针对水下目标识别中的舰船辐射噪声分类识别问题，提出一种稀疏谱特征提取方法。根据傅里叶变换机理，将其指数函数视为基函数，通过线性和非线性化基函数中的频率因子，构造了频率线性变化和非线性变化2种稀疏基，通过直接信号基分解提取了舰船辐射噪声基分解稀疏谱特征，通过求解稀疏分解优化问题，获得了舰船辐射噪声的稀疏系数谱特征。与常规功率谱特征相比，在相同特征维度下，非线性频率基分解稀疏谱特征和非线性频率稀疏系数谱特征能提供更高的低频频率分辨率，增强了舰船特征谱中的低频信息。利用实测舰船辐射噪声数据集进行了分类识别对比实验，分别提取了常规功率谱特征、线性及非线性稀疏谱等共7种谱特征，在同一特征维度下运用最近邻分类器进行了分类识别检验。实验结果表明，稀疏谱特征的正确分类识别概率更高，其中非线性频率基分解稀疏谱特征表现最佳，具有较大的潜力用于水下无源声呐目标识别。
A sparse spectral feature extraction method is proposed for the classification and identification of ship radiated noise in underwater target recognition. Based on the Fourier transform mechanism, the exponential function is regarded as the basis function. By linearizing and nonlinearizing the frequency factors of the basis function, two types of sparse bases, namely, frequency linear variation and nonlinear variation, are constructed. The sparse spectral features of ship radiated noise are extracted through direct signal basis decomposition. By solving the sparse decomposition optimization problem, the sparse coefficient spectral features of ship radiated noise are obtained. Compared with conventional power spectral features, the nonlinear frequency basis decomposition sparse spectral features and nonlinear frequency sparse coefficient spectral features can provide higher low-frequency resolution in the same feature dimension, enhancing the low-frequency information in the ship spectrum feature. Using a measured dataset of ship radiated noise, we conducted classification and recognition experiments. Seven spectral features were extracted, including conventional power spectrum features, as well as linear and nonlinear sparse spectrum features, et al. Classification and recognition tests were performed using a nearest neighbor classifier within the same feature dimension. The experimental results indicate that sparse spectrum features exhibit a higher probability of correct classification and recognition. Notably, the nonlinear frequency basis decomposition sparse spectrum feature performs optimally, indicating significant potential for underwater passive sonar target recognition.
2025,47(19): 157-161 收稿日期：2024-12-24
DOI：10.3404/j.issn.1672-7649.2025.19.025
分类号：U674；TB566
基金项目：国家自然科学面上基金项目（62371464）
作者简介：康春玉(1975-),男,博士,教授,研究方向为水声探测与目标识别
参考文献：
[1] 方世良, 杜栓平, 罗昕炜, 等. 水声目标特征分析与识别技术[J]. 中国科学院院刊, 2019, 34(3): 297-305.
FANG S L, DU S P, LUO X W, et al. Development of underwater acoustic target feature analysis and recognition technology[J]. Bulletin of Chinese Academy of Sciences, 2019, 34(3): 297-305.
[2] 程玉胜, 李智忠, 邱家兴. 水声目标识别[M], 北京: 科学出版社, 2018.
[3] 张雨萌. 水声目标调制线谱特征智能化提取方法研究[D]. 西安: 西安电子科技大学, 2021.
[4] 李新. 功率谱估计在舰船噪声特征提取中的应用仿真[J]. 舰船科学技术, 2022, 44(4): 43-46.
LI X. Application simulation of power spectrum estimation in ship noise feature extraction[J]. Ship Science and Technology, 2022, 44(4): 43-46.
[5] 李岩洲, 康春玉, 夏志军, 等. 舰船辐射噪声听觉谱张量特征提取与分类识别[J]. 科技与创新, 2022, (16): 38-40.
LI Y Z, KANG C Y, XIA Z J, et al. Auditory spectrum tensor feature extraction and recognition of the ship radiated noise[J]. Science and Technology & Innovation, 2022, (16): 38-40.
[6] 沈鑫玉, 陈涛, 郭良浩, 等. 遗传算法优化变分模态分解提取舰船辐射噪声特征线谱方法[J]. 应用声学, 2023, 43(1): 1-14.
SHEN X Y, CHEN T, GUO L H, et al. Feature line spectrum extraction of ship radiated noise based on variable mode decomposition optimized by genetic algorithm[J]. Journal of Applied Acoustics, 2023, 43(1): 1-14.
[7] COIFMAN R, WICKERHAUSER M. Entropy-based algorithms for best basis selection[J]. IEEE Transactions Information Theory, 1992, 38: 1713-1716.
[8] 王佳维, 许枫, 杨娟. 联合稀疏表示在水下目标分类中的应用[J]. 声学学报, 2022, 47(4): 471-480.
WANG J W, XU F, YANG J. Application of joint sparse representation in underwater target classification[J]. ACTA ACUSTICA, 2022, 47(4): 471-480.
[9] OIKONOMOU V P, GEORGIADIS K, KALAGANIS F, et al. A sparse representation classification scheme for the recognition of affective and cognitive brain processes in neuromarketing[J]. Sensors, 2023, 23(5): 2480-2497.
[10] 王源源. 基于改进稀疏表示的SAR图像目标识别方法[J]. 电光与控制, 2023, 30(9): 42-46.
WANG Y Y. SAR target recognition based on modified sparse representation[J]. Electronics Optics & Control, 2023, 30(9): 42-46.
[11] 李非燕, 霍宏涛, 李静, 等. 基于多特征和改进稀疏表示的高光谱图像分类[J]. 光学学报, 2019, 39(5): 351-359.
LI F Y, HUO H T, LI J, et al. Hyperspectral image classification via multiple-feature based improved sparse representation[J]. Acta Optica Sinica, 2019, 39(5): 351-359.
[12] 马杲东. 基于稀疏表示的小样本人脸图像集分类研究[D]. 南京: 南京邮电大学, 2022.
[13] 焦义民. 舰船辐射噪声稀疏特征提取与应用研究[D]. 大连: 大连舰艇学院, 2015.
[14] 曾理, 黄建军, 刘亚峰, 等. 第2讲 压缩感知的关键技术及其研究进展[J]. 军事通信技术, 2011, 32(4): 88-94.
ZENG L, HUANG J J, LIU Y F, et al. Advances in theory of compressive sensing[J]. Journal of Military Communications Technology, 2011, 32(4): 88-94.
[15] 康春玉, 夏志军, 章新华, 等. 水下无源声呐目标听觉域张量特征提取方法[J]. 声学学报, 2020, 45(6): 824-829.
KANG C Y, XIA Z J, ZHANG X H, et al. Tensor feature extraction of underwater passive sonar target based on auditory model[J]. Acta Acustica, 2020, 45(6): 824-829.
