Page 35 - 《应用声学》2023年第3期

第 42 卷 第 3 期 王珍珠等： 卷积神经网络主动目标方位估计 473

[5] Niu H, Reeves E, Gerstoft P. Source localization in an
3 结论 ocean waveguide using supervised machine learning[J].
The Journal of the Acoustical Society of America, 2017,
本文提出了一种基于 CNN 模型的主动目标方 142(3): 1176–1188.
[6] Niu H, Ozanich E, Gerstoft P. Ship localization in Santa
位估计方法，将接收阵列信号 CBF 后的结果输入
Barbara Channel using machine learning classifiers[J]. The
网络，经过仿真实验证实该方法在不同 SNR 下具 Journal of the Acoustical Society of America, 2017, 142(5):
有较高的鲁棒性，可以有效估计目标波达方向。在 EL455–EL460.
实际海洋环境下进行仿真，该方法与MVDR、CBF、 [7] Wang Y, Peng H. Underwater acoustic source localization
using generalized regression neural network[J]. The Jour-
MSNR 波束形成方法的比对结果证实，其具有更高
nal of the Acoustical Society of America, 2018, 143(4):
的估计精度。但本文提出的方法只是相比传统方法 2321–2331.
有较好的性能，通过模拟仿真对模型进行了验证，但 [8] Huang Z, Xu J, Gong Z, et al. Source localization using
实测的水声数据难以获取，模型依赖于训练数据，当 deep neural networks in a shallow water environment[J].
The Journal of the Acoustical Society of America, 2018,
训练环境与测试环境差别较大时，仍然存在环境模 143(5): 2922–2932.
型失配的问题。 [9] Elbir A M. Deep MUSIC: Multiple signal classification via
deep learning[J]. IEEE Sensors Letters, 2020, 4(4): 1–4.
[10] Yao Y, Lei H, He W. A-CRNN-based method for coherent
参 考 文 献 DOA estimation with unknown source number[J]. Sensors
(Basel, Switzerland), 2020, 20(8): 2296.
[11] Zhu W, Zhang M, Li P, et al. Two-dimensional DOA esti-
[1] 黄海宁, 李宇. 水声目标探测技术研究现状与展望 [J]. 中国科 mation via deep ensemble learning[J]. IEEE Access, 2020,
学院院刊, 2019, 34(3): 264–271. 8: 124544–124552.
Huang Haining, Li Yu. Underwater acoustic detection: [12] Liu Y, Chen H, Wang B, DOA estimation based on CNN
current status and future trends[J]. Bulletin of Chinese for underwater acoustic array[J]. Applied Acoustics, 2021,
Academy of Sciences, 2019, 34(3): 264–271. 172: 107594.
[2] Bucker H P. Use of calculated sound fields and matched- [13] 曹怀刚. 单矢量水听器波达方向估计的深度学习模型构建与
field detection to locate sound sources in shallow water[J]. 应用 [D]. 北京: 中国科学院大学, 2021.
The Journal of the Acoustical Society of America, 1976, [14] 姚琦海, 汪勇, 黎佳艺, 等. 基于广义回归神经网络的强干
59(2): 368–373. 扰下垂直阵目标距离估计方法 [J]. 应用声学, 2021, 40(5):
[3] Baggeroer A B. Matched field processing: source localiza- 723–730.
tion in correlated noise as an optimum parameter estima- Yao Qihai, Wang Yong, Li Jiayi, et al. Source range esti-
tion problem[J]. The Journal of the Acoustical Society of mation method of vertical array under strong interference
America, 1988, 83(2): 571–587. based on GRNN[J]. Journal of Applied Acoustics, 2021,
[4] Steinberg B Z, Beran M J, Chin S H, et al. A neural net- 40(5): 723–730.
work approach to source localization[J]. The Journal of the [15] Kingma D, Ba J. Adam: a method for stochastic opti-
Acoustical Society of America, 1991, 90(4): 2081–2090. mization[J]. Computer Science, 2014, arXiv: 1412.6980.