Article
Acoustics
Yunan Zhu, Biao Wang, Youwen Zhang, Jianghui Li, Chengxi Wu
Summary: This paper proposes a CNN-based FBMC system for underwater acoustic communication. By using a pre-trained CNN model as the receiver, the system can directly recover transmitted symbols and avoid inherent imaginary interference. Analysis and simulation under various system parameters show admirable performance of the proposed system for signal detection.
Article
Acoustics
Yaohui Hu, Shuping Han, Houquan Li, Heng Zhao, Gang Yang, Jingfeng Xu
Summary: A deep learning-based direct sequence spread spectrum (DSSS) underwater acoustic (UWA) communication system was proposed to enhance performance under Doppler effects and low SNR. The TCN model was used as the receiver, which outperformed the conventional system under Doppler effects and complex shallow water acoustic channel. A transfer training algorithm was employed to fine-tune the model using real-time sea data, resulting in improved communication effectiveness.
Article
Computer Science, Information Systems
Fangtong Xie, Yunan Zhu, Biao Wang, Wu Wang, Pian Jin
Summary: This paper proposes a data-driven underwater acoustic filter bank multicarrier (FBMC) communication system based on convolutional autoencoder networks, which incorporates deep learning theory into traditional communication systems to address the problems of high complexity and poor bit error rate (BER) performance in underwater acoustic environments. The proposed system achieves global optimization through two one-dimensional convolutional (Conv1D) modules at the transmitter and receiver, realizes signal reconstruction through end-to-end training, effectively avoids inherent interference, and improves the reliability of the communication system. Furthermore, dense-block modules are introduced for feature reuse in the network. Simulation results demonstrate that the proposed method outperforms conventional FBMC systems with channel equalization algorithms under specific measured channel conditions in the Qingjiang River at a certain moment.
Article
Computer Science, Hardware & Architecture
Lihuan Huang, Yue Wang, Qunfei Zhang, Jing Han, Weijie Tan, Zhi Tian
Summary: In this study, machine learning techniques are used to enhance underwater acoustic communication with intelligent capabilities. The research explores ML algorithms relevant to UAC networks. Due to the unique characteristics of marine environments, traditional model-based design methods are no longer effective or reliable in UAC systems.
IEEE WIRELESS COMMUNICATIONS
(2022)
Article
Chemistry, Multidisciplinary
Evan Lucas, Zhaohui Wang
Summary: This article uses convolutional neural networks to learn useful features in underwater acoustic communication channels for predicting performance, outperforming traditional supervised learning models. The study also demonstrates the universality of the learned features across different channels.
APPLIED SCIENCES-BASEL
(2022)
Article
Acoustics
Cheng Fan, Li Wei, Zhaohui Wang
Summary: This work investigates the adaptation of communication strategies to the dynamics of the underwater acoustic (UWA) channel. Three communication strategies, including Frequency-Hopped Binary Frequency Shift Keying (FH-BFSK), Single-Carrier (SC) communication, and multicarrier communication, are considered. A reinforcement learning algorithm, the Deep Deterministic Policy Gradient (DDPG) method with a Gumbel-softmax scheme, is employed for intelligent and adaptive switching among these strategies. Simulation and experimental results demonstrate that the proposed method outperforms random selection and direct feedback methods in time-varying channels.
Article
Acoustics
Dajun Sun, Xiaoping Hong, Hongyu Cui
Summary: This paper investigates a method for dynamically tracking and compensating the Doppler spread of spread spectrum signals in underwater acoustic communications using a novel Kalman-based algorithm. The accurate estimation of waveform dilation/compression in the time domain and compensating magnitude distortion induced by velocity variation are achieved through processing passband signals with higher accuracy.
Article
Engineering, Electrical & Electronic
Abigail Lee-Leon, Chau Yuen, Dorien Herremans
Summary: In this paper, a novel receiver system utilizing Deep Belief Network (DBN) to combat signal distortion in underwater environments caused by Doppler effect and multi-path propagation is proposed. Results show that the DBN based receiver system outperforms traditional methods in handling received signals affected by the Doppler effect and multi-path propagation.
IEEE TRANSACTIONS ON COMMUNICATIONS
(2021)
Article
Engineering, Marine
Yuzhi Zhang, Shumin Zhang, Bin Wang, Yang Liu, Weigang Bai, Xiaohong Shen
Summary: This paper proposes a deep learning-based signal detection method for underwater acoustic (UWA) OTFS communication, and compares it with conventional methods, showing that the proposed method achieves a lower bit error rate (BER).
JOURNAL OF MARINE SCIENCE AND ENGINEERING
(2022)
Article
Acoustics
Daihui Li, Feng Liu, Tongsheng Shen, Liang Chen, Dexin Zhao
Summary: This study proposes a data augmentation method based on underwater acoustic channel modeling and transfer learning to address the challenges of data scarcity and noise interference in underwater acoustic target recognition. The augmented signal is generated using underwater acoustic channel modeling, and a feature-based transfer learning method is used to narrow the distribution differences between augmented and observed data. The effectiveness of the proposed methods is proved by utilizing data augmentation in the model training process, which improves the accuracy and noise robustness of the recognition model, especially when observed data is scarce.
Article
Engineering, Marine
Yuzhi Zhang, Shumin Zhang, Yang Wang, Qingyuan Liu, Xiangxiang Li
Summary: This paper proposes a novel approach for underwater acoustic communications using a model-driven deep learning technique. The proposed method incorporates a residual neural network into the OTFS channel estimation process, achieving superior accuracy in channel estimation and reducing the system's bit error rate.
JOURNAL OF MARINE SCIENCE AND ENGINEERING
(2023)
Article
Robotics
Matheus M. Dos Santos, Giovanni G. De Giacomo, Paulo L. J. Drews-Jr, Silvia S. C. Botelho
Summary: This study proposes a cross-domain and cross-view localization framework that improves the localization of underwater vehicles in partially structured environments by identifying the correlation between color aerial images and underwater acoustic images.
IEEE ROBOTICS AND AUTOMATION LETTERS
(2022)
Article
Engineering, Electrical & Electronic
Xiao Feng, Junfeng Wang, Xiaoyan Kuai, Mingzhang Zhou, Haixin Sun, Jianghui Li
Summary: This paper proposes a new framework based on sparse Bayesian learning and generalized approximate message passing for joint impulsive noise mitigation and channel estimation. The GAMP algorithm is introduced to reduce computational complexity, and a novel GAMP-based temporal framework is proposed for slow time-varying scenarios. Simulation and sea-trial results demonstrate the superiority of the proposed algorithms compared to existing methods in terms of multiple performance metrics.
IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY
(2022)
Article
Acoustics
Yuji Liu, Huixiu Chen, Biao Wang
Summary: This paper proposes a method for estimating the arrival direction of underwater acoustic signals, using two-channel real and imaginary covariance matrices as input signals for a convolutional neural network. Compared to the traditional MUSIC algorithm, the CNN algorithm shows higher accuracy and shorter estimation time in low SNR environments.
Article
Acoustics
Kwang B. Yoo, Geoffrey F. Edelmann
Summary: This study demonstrates a method to achieve low complexity and low error decoding of direct-sequence spread spectrum signals in underwater acoustic communications, by measuring and updating a passive conjugate matched filter to mitigate multipath effects and strong Doppler.
Article
Geography, Physical
Julien Chaput, Rick Aster, Marianne Karplus, Nori Nakata, P. Gerstoft, P. D. Bromirski, A. Nyblade, R. A. Stephen, D. A. Wiens
Summary: Ambient seismic recordings in Antarctica reveal time-variable resonance peaks in near-surface firn layers, resulting from trapped seismic waves. The splitting of these peaks on the horizontal components indicates frequency-dependent anisotropy in the firn and underlying ice caused by overlapping mechanisms driven by ice flow. A novel algorithm was used to estimate the splitting magnitudes and axes, which were compared with active source anisotropy measurements, showing good agreement. The study also discovered a novel plastic stretching mechanism of anisotropy in the near-surface firn, where the fast direction aligns with accelerating ice shelf flow.
JOURNAL OF GLACIOLOGY
(2023)
Article
Acoustics
Domenico De Salvio, Michael J. J. Bianco, Peter Gerstoft, Dario D'Orazio, Massimo Garai
Summary: Noise exposure affects people's comfort and well-being in various contexts, such as work or learning environments. To assess sound environments, it is crucial to be able to separate sound sources in real contexts. Long-term monitoring provides abundant data that can be analyzed using machine and deep learning algorithms.
JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA
(2023)
Article
Acoustics
Yongsung Park, Florian Meyer, Peter Gerstoft
Summary: This paper presents a Bayesian estimation method for sequential direction finding. The method estimates the number of directions of arrivals (DOAs) and their DOAs using a factor graph and belief propagation. Variational Bayesian inference is then used to update the estimates. The proposed method shows improved performance compared to nonsequential approaches in scenarios involving multiple time steps and time-varying DOAs.
JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA
(2023)
Article
Computer Science, Interdisciplinary Applications
Chih-Chieh Chien, William F. Jenkins II, Peter Gerstoft, Mark Zumberge, Robert Mellors
Summary: This study investigates the relationship between fluid injection in enhanced geothermal systems and induced seismicity from hydraulic fracturing using unsupervised machine learning. Spectrograms of the detected signals are dimensionally reduced to a lower-dimensional latent feature space with a deep neural network called autoencoder. Gaussian mixture model clustering is then performed on this feature space to assign each signal to one of 7 classes. The results show bimodal spatiotemporal distributions, with a shallow mode occurring between 250 and 500 m and a deep mode centered around 750 m. The correlation between the clustered signal classes and injection-related activities is weak or non-existent. This study demonstrates the capability to analyze not only when and where signals are detected, but also their types, facilitating rapid and targeted data exploration and providing insights into source mechanisms.
COMPUTERS AND GEOTECHNICS
(2023)
Article
Acoustics
Efren Fernandez-Grande, Xenofon Karakonstantis, Diego Caviedes-Nozal, Peter Gerstoft
Summary: This study investigates the use of generative adversarial networks to reconstruct sound fields from experimental data. It demonstrates that generative models can improve the spatio-temporal reconstruction of a sound field by extending its bandwidth. The results show that these models can recover lost energy at high frequencies and have potential applications in computational acoustics.
JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA
(2023)
Article
Acoustics
Ruixian Liu, Peter Gerstoft, Michael J. Bianco, Bhaskar D. D. Rao
Summary: A data-driven method is proposed to map the spatial variations of physical properties for a material by identifying spatially dependent partial differential equations (PDEs) from observations of dynamical behaviors. This method based on L1-norm minimization does not require any assumed active PDE terms and is capable of identifying spatially dependent PDEs from measurements of phenomena. It has been demonstrated in various experimental settings, including real laser measurements, and is efficient and robust against noise.
JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA
(2023)
Article
Acoustics
Zoi-Heleni Michalopoulou, Peter Gerstoft
Summary: This work utilizes Gaussian processes (GPs) to capture correlation of the acoustic field at different depths in the ocean for pre-processing acoustic data in underwater waveguide. The data are denoised and interpolated using GPs, generating densely populated acoustic fields at virtual arrays for source localization and environmental inversion. Field predictions are made by computing replicas at virtual receivers, and the correlations among field measurements are selected through kernel functions with estimated hyperparameters. The approach is found to be superior to conventional beamformer MFI and the Matern kernel is preferred over the Gaussian kernel.
JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA
(2023)
Article
Acoustics
William F. Jenkins, Peter Gerstoft, Yongsung Park
Summary: This paper proposes a sample-efficient sequential Bayesian optimization strategy for source localization with a geoacoustic model. By modeling the objective function as a Gaussian process surrogate model and using a heuristic acquisition function, the proposed method can converge on optimal solutions rapidly.
JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA
(2023)
Article
Acoustics
Ishan D. Khurjekar, Peter Gerstoft
Summary: Uncertainty quantification (UQ) of deep learning (DL)-based acoustic estimation methods is crucial for real-world applicability, and conformal prediction (CP) provides statistically rigorous confidence intervals without distributional assumptions.
JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA
(2023)
Article
Acoustics
Jeung-Hoon Lee, Yongsung Park, Peter Gerstoft
Summary: Direction-of-arrival estimation is challenging for spatially undersampled signals, but frequency-difference beamforming offers an alternative approach. However, unconventional beamforming sacrifices spatial resolution, making it difficult to distinguish closely spaced targets. To overcome this limitation, a simple yet effective method is proposed by formulating frequency-difference beamforming as a sparse signal reconstruction problem. The proposed compressive frequency-difference beamforming outperforms the conventional method in terms of separation when the signal-to-noise ratio exceeds 4 dB, as supported by ocean data from the FAF06 experiment.
JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA
(2023)
Proceedings Paper
Computer Science, Hardware & Architecture
Venkatesh Sathyanarayanan, Peter Gerstoft, Aly El Gamal
Summary: Deep learning has achieved significant performance improvements in modulation classification. Previous research mainly focused on model construction and benchmark dataset RML16, while this study adopted a data-centric deep learning approach. By addressing the errors and ad-hoc parameter choices in RML16, a more realistic dataset RML22 was introduced, and the Python source code used to generate RML22 was shared for further improvements.
2023 15TH INTERNATIONAL CONFERENCE ON COMMUNICATION SYSTEMS & NETWORKS, COMSNETS
(2023)
Article
Engineering, Electrical & Electronic
Yifan Wu, Michael B. Wakin, Peter Gerstoft
Summary: This paper addresses the continuous DOA estimation problem with multiple frequencies using an atomic norm minimization approach. The problem is formulated as a semi-definite program, which can be solved by an SDP solver. The authors provide a dual certificate condition to certify the optimality of the SDP solution and propose a robust solution to combat spatial aliasing. Numerical results support the theoretical findings and demonstrate the effectiveness of the proposed method.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2023)
Proceedings Paper
Acoustics
Christoph F. Mecklenbraeuker, Peter Gerstoft, Esa Ollila
Summary: Based on the assumption of robustness, we derive a sparse direction estimation method that accurately estimates the direction of arrival in complex noise environments. The method performs well under various loss functions and has similar performance to the classical method for Gaussian noise.
2022 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP)
(2022)
Proceedings Paper
Acoustics
Michael J. Bianco, Peter Gerstoft
Summary: Machine learning approaches have been successful in source localization, but often suffer from limited labeled data. By combining semi-supervised learning and traditional signal processing, a hybrid approach can achieve better source localization.
2022 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP)
(2022)