismatches[J]. IEEE/ACM Transac- mization with passband response variance constraint[J].
tions on Audio, Speech, and Language Processing, 2018, IEEE Transactions on Audio, Speech, and Language Pro-
26(12): 2475–2488. cessing, 2012, 20(5): 1565–1572.