以机器学习为代表的智能技术迅猛发展，也为被动声呐目标识别提供了新的思路。利用机器学习算法挖掘水声目标信号深层特征，实现目标自动识别、辅助识别，成为被动声呐目标识别的新发展方向。本文针对水下噪声目标的信号特性，结合人耳在低信噪比、多目标环境下的优异识别性能，提取被动声呐目标经典听觉感知特征——梅尔倒谱（MFCC），并引入KNN、SVM、CNN和DBN四种机器学习算法对两类水声目标进行监督学习和识别分析。试验结果表明，监督学习方法应用于被动声呐目标识别具有可行性，且其中DBN方法对目标MFCC特征的识别性能最佳。
The intelligent target recognition techniques, especially the machine learning method, have been booming in recent years, which may provide many new solutions to passive sonar target recognition. Mining the intrinsic features of underwater acoustic target via machine learning algorithms, is leading an evolution in passive sonar target recognition technology. In this paper, the Mel Frequency Cepstrum Coefficient (MFCC) database, extracted from several underwater acoustic sections, has been established. And several machine learning methods, K-nearest Neighbor (KNN), Support Vector Machine (SVM), Convolutional Neural Network (CNN) and Deep Belief Network (DBN), have been applied in passive sonar target recognition. The results show that supervised learning methods have excellent performance in passive sonar target recognition. And the recognition accuracy of SVM, CNN, DBN are close while the DBN is the best.
2018,40(9): 116-121 收稿日期：2017-12-21
DOI：10.3404/j.issn.1672-7649.2018.09.022
分类号：TN929.3
基金项目：装备预研重点实验室基金资助项目（614221403031703）
作者简介：程锦盛(1991-),男,硕士研究生,研究方向为水声信号处理
参考文献：
[1] 王磊, 彭圆. 听觉外周计算模型在水中目标分类识别中的应用[J]. 电子学报, 2012, 40(1):199-203 WANG Lei, PENG Yuan. The application of computational auditory peripheral model in underwater target classification[J]. Acta Electronical Sinica, 2012, 40(1):199-203
[2] 李秀坤, 孟祥夏. 水下目标回波和混响在听觉感知特征空间的分类[J]. 哈尔滨工程大学学报, 2015, 36(9):1183-1187 LI Xiu-kun, MENG Xiang-xia. Classification of under-water target echoes and reverberation in auditory per-ception feature spaces[J]. Jornal of Harbin Engineeri-ng University, 2015, 36(9):1183-1187
[3] 常国勇, 袁富宇, 崔杰. 基于不同能量值特征优选的水声目标识别[J]. 指挥控制与仿真, 2016, 38(2):60-65 CHANG Guo-yong, YUAN Fu-yu, CUI Jie. Underwat-er acoustic target recognition based on feature select-ion of different energy feature[J]. Command Control and Simulation, 2016, 38(2):60-65
[4] 吴姚振. 水下目标辐射噪声生理感知特征的提取方法研究[D]. 西安:西北工业大学, 2014.
[5] 曾向阳. 智能水中目标识别[M]. 北京:国防工业出版社, 2016.
[6] 周志华, 机器学习[M]. 北京:清华大学出版社, 2016.
[7] HARRINGTON P. Machine learning in action[M]. Manning Publications Co. 2012.
[8] 戴冬, 卫娟. K-means与SVM结合的水下目标分类方法[J]. 舰船科学技术, 2015, 37(2):204-207 DAI Dong, WEI Juan. Research on target identificat-ion underwater method by K-means and SVM[J]. Ship Science and Technology, 2015, 37(2):204-207
[9] NEUMANN J, SCHNÖRR C, STEIDL G. Combined SVM based feature selection and classification[J]. Machine Le-arning, 2005, 61:129-150
[10] 王强, 曾向阳. 深度学习方法及其在水下目标识别中的应用[J]. 声学技术, 2015, 34(2):138-140 WANG Qiang, ZENG Xiang-yang. Deep learning methods and their applications in underwater targets rec-ognition[J]. Technical Acoustics, 2015, 34(2):138-140
[11] LECUN Y, BENGIO Y, HINTON G. Deep learning[J]. Nature, 2015, 521(7553):436-444
[12] 张大伟, 章新华. 基于听觉感知与卷积神经网络识别舰船目标[J]. 声学技术, 2015, 34(6):182-184 ZHANG Da-wei, ZHANG Xin-hua. Recognition of ships based on auditory sense and convolutional neural networks[J]. Technical Acoustics, 2015, 34(6):182-184
[13] 史鹤欢, 许悦雷. PCA预训练的卷积神经网络目标识别算法[J]. 西安电子科技大学学报, 2016, 43(3):161-166 SHI He-huan, XU Yue-lei. Convolutional neural networks recognition algorithm based on PCA[J]. Journal of Xidian University, 2016, 43(3):161-166
[14] 鲁铮. 基于T-RBM算法的DBN分类网络的研究[D]. 吉林:吉林大学, 2014.
[15] LI D, DONG Y. Deep learning methods and applicati-ons[M]. Foundations and Trends in Signal Processing, 2015.
[16] BENGIO Y. Learning deep architecture for AI[M]. Fo-undations and Trends in Machine Learning:2009.
[17] 焦李成, 赵进, 杨淑媛, 等. 深度学习优化与识别[M].北京:清华大学出版社, 2017.JIAO Licheng, ZHAO Jin, YANG Shuyuan, et al. D-eep learning, optimization and recognition[M]. Beijing:Tsinghua University Press, 2017.
[18] GOODFELLOW I, BENGIO Y, COURVILLE A. 深度学习[M]. 北京:人民邮电出版社, 2017.
