Page 173 - 《应用声学》2025年第2期

第 44 卷 第 2 期 明超等： 结合有限新息率重构的水声宽带信号方位估计方法 433

[4] Gerstoft P, Mecklenbräuker C F, Xenaki A, et al. Multi-
10 1
snapshot sparse Bayesian learning for DOA[J]. IEEE Sig-
nal Processing Letters, 2016, 23(10): 1469–1473.
CPUᤂᛡ௑ᫎ/s ISSM ing[J]. The Journal of the Acoustical Society of America,
[5] Xenaki A, Gerstoft P. Grid-free compressive beamform-
0
10
RSS
2015, 137(4): 1923–1935.
MD-SBL
[6] Vetterli M, Marziliano P, Blu T. Sampling signals with
BF-FRI
-1
10
finite rate of innovation[J]. IEEE transactions on Signal
Processing, 2002, 50(6): 1417–1428.
[7] Blu T, Dragotti P L, Vetterli M, et al. Sparse sampling of
10 -2
0 5 10 15 signal innovations[J]. IEEE Signal Processing Magazine,
௑ᫎ/min
2008, 25(2): 31–40.
[8] Pan H, Blu T, Vetterli M. Towards generalized FRI
图 12 实验数据中各算法的 CPU 运行时间对比
sampling with an application to source resolution in ra-
Fig. 12 Comparison of CPU runtime of each al- dioastronomy[J]. IEEE Transactions on Signal Processing,
gorithm in the experimental data 2016, 65(4): 821–835.
[9] Pan Y, Luo G Q, Jin H, et al. DOA estimation with planar
5 结论 array via spatial finite rate of innovation reconstruction[J].
Signal Processing, 2018, 153: 47–57.
本文引入了一种基于多项式比值模型的 FRI [10] Chen T, Shi L, Yu Y. Gridless DOA estimation with fi-
nite rate of innovation reconstruction based on symmetric
重构方法。为解决该 FRI 重构方法在宽带信号下
Toeplitz covariance matrix[J]. EURASIP Journal on Ad-
无法应用的问题，结合 CSSM 方法中的宽带聚焦思 vances in Signal Processing, 2020, 2020(1): 1–16.
想，提出了一种适用于宽带信号的 DOA 估计方法 [11] Gilliam C, Blu T. Fitting instead of annihilation: Im-
——BF-FRI 方法。数值仿真验证了该方法在宽带 proved recovery of noisy FRI signals[C]. 2014 IEEE In-
ternational Conference on Acoustics, Speech and Signal
信号下的可行性。与基于CS理论的DOA估计方法 Processing (ICASSP). IEEE, 2014: 51–55.
相比，该方法具有较低的运算复杂度，在实时处理上 [12] Li Y, Guo R, Blu T, et al. Generic FRI-based DOA es-
具有显著优势，同时避免了因网格失配导致的估计 timation: A model-fitting method[J]. IEEE Transactions
on Signal Processing, 2021, 69: 4102–4115.
误差。数值仿真表明，该方法目前在低 SNR 环境下
[13] Hoffmann F M, Nelson P A, Fazi F M. DOA estimation
性能表现相较基于 CS 理论的 DOA 估计方法和经 performance with circular arrays in sound fields with finite
典的 ISSM 方法略有下降。海试数据处理结果验证 rate of innovation[J]. IEEE/ACM Transactions on Audio,
Speech, and Language Processing, 2019, 28: 171–184.
了该方法在宽带信号DOA估计中的有效性。
[14] Cadzow J A. Signal enhancement-a composite property
本文为在宽带情况下采用 FRI 重构方法进行 mapping algorithm[J]. IEEE Transactions on Acoustics,
DOA 估计这一问题引进了新的思路，然而，由于该 Speech, and Signal Processing, 1988, 36(1): 49–62.
方法目前在低SNR环境下性能表现尚有不足，需进 [15] Li Y, Guo R, Blu T, et al. Robust sparse reconstruction of
attenuated acoustic field with unknown range of source[J].
一步研究和改进以提升其在低SNR环境下的表现。
The Journal of the Acoustical Society of America, 2022,
152(6): 3523–3534.
致谢 感谢参与该水声实验的所有工作人员，是
[16] Guo R, Li Y, Blu T, et al. Vector-FRI recovery of
你们的奉献和专业精神为本文研究打下了坚实的 multi-sensor measurements[J]. IEEE Transactions on Sig-
基础。 nal Processing, 2022, 70: 4369–4380.
[17] 李永飞, 郭瑞明, 赵航芳. 浅海内波环境下声场干涉条纹的稀
参 考 文 献 疏重建 [J]. 物理学报, 2023, 72(7): 241–251.
Li Yongfei, Guo Ruiming, Zhao Hangfang. Sparse recon-
[1] Hung H, Kaveh M. Focusing matrices for coherent signal-
struction of acoustic interference fringes in shallow water
subspace processing[J]. IEEE Transactions on Acoustics,
and internal wave environment[J]. Acta Physica Sinica,
Speech, and Signal Processing, 1988, 36(8): 1272–1281.
2023, 72(7): 241–251.
[2] Valaee S, Kabal P. Wideband array processing using a
[18] Cao J, Liu Z, Xu Y. New algorithm requiring no prepro-
two-sided correlation transformation[J]. IEEE Transac-
cessing for wideband DOA estimation[C]. 2008 9th Inter-
tions on Signal processing, 1995, 43(1): 160–172.
national Conference on Signal Processing. IEEE, 2008:
[3] Li C, Liang G, Qiu L, et al. An efficient sparse method
394–397.
for direction-of-arrival estimation in the presence of strong
[19] Nannuru S, Gemba K L, Gerstoft P, et al. Sparse Bayesian
interference[J]. The Journal of the Acoustical Society of
learning with multiple dictionaries[J]. Signal Processing,
America, 2023, 153(2): 1257–1271.
2019, 159: 159–170.