水下目标探测是水下无人航行器的重要应用领域，而被动声呐则是水下无人航行器的最重要探测载荷之一。对于被动声呐而言，从接收到的微弱目标辐射噪声中检测出稳定的线谱分量是一个十分重要的研究内容。通常被动声呐在检测线谱之前会对接收信号进行线谱分量的增强，往往使用自适应线谱增强器（ALE）来实现线谱分量的的增强。但是，ALE算法对系统的输入信噪比（SNR）有一定要求。当输入信噪比太低时，ALE将无法正常工作。为了克服ALE在输入SNR方面的限制，本文提出一种基于无监督深度学习的线谱分量增强算法。仿真表明，当输入信噪比为−30 dB时，本文算法仍可实现20 dB的信噪比增益，ALE算法则无法对信号进行增强。利用海试数据进行验证，证明本文所提出的无监督深度学习线谱增强器算法在低信噪比环境下的优势。
Detecting underwater targets is an important application of unmanned underwater vehicles. A passive sonar is one of key payloads of UUVs. Detection of acoustic tonals radiated from surfaces and underwater vehicles is important for passive sonars. Enhancing the tonals is usually necessary in passive sonars before detection. Conventionally, passive sonars use an adaptive line enhancer (ALE) to enhance the tonals. However, ALEs require their input signal-to-noise ratio (SNR) should higher than a threshold. If the input SNR is too low, ALEs will fail. To overcome the limit of ALEs, we proposes to use unsupervised deep learning to enhance the tonals for passive sonars. The proposed line enhancing algorithm is based on an deep neural network which is popular in unsupervised deep learning. Simulation shows that at the input SNR of −30 dB, the proposed line enhancing algorithm still achieves an SNR gain of 20 dB, but the reference ALE cannot work. The experiment also demonstrates the priority of the proposed method.
2020,42(12): 117-120 收稿日期：2020-08-05
DOI：10.3404/j.issn.1672-7649.2020.12.023
分类号：TN911.4
基金项目：中国科学院百人计划资助项目
作者简介：鞠东豪(1993-),男,博士研究生,研究方向为水下无人平台探测与识别技术
参考文献：
[1] 潘光, 宋保维, 黄桥高, 等. 水下无人系统发展现状及其关键技术[J]. 水下无人系统学报, 2017, 25(2): 44-51
[2] 钟宏伟, 李国良, 宋林桦, 等. 国外大型无人水下航行器发展综述[J]. 水下无人系统学报, 2018, 26(4): 273-282
[3] 李启虎. 数字式声纳设计原理[M]. 合肥: 安徽教育出版社, 1985.
[4] 李启虎. 用于拖曳式线列阵的一种新的线谱增强系统[J]. 声学学报, 1988(3): 9-15
[5] RICKARD J. T., ZEIDLER J. R., “Second-order output statistics of the adaptive line enhancer,” IEEE Trans. Acoust. Speech Signal Process. 27, 31−39 (1979).
[6] NEHORAI A., MALAH D D.. “On the stability and performance of the adaptive line enhancer,” in: Proceedings of IEEE ICASSP, Denver, USA (April 9-11, 1980), pp. 478−481.
[7] BIANCO M. J., GERSTOFT P., TRAER J., et al. “Machine learning in acoustics: Theory and application,” J. Acoust. Soc. Am. 146, 3590−3628 (2019).
[8] HINTON G E, SALAKHUTDINOV R R. Reducing the dimensionality of data with neural networks[J]. science, 2006, 313(5786): 504-507
[9] VINCENT P, LAROCHELLE H, BENGIO Y, et al. Extracting and composing robust features with denoising autoencoders[C]//Proceedings of the 25th international conference on Machine learning. 2008: 1096−1103.
