| Li Qianqian,Shi Juan,Li Zhenglin,Luo Yu,Yang Fanlin,Zhang Kai.Acoustic sound speed profile inversion based on orthogonal matching pursuit[J].Acta Oceanologica Sinica,2019,38(11):149-157 |
| 基于正交匹配追踪的声层析方法 |
| Acoustic sound speed profile inversion based on orthogonal matching pursuit |
| 投稿时间：2018-06-20 |
| DOI：10.1007/s13131-019-1505-4 |
| 中文关键词: 海水声速 水声学 压缩感知 正交匹配追踪 |
| 英文关键词:acoustic sound speed|ocean acoustics|compressive sensing|orthogonal matching pursuit |
| 基金项目:The National Natural Science Foundation of China under contract No. 11704225; the Shandong Provincial Natural Science Foundation under contract No. ZR2016AQ23; the State Key Laboratory of Acoustics, Chinese Academy of Sciences under contract No. SKLA201902; the National Key Research and Development Program of China contract No. 2018YFC1405900; the SDUST Research Fund under contract No. 2019TDJH103; the Talent Introduction Plan for Youth Innovation Team in Universities of Shandong Province (Innovation Team of Satellite Positioning and Navigation). |
| 作者 | 单位 | E-mail | | 李倩倩 | 山东科技大学, 测绘科学与工程学院海洋测绘系, 青岛 266590
哈尔滨工程大学 水声工程学院, 哈尔滨 150001
中国科学院声学研究所, 声场声信息国家重点实验室, 北京 100190 | lqq@mail.ioa.ac.cn | | 史娟 | 山东科技大学, 测绘科学与工程学院海洋测绘系, 青岛 266590 | | | 李整林 | 中国科学院声学研究所, 声场声信息国家重点实验室, 北京 100190 | | | 罗宇 | 山东科技大学, 测绘科学与工程学院海洋测绘系, 青岛 266590 | | | 阳凡林 | 山东科技大学, 测绘科学与工程学院海洋测绘系, 青岛 266590 | | | 张凯 | 山东科技大学, 测绘科学与工程学院海洋测绘系, 青岛 266590 | |
|
| 摘要点击次数: 1586 |
| 全文下载次数: 1003 |
| 中文摘要: |
| 声速剖面的变化会对声传播产生较大的影响，经验正交函数模型经常用来实现对声速剖面数据的简化描述。然而在内波、湍流等海水不均匀性存在时，这种正则化操作会造成声速重构精度的大幅降低。本文利用字典学习生成声速剖面的非正交原子，在稀疏编码时采用正交匹配追踪（OMP，Orthogonal Matching Pursuit）算法，更新字典则使用KSVD （Kernel Singular Value Decomposition）的字典更新算法。由于字典学习不需要强制使用正交条件，对于训练数据更加灵活，从而可以使用少数的原子组合达到更高的重构精度。利用一次浅海声学实验多次测量的声速剖面研究了海水声速剖面的经验正交函数表示和字典学习，研究表明：相比于正交函数表示，学习字典可以利用少数原子（甚至一个原子）更好的解释声速剖面扰动。字典学习可以提高声速剖面的稀疏性，从而提高声速剖面的反演精度。 |
| 英文摘要: |
| The estimation of ocean sound speed profiles (SSPs) requires the inversion of an acoustic field using limited observations. Such inverse problems are underdetermined, and require regularization to ensure physically realistic solutions. The empirical orthonormal function (EOF) is capable of a very large compression of the data set. In this paper, the non-linear response of the sound pressure to SSP is linearized using a first order Taylor expansion, and the pressure is expanded in a sparse domain using EOFs. Since the parameters of the inverse model are sparse, compressive sensing (CS) can help solve such underdetermined problems accurately, efficiently, and with enhanced resolution. Here, the orthogonal matching pursuit (OMP) is used to estimate range-independent acoustic SSPs using the simulated acoustic field. The superior resolution of OMP is demonstrated with the SSP data from the South China Sea experiment. By shortening the duration of the training set, the temporal correlation between EOF and test sets is enhanced, and the accuracy of sound velocity inversion is improved. The SSP estimation error versus depth is calculated, and the 99% confidence interval of error is within ±0.6 m/s. The 82% of mean absolute error (MAE) is less than 1 m/s. It is shown that SSPs can be well estimated using OMP. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|