为了改善传统的单一识别网络难以充分考虑水下声音样本各方面特征的缺陷，本文利用联合一维卷积神经网络与长短期记忆网络2种网络串行的方式，构建一种新的网络框架，首次将联合网络运用到水声目标识别中。其次，用船舶音频数据作为数据集输入网络，对网络性能进行评价，进行识别结果的可视化分析。通过结果分析得出，该网络能够实现对水声目标的识别分类，与单一神经网络相比，联合网络的识别精度更高，正确率等相关指标均优于单一识别网络，为水声目标识别领域的深度学习发展提供了新的参考方向。
In order to improve the defect that the traditional single recognition network is difficult to fully consider all features of underwater sound samples, this paper uses the joint one-dimensional convolutional neural network and long short-term memory network to construct a new network framework, and applies the joint network to underwater acoustic target recognition for the first time. Secondly, the ship audio data is used as the data set to input the network, and evaluate the network performance, and the visual analysis of the identification results is carried out. The interpretation of result shows that this method can realize the recognition and classification of underwater acoustic target. Compared with the single neural network, the recognition accuracy of the joint network is higher, and the relevant indexes such as the accuracy are better than that of the single neural network, which provides a new reference direction for the application of deep learning in the field of underwater acoustic target recognition.
2022,44(1): 136-141 收稿日期：2021-03-22
DOI：10.3404/j.issn.1672-7649.2022.01.026
分类号：TB566;
基金项目：中北大学2019年校科研基金资助项目(XJJ201927)
作者简介：任晨曦(1997-)，女，硕士研究生，研究方向为水声信号处理
参考文献：
[1] 章业成. 水声目标识别技术的现状与发展[J]. 电子技术与软件工程, 2019(18): 97−98.
KANG Y, WANG Y, WANG Y, et al. A new method for detecting underwater acoustic targets [J]. IEEE Transactions on Geoscience and Remote Sensing, 2015, 18 (1): 69−72.
[2] 王培兵, 彭圆. 深度学习在水声目标识别中的应用研究[J]. 数字海洋与水下攻防, 2020, 3(1): 11–17
WANG Peibing, PENG Yuan. Application of deep learning in underwater acoustic target recognition[J]. Digital Oceans and Underwater Offense and Defense, 2020, 3(1): 11–17
[3] HU Gang, WANG Kejun, LIU Liangliang. Underwater acoustic target recognition based on depthwise separable convolution neural networks[J]. Sensors, 2021, 21(4).
[4] 王小宇, 李凡, 曹琳, 等. 改进的卷积神经网络实现端到端的水下目标自动识别[J]. 信号处理, 2020, 36(6): 958–965
[5] 张少康, 王超, 孙芹东. 基于多类别特征融合的水声目标噪声识别分类技术[J]. 西北工业大学学报, 2020, 38(2): 366–376
ZHANG Shaokang, WANG Chao, SUN Qindong. Recognition and classification of underwater acoustic target noise based on multi-category feature fusion[J]. Journal of Northwestern Polytechnical University, 2020, 38(2): 366–376
[6] CHENG Wensheng, YANG Wen, et al. Context aggregation network for semantic labeling in aerial images[J]. Remote Sensing, 2019, 11(10): 1158
[7] 王升贵, 胡桥, 陈迎亮, 等. 基于深度学习的水下目标识别方法研究[J]. 舰船科学技术, 2020, 42(23): 141–145
WANG Shenggui, HU Qiao, CHEN Yingliang, et al. Research on underwater target recognition method based on deep learning[J]. Ship Science and Technology, 2020, 42(23): 141–145
[8] 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
ZHANG Shaokang, WANG Chao, TIAN Deyan, et al. Underwater target noise intelligent recognition method based on short and long time memory network[J]. Ship Science and Technology, 2019, 41(23): 181–185
[9] 卢安安. 基于深度学习方法的水下声音目标识别研究[D]. 哈尔滨: 哈尔滨工程大学, 2017.
[10] Santos-Domínguez D, Torres-Guijarro S, Cardenal-López A, et al. Shipsear: An underwater vessel noise database[J]. Applied Acoustics, 2016, 113: 64–69.
[11] BOGHDADY N El, LANGNER F, GAUDRAIN E, et al. Effect of spectral contrast enhancement on speech-on-speech intelligibility and voice cue sensitivity in cochlear implant users. [J]. Ear and Hearing, 2020.
[12] 杨路飞, 章新华, 吴秉坤, 等. 基于MFCC特征的被动水声目标深度学习分类方法[J]. 舰船科学技术, 2020, 42(19): 129−133.
YANG Lufei, ZHANG Xinhua, WU Bingkun, et al. A deep learning classification method for passive underwater acoustic target based on mfcc feature [J]. Ship Science and Technology, 202, 42(19): 129-133.
[13] 李阳. 面向声音事件的特征提取与识别方法研究[D]. 成都: 电子科技大学, 2020.
[14] JIANG Dan-Ning, LIE Lu, et. al. Music type classification by spectral contrast feature[C]//Proceedings of 2002 International Conference on Multimedia and Expo, Lausanne, 2002: 113-116.
[15] HOOK J. Generic sequences and the generic tonnetz[J]. Journal of Music Theory, 2020, 64(1).
