针对水下被动声呐目标分类识别难的问题，根据张量结构具有鲁棒性的特点，提出一种基于小波变换的船舶辐射噪声3阶张量特征提取方法。将船舶辐射噪声进行分帧，对每一帧信号进行小波变换，再对小波变换系数进行MFCC特征提取，最后构建一个帧数 小波分解分量 MFCC特征维数的3阶特征张量。将整个船舶辐射噪声样本集划分为训练集和测试集，对训练集的3阶特征张量采用张量分解得到所有训练样本的核张量和因子矩阵，利用测试样本的3阶特征张量和训练样本的因子矩阵，得到测试样本的核张量，通过比较训练样本与测试样本核张量之间的范数大小，实现船舶辐射噪声的分类识别。实测样本识别结果表明，该方法具有较高的正确分类识别概率。
Aiming at the difficulty of underwater passive sonar target classification and identification, according to the robustness of tensor structure, a feature extraction method of the ship-radiated noise based on wavelet transform and the third-order tensor is proposed. Firstly, the ship radiated noise is divided into frames, then wavelet transform is carried out for each frame, and then the wavelet transform coefficient is extracted with MFCC feature, and finally a third-order feature tensor (frame number × wavelet decomposition × MFCC) is created. The whole ship-radiated noise sample sets is divided into the training and test sets, the core tensor and factor matrices of the training samples were obtained via tensor decomposition, the core tensor of test samples were obtained through using the test sample tensor and the training sample factor matrices. The classification and identification of ship-radiated noise was realized by comparing the core tensor of the training samples and test samples. The results of measured samples show that this method is useful on the classification and recognition.
2020,42(9): 171-175 收稿日期：2019-10-31
DOI：10.3404/j.issn.1672-7649.2020.09.033
分类号：TB566
基金项目：国家自然科学基金资助项目(61471378)
作者简介：郭德鑫(1995-),男,硕士研究生,主要从事水声目标识别研究
参考文献：
[1] KOLDA T G, BADER B W. Tensor decompositions and applications[J]. SIAM Re-view, 2009, 51(3): 455–500
[2] 周春光, 孙明芳, 王甦菁, 等. 基于稀疏张量的人脸图像特征提取[J]. 吉林大学学报(工学版), 2012, 42(6): 1521–1526
ZHOU Chun-guang, SUN Ming-fang, WANG Su-jing, et al. Face image feature extraction method based on sparse tensor[J]. Journal of Jilin University (Engineering and Technology Edition), 2012, 42(6): 1521–1526
[3] 计雨含, 王晓东, 姚宇. 基于张量奇异值分解的人脸识别方法[J]. 计算机应用, 2015, 35(S1): 117–121, 126
JI Yuhan*, WANG Xiaodong, YAO Yu. Face identification based on tensor singular value decomposition[J]. Journal of Computer Applications, 2015, 35(S1): 117–121, 126
[4] ELIZABETH N, MISHA K, LIOR H. Image classification using local tensor singular value decompositions[J]. 2017 IEEE 7th International Workshop on Computational Advances in Multi-Sensor Adaptive Processing (CAMSAP), pp.1-5.
[5] 梁秋霞, 何光辉, 陈如丽, 等. 基于非负张量分解的人脸识别算法研究[J]. 计算机科学, 2016, 43(10): 312–316
LIANG Qiu-xia, HE Guang-hui, CHEN Ru-li, CHU Jian-pu. Research of face recognition algorithm based on nonnegative tensor factorization[J]. Computrter Science, 2016, 43(10): 312–316
[6] 任卿龙. 基于支持张量机的遥感图像目标分级识别研究[D]. 哈尔滨: 哈尔滨工业大学. 2017.
Ren Qinglong. Research on target hierarchical recognition of remote sensing images based on support tensor machine[D]. Harbin: Harbin Institute of Technology, 2017.
[7] 王珊. 基于非负张量分解的多通道音频信号盲分离研究[D]. 南京: 东南大学, 2015.
WANG Shan. Multichannnel audio signals blind sepatation based on nonnegative tensor factorization[D]. Nanjing: Southeast University, 2015.
[8] XING L, MA Q Zhu M. Tensor semantic model for an audio classification system[J]. Science China Information Sciences, 2013, 56(6): 1–9
[9] 杨立东.基于张量分析的多因素音频信号建模与应用研究[D]. 北京: 北京理工大学. 2016.
YANG Lidong. Study on multifactor audio signal modeling and applications based on tensor analysis[D]. Beijing: Beijing Institute of Technology, 2016.
[10] 樊帆. 基于张量分解的声矢量传感器阵列信号处理方法研究[D]. 吉林: 吉林大学. 2017.
FAN Fan. Research on signal processing of acoustic vector sensor array based on tensor decomposition[D]. Jilin: JiLin University, 2016.
[11] 邱静雯, 官金安. 张量时频空模式在脑电信号特征提取中的研究[J]. 计算机与数字工程, 2015(43): 5: 797–804
QIU Jingwen, GUAN Jinan. Fan fan tensor temporal-frequency-spatial pattern in feature extraction of EEG Signal[J]. Computer & Digital Engineering, 2015(43): 5: 797–804
[12] 佘青山, 马鹏刚, 马玉良, 等. 基于张量线性拉普拉斯判别的肌电特征提取方法[J]. 东南大学学报, 2017(47): 6:1117–1122
SHE Qingshan, MA Penggang, MA Yuliang, et al. EMG feature extraction based on tensor linear Laplacian discriminant[J]. Journal of Southeast University (Natural Science Edition), 2017(47): 6:1117–1122
[13] 徐曼. 基于张量分解的胎儿心电信号提取研究[D]. 重庆: 重庆大学. 2017.
XU Man. Study on method of fetal electrocar diagram extraction based on tensor decomposition[D]. Chongqing: Chongqing University.
[14] ND Sidiropoulos, LD Lathauwer, et al. Tensor decomposition for signal processing and machine learning[J]. IEEE Transactions on Signal Processing, 2016, 65(13): 3551–3582
[15] STPHAN R, OLEKSANDR S, STEPHAN G. Introduction to tensor decompositions and their applications in machine learning[J]. ArXiv preprint, 1711. 10781, 2017.
[16] 尹凤, 雷晓军, 黄光鑫. 关于高阶张量的秩 分解方法[J]. 成都理工大学学报(自然科学版), 2019, 42(1): 123–128
YIN Feng, LEI Xiaojun, HUANG Guangxin. On the research of rank decomposition of trnsor with high order[J]. Journal of Chengdu University of Technology (Science & Technology Edition), 2019, 42(1): 123–128
[17] LATHAUWER L D, MOOR B D, VANDEWALLE J. A multilinear singular value decomposition[J]. SIAM Journalon Matrix Analysis and Applications, 2000, 21(4): 1253–1278
[18] 王梦璇. 基于小波包分解的舰船辐射噪声特征提取方法研究[J]. 电子设计工程, 2014, 22(4): 81–83
WANG Meng-xuan. Research of ship-radiated noise feature extraction method based on wavelet packet decomposition[J]. Electronic Design Engineering, 2014, 22(4): 81–83
[19] 程锦盛, 杜选民, 周胜增, 等. 基于目标MFCC特征的监督学习方法在被动声呐目标识别中的应用研究[J]. 舰船科学技术, 2018, 40(9): 116–121
CHENG Jin-sheng, DU Xuan-min, ZHOU Sheng-zeng. Application of the MFCC feature based supervised learning method in passive sonar target recognition[J]. Ship Science and Technology, 2018, 40(9): 116–121
