基于深度流形学习的水中目标声信号特征提取

周钰,王津,滕飞,潘必胜,王友瑞,雷迎科

振动与冲击 ›› 2024, Vol. 43 ›› Issue (9) : 50-59.

PDF(2614 KB)
PDF(2614 KB)
振动与冲击 ›› 2024, Vol. 43 ›› Issue (9) : 50-59.
论文

基于深度流形学习的水中目标声信号特征提取

  • 周钰,王津,滕飞,潘必胜,王友瑞,雷迎科
作者信息 +

Feature extraction of underwater target acoustic signals based on deep manifold learning

  • ZHOU Yu, WANG Jin, TENG Fei, PAN Bisheng, WANG Yourui, LEI Yingke
Author information +
文章历史 +

摘要

随着海洋环境日益复杂,水中目标声信号观测数据呈现出高维数、非线性、非结构化等特点,无疑给水中目标声信号特征提取带来严峻挑战,该文提出了一种基于流形自编码器的水中目标声信号特征提取方法。首先,通过自编码器重建误差对原始数据进行全局优化,找出潜在的低维表示结果;然后,利用流形学习保持近邻重构权值的思想对潜在表示实施局部约束,保留其内在拓扑结构;最后,引入生成对抗网络架构进行正则化处理,使潜在表示服从特定分布,从而实现一种局部与全局的联合保持低维嵌入方法。在DeepShip深水船公开数据集上进行实验,使用该文方法提取4种深水船数据的本质特征,为评估该类特征的质量水平,利用经典分类器SVM进行分类识别,与现有深度学习以及流形学习特征提取方法对比,识别精度平均提高14.96%。

Abstract

With the increasing complexity of the marine environment, the data obtained from observing underwater target acoustic signals exhibits several challenging characteristics, including high dimensionality, nonlinearity, and lack of structure. These characteristics undoubtedly pose significant difficulties in extracting features from underwater target acoustic signals. In this study, a novel method for extracting features from underwater target acoustic signals is proposed, utilizing manifold autoencoders. Initially, the original data is globally optimized by leveraging the autoencoder reconstruction error to identify potential low-dimensional representations. Subsequently, the concept of preserving neighboring reconstruction weights through manifold learning is employed to enforce local constraints on the latent representation, thereby preserving its inherent topological structure. Finally, a generative adversarial network architecture is introduced for regularization, ensuring that the latent representation adheres to a specific distribution. This approach achieves a synergistic preservation of both local and global low-dimensional embedding. The proposed method was applied to extract essential features from four types of deep-water ships in the DeepShip open dataset. The quality of these features was evaluated by employing the classic classifier SVM for classification recognition. A comparison was conducted with existing methods for feature extraction in deep learning and manifold learning. The results showed an average improvement of 14.96% in recognition accuracy.

关键词

流形 / 自编码 / 生成对抗 / 特征提取

Key words

manifold / auto-encoder / generating adversarial / feature extraction

引用本文

导出引用
周钰,王津,滕飞,潘必胜,王友瑞,雷迎科. 基于深度流形学习的水中目标声信号特征提取[J]. 振动与冲击, 2024, 43(9): 50-59
ZHOU Yu, WANG Jin, TENG Fei, PAN Bisheng, WANG Yourui, LEI Yingke. Feature extraction of underwater target acoustic signals based on deep manifold learning[J]. Journal of Vibration and Shock, 2024, 43(9): 50-59

参考文献

[1] Vaccaro R J. The past, present, and the future of underwater acoustic signal processing[J]. IEEE Signal Processing Magazine, 1998, 15(4): 21-51. [2] Zhao Y, Zhang X, Huang J, et al. Techniques on ship radiated noise power spectrum feature extraction and realization[C]//2010 The 2nd International Conference on Computer and Automation Engineering (ICCAE). IEEE, 2010, 1: 554-558. [3] Huang J, Zhao J, Xie Y. Source classification using pole method of AR model[C]//1997 IEEE International Conference on Acoustics, Speech, and Signal Processing. IEEE, 1997, 1: 567-570. [4] Yao Q, Wang Y, Yang Y. Underwater Acoustic Target Recognition Based on Data Augmentation and Residual CNN[J]. Electronics, 2023, 12(5): 1206. [5] Zhang Q, Da L, Zhang Y, et al. Integrated neural networks based on feature fusion for underwater target recognition[J]. Applied Acoustics, 2021, 182: 108261. [6] Azimi-Sadjadi M R, Yao D, Huang Q, et al. Underwater target classification using wavelet packets and neural networks[J]. IEEE Transactions on neural networks, 2000, 11(3): 784-794. [7] Li G, Liu F, Yang H. Research on feature extraction method of ship radiated noise with K-nearest neighbor mutual information variational mode decomposition, neural network estimation time entropy and self-organizing map neural network[J]. Measurement, 2022, 199: 111446. [8] Liu F, Li G, Yang H. A new feature extraction method of ship radiated noise based on variational mode decomposition, weighted fluctuation-based dispersion entropy and relevance vector machine[J]. Ocean Engineering, 2022, 266: 113143. [9] Xie D, Hong S, Yao C. Optimized variational mode decomposition and permutation entropy with their application in feature extraction of ship-radiated noise[J]. Entropy, 2021, 23(5): 503. [10] Kantz H, Schreiber T. Nonlinear time series analysis[M]. Cambridge university press, 2004. [11] Khishe M. Drw-ae: A deep recurrent-wavelet autoencoder for underwater target recognition[J]. IEEE Journal of Oceanic Engineering, 2022, 47(4): 1083-1098. [12] Chen Y, Xu X. The research of underwater target recognition method based on deep learning[C]//2017 IEEE International Conference on Signal Processing, Communications and Computing (ICSPCC). IEEE, 2017: 1-5. [13] Luo X, Feng Y. An underwater acoustic target recognition method based on restricted Boltzmann machine[J]. Sensors, 2020, 20(18): 5399. [14] Jain V, Saul L K. Exploratory analysis and visualization of speech and music by locally linear embedding[C]//2004 IEEE International Conference on Acoustics, Speech, and Signal Processing. IEEE, 2004, 3: iii-984. [15] Errity A. Exploring the dimensionality of speech using manifold learning and dimensionality reduction methods[D]. Dublin City University, 2010. [16] 刘辉, 杨俊安, 王一. 基于流形学习的声目标特征提取方法研究[J]. 物理学报, 2011, 60(7): 437-443. Liu H, Yang J A, Wang Y. Research on sound target feature extraction method based on manifold learning. Journal of Physics, 2011, 60(7): 437-443. [17] 刘辉, 杨俊安, 王一. 基于去相关邻域保持判别投影的声目标特征提取[J]. 电子测量与仪器学报, 2010, 24(10): 905-910. Liu H, Yang J A, Wang Y. Acoustic target feature extraction based on decorrelated neighborhood preserving discriminant projection[J]. Journal of Electronic Measurement and Instrumentation, 2010, 24(10): 905-910. [18] 刘辉, 杨俊安, 王一, 等. 基于改进测地距离的等度规映射及其在声目标特征提取中的应用[J]. 兵工学报, 2012, 33(10): 1178. Liu H, Yang J A, Wang Y. Improved geodesic distance-based isometric mapping and its application to acoustic target feature extraction[J]. Journal of Military Engineering, 2012, 33(10): 1178. [19] 吕志超, 王好忠, 白一奇. 流形学习在浅海水声通信中的应用[J]. 电 子 与 信 息 学 报, 2021, 43: 3. Lv Z C, Wang H Z, Bai Y Q. Application of manifold learning in shallow sea hydroacoustic communication[J]. Journal of Electronics and Information, 2021, 43: 3. [20] Roweis S T, Saul L K. Nonlinear dimensionality reduction by locally linear embedding[J]. science, 2000, 290(5500): 2323-2326. [21] Yu W, Zeng G, Luo P, et al. Embedding with autoencoder regularization[C]//Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2013, Prague, Czech Republic, September 23-27, 2013, Proceedings, Part III 13. Springer Berlin Heidelberg, 2013: 208-223. [22] Ng A. Cs294a lecture notes: Sparse autoencoder[J]. URL: https://web. stanford. edu/class/cs294a/sparseAutoencoder 2011new. pdf, 2010. [23] Goodfellow I, Pouget-Abadie J, Mirza M, et al. Generative adversarial networks[J]. Communications of the ACM, 2020, 63(11): 139-144. [24] Makhzani A, Shlens J, Jaitly N, et al. Adversarial autoencoders[J]. arXiv preprint arXiv:1511.05644, 2015. [25] 曾向阳,智能水中目标识别. 北京:国防工业出版社,2016. Zeng X Y, Intelligent water target recognition. Beijing: Defense Industry Press, 2016. [26] Rumelhart D E, Hinton G E, Williams R J. Learning representations by back-propagating errors[J]. nature, 1986, 323(6088): 533-536. [27] Ho Y, Wookey S. The real-world-weight cross-entropy loss function: Modeling the costs of mislabeling[J]. IEEE access, 2019, 8: 4806-4813. [28] Irfan M, Jiangbin Z, Ali S, et al. DeepShip: An underwater acoustic benchmark dataset and a separable convolution based autoencoder for classification[J]. Expert Systems with Applications, 2021, 183: 115270. [29] Laurens V D M, Hinton G. Visualizing data using t-SNE[J]. Journal of machine learning research, 2008, 9(2605): 2579-2605.

PDF(2614 KB)

Accesses

Citation

Detail

段落导航
相关文章

/