Page 149 - 《应用声学）》2023年第5期

第 42 卷 第 5 期 赵乾坤等： 基于时延神经网络模型的舰船辐射噪声目标识别 1041

஝᧚/Զ
[6] Kamal S, Mohammed S K, Pillai P, et al. Deep learning
600 architectures for underwater target recognition[C]. 2013
Х̵ 646 68 41 21 Ocean Electronics (SYMPOL), 2013.
500
[7] 王强, 曾向阳. 深度学习方法及其在水下目标识别中的应
用 [C]//中国声学学会水声学分会 2015 年学术会议论文集,
ࠇᓕ 47 425 19 23 400
ᄾࠄ 2015.
300 [8] Yue H, Zhang L, Wang D, et al. The classification of
᠎ᓕ 69 83 349 136 underwater acoustic targets based on deep learning meth-
200 ods[C]. In 2017 2nd International Conference on Control,
Automation and Artificial Intelligence (CAAI 2017), 2017.
࠵ດᓕ 41 24 64 473 100 [9] 张少康, 王超, 田德艳, 等. 长短时记忆网络水下目标噪声智
能识别方法 [J]. 舰船科学技术, 2019, 41(23): 181–185.
Х̵ ࠇᓕ ᠎ᓕ ࠵ດᓕ Zhang Shaokang, Wang Chao, Tian Deyan, et al. Under-
ᮕ฾
water target noise intelligent recognition method based on
图 10 分类混淆矩阵 short and long time memory network[J]. Ship Science and
Fig. 10 Confusion matrices for of targets Technology, 2019, 41(23): 181–185.
[10] Li C, Huang Z, Xu J, et al. Underwater target classifica-
3 结论 tion using deep learning[C]. OCEANS 2018 MTS/IEEE
Charleston, 2018.
[11] Li J, Wang B, Cui X, et al. Underwater acoustic target
本文以典型的船舶类水下辐射噪声信号为研
recognition based on attention residual network[J]. En-
究对象，以水声信号的分类识别为目的，研究了采 tropy, 2022, 24(11): 1657.
用一种基于注意力机制的 TDNN 网络模型在水声 [12] 徐承, 李勇, 张梦, 等. 基于特征融合和自注意力机制的水下
目标识别 [J]. 移动通信, 2022, 46(6): 91–98.
信号分类识别的应用能力。分别对ShipsEar开源数
Xu Cheng, Li Yong, Zhang Meng, et al. Underwater
据集和课题组自行采集的实验数据进行了实验，提 target recognition based on feature fusion and self atten-
取信号梅尔频谱作为输入特征，识别准确率分别达 tion mechanism[J]. Mobile Communications, 2022, 46(6):
91–98.
到 79.2% 和 73.9%，验证了实验模型在水声目标识
[13] 崔琳, 王芷悦. 基于 LFBank 与 FBank 混合特征的声纹识别
别问题上的有效性。下一步将验证多特征融合输入 研究 [J]. 计算机科学, 2022, 49(S2): 621–625.
是否会提高模型得识别准确率。 Cui Lin, Wang Zhiyue. Research on voiceprint recogni-
tion based on mixed features of LFBank and FBank[J].
Computer Science, 2022, 49 (S2): 621–625.
参 考 文 献 [14] Snyder D, Garcia-Romero D, Sell G, et al. X-
Vectors: robust DNN embeddings for speaker recog-
[1] 王培兵, 彭圆. 深度学习在水声目标识别中的应用研究 [J]. 数 nition[C]//ICASSP 2018-2018 IEEE International Con-
字海洋与水下攻防, 2020, 3(1): 11–17. ference on Acoustics, Speech and Signal Processing
Wang Peibing, Peng Yuan. Application of deep learning in (ICASSP). IEEE, 2018.
underwater acoustic target recognition[J]. Digital Oceans [15] Desplanques B, Thienpondt J, Demuynck K. ECAPA-
and Underwater Offense and Defense, 2020, 3(1): 11–17. TDNN: emphasized channel attention, propagation and
[2] Leal N, Leal E, Sanchez G. Marine vessel recognition by aggregation in TDNN based speaker verification[J]. arXiv
acoustic signature[J]. ARPN Journal of Engineering and Preprint, arXiv: 2005.07143, 2020.
Applied Sciences, 2015, 10(20): 9633–9639. [16] Okabe K, Koshinaka T, Shinoda K. Attentive statistics
[3] Wang W, Li S, Yang J, et al. Feature extraction pooling for deep speaker embedding[J]. arXiv Preprint,
of underwater target in auditory sensation area based arXiv: 1803.10963, 2018.
on MFCC[C]//2016 IEEE/OES China Ocean Acoustics [17] Deng J, Guo J, Zafeiriou S. ArcFace: additive angular
(COA). IEEE, 2016. margin loss for deep face recognition[J]. arXiv Preprint,
[4] Chen Y, Xu X. The research of underwater target recog- arXiv: 1801.07698, 2022.
nition method based on deep learning[C]. IEEE Interna- [18] 刘峰, 罗再磊, 沈同圣, 等. 时频谱图和数据增强的水声信号
tional Conference on Signal Processing, Communications 深度学习目标识别方法 [J]. 应用声学, 2021, 40(4): 518–524.
and Computing, 2017. Liu Feng, Luo Zailei, Shen Tongsheng, et al. Recogni-
[5] Cao X, Zhang X, Yu Y, et al. Deep learning-based recog- tion method of underwater acoustic signal depth learn-
nition of underwater target[C]. IEEE International Con- ing based on spectrogram map and data enhancement[J].
ference on Digital Signal Processing, 2016. Journal of Applied Acoustics, 2021, 40(4): 518–524.