Feature Extraction Methods for Underwater Acoustic Target Recognition of Divers.

The extraction of typical features of underwater target signals and excellent recognition algorithms are the keys to achieving underwater acoustic target recognition of divers. This paper proposes a feature extraction method for diver signals: frequency-domain multi-sub-band energy (FMSE), aiming to achieve accurate recognition of diver underwater acoustic targets by passive sonar. The impact of the presence or absence of targets, different numbers of targets, different signal-to-noise ratios, and different detection distances on this method was studied based on experimental data under different conditions, such as water pools and lakes. It was found that the FMSE method has the best robustness and performance compared with two other signal feature extraction methods: mel frequency cepstral coefficient filtering and gammatone frequency cepstral coefficient filtering. Combined with the commonly used recognition algorithm of support vector machines, the FMSE method can achieve a comprehensive recognition accuracy of over 94% for frogman underwater acoustic targets. This indicates that the FMSE method is suitable for underwater acoustic recognition of diver targets.

In-Lab Demonstration of an Underwater Acoustic Spiral Source.

Underwater acoustic spiral sources can generate spiral acoustic fields where the phase depends on the bearing angle. This allows estimating the bearing angle of a single hydrophone relative to a single source and implementing localization equipment, e.g., for target detection or unmanned underwater vehicle navigation, without requiring an array of hydrophones and/or projectors. A spiral acoustic source prototype made out of a single standard piezoceramic cylinder, which is able to generate both spiral and circular fields, is presented. This paper reports the prototyping process and the multi-frequency acoustic tests performed in a water tank where the spiral source was characterized in terms of the transmitting voltage response, phase, and horizontal and vertical directivity patterns. A receiving calibration method for the spiral source is proposed and showed a maximum angle error of 3° when the calibration and the operation were carried out in the same conditions and a mean angle error of up to 6° for frequencies above 25 kHz when the same conditions were not fulfilled.

Low-to-Mid-Frequency Monopole Source Levels of Underwater Noise from Small Recreational Vessels in the St. Lawrence Estuary Beluga Critical Habitat.

Anthropogenic noise from navigation is a major contributor to the disturbance of the acoustic soundscape in underwater environments containing noise-sensitive life forms. While previous studies mostly developed protocols for the empirical determination of noise source levels associated with the world's commercial fleet, this work explores the radiated noise emitted by small recreational vessels that thrive in many coastal waters, such as in the St. Lawrence Estuary beluga population's summer habitat. Hydrophone-based measurements in the Saguenay River (QC, Canada) were carried out during the summers of 2021 and 2022. Shore-based observations identified 45 isolated transits of small, motorized vessels and were able to track their displacement during their passage near the hydrophone. Received noise levels at the hydrophone typically fell below the hearing audiogram of the endangered St. Lawrence Estuary beluga. Monopole source levels at low frequencies (0.1-≲2 kHz) held on average twice the acoustic power compared to their mid-frequency (≳2-30 kHz) counterparts. The speed over ground of recreational vessel showed a positive correlation with the back-propagated monopole source levels. Estimations of the mid-frequency noise levels based on low-frequency measurements should be used moderately.

A Survey of Underwater Acoustic Data Classification Methods Using Deep Learning for Shoreline Surveillance.

This paper presents a comprehensive overview of current deep-learning methods for automatic object classification of underwater sonar data for shoreline surveillance, concentrating mostly on the classification of vessels from passive sonar data and the identification of objects of interest from active sonar (such as minelike objects, human figures or debris of wrecked ships). Not only is the contribution of this work to provide a systematic description of the state of the art of this field, but also to identify five main ingredients in its current development: the application of deep-learning methods using convolutional layers alone; deep-learning methods that apply biologically inspired feature-extraction filters as a preprocessing step; classification of data from frequency and time-frequency analysis; methods using machine learning to extract features from original signals; and transfer learning methods. This paper also describes some of the most important datasets cited in the literature and discusses data-augmentation techniques. The latter are used for coping with the scarcity of annotated sonar datasets from real maritime missions.

Underwater Acoustic Target Recognition Based on Attention Residual Network.

Underwater acoustic target recognition is very complex due to the lack of labeled data sets, the complexity of the marine environment, and the interference of background noise. In order to enhance it, we propose an attention-based residual network recognition method (AResnet). The method can be used to identify ship-radiated noise in different environments. Firstly, a residual network is used to extract the deep abstract features of three-dimensional fusion features, and then a channel attention module is used to enhance different channels. Finally, the features are classified by the joint supervision of cross-entropy and central loss functions. At the same time, for the recognition of ship-radiated noise in other environments, we use the pre-training network AResnet to extract the deep acoustic features and apply the network structure to underwater acoustic target recognition after fine-tuning. The two sets of ship radiation noise datasets are verified, the DeepShip dataset is trained and verified, and the average recognition accuracy is 99%. Then, the trained AResnet structure is fine-tuned and applied to the ShipsEar dataset. The average recognition accuracy is 98%, which is better than the comparison method.

New Insights into the Design and Application of a Passive Acoustic Monitoring System for the Assessment of the Good Environmental Status in Spanish Marine Waters.

Passive acoustic monitoring systems allow for non-invasive monitoring of underwater species and anthropogenic noise. One of these systems has been developed keeping in mind the need to create a user-friendly tool to obtain the ambient noise indicators, while at the same time providing a powerful tool for marine scientists and biologists to progress in studying the effect of human activities on species and ecosystems. The device is based on a low-power processor with ad-hoc electronics, ensuring that the system has efficient energy management, and that the storage capacity is large enough to allow deployments for long periods. An application is presented using data from an acoustic campaign done in 2018 at El Gorguel (Cartagena, Spain). The results show a good agreement between theoretical maps created using AIS data and the ambient noise level indicators measured in the frequency bands of 63 Hz and 125 Hz specified in the directive 11 of the EU Marine Strategy Framework Directive. Using a 2D representation, these ambient noise indicators have enabled repetitive events and daily variations in boat traffic to be identified. The ship noise registered can also be used to track ships by using the acoustic signatures of the engine propellers' noise.

Monitoring and Reconstruction of the Shape of the Detection Units in KM3NeT Using Acoustic and Compass Sensors.

The KM3NeT underwater neutrino telescope comprises thousands of optical modules forming 3D arrays to detect the Cherenkov light produced by particles generated after a neutrino interaction in the medium. The modules are arranged in detection units-vertical structures with 18 modules at different heights, anchored to the seabed and kept vertical by the buoyancy of the optical modules and a top buoy. The optical modules are, thus, subject to movements due to sea currents. For a correct reconstruction of events detected by the telescope, it is necessary to know the relative position and orientation of modules with 10 cm and a few degrees accuracy, respectively. For this, an Acoustic Positioning System with a piezoceramic transducer installed in each module and a long baseline of acoustic transmitters and receivers on the seabed are used. In addition, there is a system of compass and accelerometers inside the optical modules to determine their orientation. A model of mechanical equations is used to reconstruct the shape of the detection unit taking as input the information from the positioning/orientation sensors and using the sea current velocity and direction as free parameters. The mechanical equations take the buoyancy and the drag force of the elements of the detection unit into account. This work describes the full process that is implemented in KM3NeT to monitor the modules and the shape of the detection units from the measured position and orientation data.

Ship Type Classification by Convolutional Neural Networks with Auditory-like Mechanisms.

Ship type classification with radiated noise helps monitor the noise of shipping around the hydrophone deployment site. This paper introduces a convolutional neural network with several auditory-like mechanisms for ship type classification. The proposed model mainly includes a cochlea model and an auditory center model. In cochlea model, acoustic signal decomposition at basement membrane is implemented by time convolutional layer with auditory filters and dilated convolutions. The transformation of neural patterns at hair cells is modeled by a time frequency conversion layer to extract auditory features. In the auditory center model, auditory features are first selectively emphasized in a supervised manner. Then, spectro-temporal patterns are extracted by deep architecture with multistage auditory mechanisms. The whole model is optimized with an objective function of ship type classification to form the plasticity of the auditory system. The contributions compared with an auditory inspired convolutional neural network include the improvements in dilated convolutions, deep architecture and target layer. The proposed model can extract auditory features from a raw hydrophone signal and identify types of ships under different working conditions. The model achieved a classification accuracy of 87.2% on four ship types and ocean background noise.

Automatic Bluefin Tuna Sizing with a Combined Acoustic and Optical Sensor.

A proposal is described for an underwater sensor combining an acoustic device with an optical one to automatically size juvenile bluefin tuna from a ventral perspective. Acoustic and optical information is acquired when the tuna are swimming freely and the fish cross our combined sensor's field of view. Image processing techniques are used to identify and classify fish traces in acoustic data (echogram), while the video frames are processed by fitting a deformable model of the fishes' ventral silhouette. Finally, the fish are sized combining the processed acoustic and optical data, once the correspondence between the two kinds of data is verified. The proposed system is able to automatically give accurate measurements of the tuna's Snout-Fork Length (SFL) and width. In comparison with our previously validated automatic sizing procedure with stereoscopic vision, this proposal improves the samples per hour of computing time by 7.2 times in a tank with 77 juveniles of Atlantic bluefin tuna (Thunnus thynnus), without compromising the accuracy of the measurements. This work validates the procedure for combining acoustic and optical data for fish sizing and is the first step towards an embedded sensor, whose electronics and processing capabilities should be optimized to be autonomous in terms of the power supply and to enable real-time processing.

Competitive Deep-Belief Networks for Underwater Acoustic Target Recognition.

Underwater acoustic target recognition based on ship-radiated noise belongs to the small-sample-size recognition problems. A competitive deep-belief network is proposed to learn features with more discriminative information from labeled and unlabeled samples. The proposed model consists of four stages: (1) A standard restricted Boltzmann machine is pretrained using a large number of unlabeled data to initialize its parameters; (2) the hidden units are grouped according to categories, which provides an initial clustering model for competitive learning; (3) competitive training and back-propagation algorithms are used to update the parameters to accomplish the task of clustering; (4) by applying layer-wise training and supervised fine-tuning, a deep neural network is built to obtain features. Experimental results show that the proposed method can achieve classification accuracy of 90.89%, which is 8.95% higher than the accuracy obtained by the compared methods. In addition, the highest accuracy of our method is obtained with fewer features than other methods.

DOA Estimation for Underwater Wideband Weak Targets Based on Coherent Signal Subspace and Compressed Sensing.

Direction of arrival (DOA) estimation is the basis for underwater target localization and tracking using towed line array sonar devices. A method of DOA estimation for underwater wideband weak targets based on coherent signal subspace (CSS) processing and compressed sensing (CS) theory is proposed. Under the CSS processing framework, wideband frequency focusing is accompanied by a two-sided correlation transformation, allowing the DOA of underwater wideband targets to be estimated based on the spatial sparsity of the targets and the compressed sensing reconstruction algorithm. Through analysis and processing of simulation data and marine trial data, it is shown that this method can accomplish the DOA estimation of underwater wideband weak targets. Results also show that this method can considerably improve the spatial spectrum of weak target signals, enhancing the ability to detect them. It can solve the problems of low directional resolution and unreliable weak-target detection in traditional beamforming technology. Compared with the conventional minimum variance distortionless response beamformers (MVDR), this method has many advantages, such as higher directional resolution, wider detection range, fewer required snapshots and more accurate detection for weak targets.

A Glider-Assisted Link Disruption Restoration Mechanism in Underwater Acoustic Sensor Networks.

Underwater acoustic sensor networks (UASNs) have become a hot research topic. In UASNs, nodes can be affected by ocean currents and external forces, which could result in sudden link disruption. Therefore, designing a flexible and efficient link disruption restoration mechanism to ensure the network connectivity is a challenge. In the paper, we propose a glider-assisted restoration mechanism which includes link disruption recognition and related link restoring mechanism. In the link disruption recognition mechanism, the cluster heads collect the link disruption information and then schedule gliders acting as relay nodes to restore the disrupted link. Considering the glider's sawtooth motion, we design a relay location optimization algorithm with a consideration of both the glider's trajectory and acoustic channel attenuation model. The utility function is established by minimizing the channel attenuation and the optimal location of glider is solved by a multiplier method. The glider-assisted restoration mechanism can greatly improve the packet delivery rate and reduce the communication energy consumption and it is more general for the restoration of different link disruption scenarios. The simulation results show that glider-assisted restoration mechanism can improve the delivery rate of data packets by 15-33% compared with cooperative opportunistic routing (OVAR), the hop-by-hop vector-based forwarding (HH-VBF) and the vector based forward (VBF) methods, and reduce communication energy consumption by 20-58% for a typical network's setting.

Great cormorants (Phalacrocorax carbo) can detect auditory cues while diving.

In-air hearing in birds has been thoroughly investigated. Sound provides birds with auditory information for species and individual recognition from their complex vocalizations, as well as cues while foraging and for avoiding predators. Some 10% of existing species of birds obtain their food under the water surface. Whether some of these birds make use of acoustic cues while underwater is unknown. An interesting species in this respect is the great cormorant (Phalacrocorax carbo), being one of the most effective marine predators and relying on the aquatic environment for food year round. Here, its underwater hearing abilities were investigated using psychophysics, where the bird learned to detect the presence or absence of a tone while submerged. The greatest sensitivity was found at 2 kHz, with an underwater hearing threshold of 71 dB re 1 µPa rms. The great cormorant is better at hearing underwater than expected, and the hearing thresholds are comparable to seals and toothed whales in the frequency band 1-4 kHz. This opens up the possibility of cormorants and other aquatic birds having special adaptations for underwater hearing and making use of underwater acoustic cues from, e.g., conspecifics, their surroundings, as well as prey and predators.

Potential Competitive Dynamics of Acoustic Ecology.

The top predators in coastal marine ecosystems, such as whales, dolphins, seabirds, and large predatory fishes (including sharks), may compete with each other to exploit food aggregations. Finding these patchy food sources and being first to a food patch could provide a significant competitive advantage. Our hypothesis is that food patches have specific sound signatures that marine predators could detect and that acoustic sources and animal sensory capabilities may contribute to competition dynamics. Preliminary analysis shows that diving gannets have a distinct spectral signature between 80 and 200 Hz, which falls within the hearing sensitivity of large pelagic fishes. Therefore, we suggest that diving birds may contribute to the sound signatures of food aggregations, linking competition dynamics both above and below the water surface.

Evidence for the initiation of decompression sickness by exposure to intense underwater sound.

Mass stranding of cetaceans (whales and dolphins), in close association with the activity of naval sonar systems, has been reported on numerous occasions. Necropsy showed bubble-associated lesions similar to those described in human decompression sickness (DCS). We examined the hypothesis that exposure to underwater sound may potentiate DCS. Rats were subjected to immersion and simulated dives with and without simultaneous acoustic transmissions at pressure levels and frequencies of 204 dB/8 kHz and 183.3 dB/15 kHz. DCS severity was assessed using the rotating wheel method. Recording of somatosensory evoked potentials (SSEPs) was employed under general anesthesia as an electrophysiological measure of neurologic insult. A significantly higher rate of decompression sickness was found among animals exposed to the 204-dB/8-kHz sound field. Significantly higher pathological SSEPs scores were noted for both underwater sound protocols. Pathological SSEPs scores in animals immersed during the acoustic transmissions, but without changes in ambient pressure, were comparable to those observed in animals exposed to the dive profile. The results demonstrate induction of neurological damage by intense underwater sound during immersion, with a further deleterious effect when this was combined with decompression stress. The study outcome has potential implications for human diving safety and may provide an explanation for the mass stranding of cetaceans purportedly associated with sonar activity.

A Hybrid Sender- and Receiver-Initiated Protocol Scheme in Underwater Acoustic Sensor Networks.

In this paper, we propose a method for sharing the handshakes of control packets among multiple nodes, which we call a hybrid sender- and receiver-initiated (HSR) protocol scheme. Handshake-sharing can be achieved by inviting neighbors to join the current handshake and by allowing them to send their data packets without requiring extra handshakes. Thus, HSR can reduce the signaling overhead involved in control packet exchanges during handshakes, as well as resolve the spatial unfairness problem between nodes. From an operational perspective, HSR resembles the well-known handshake-sharing scheme referred to as the medium access control (MAC) protocol using reverse opportunistic packet appending (ROPA). However, in ROPA the waiting time is not controllable for the receiver's neighbors and thus unexpected collisions may occur at the receiver due to hidden neighbors, whereas the proposed scheme allows all nodes to avoid hidden-node-induced collisions according to an elaborately calculated waiting time. Our computer simulations demonstrated that HSR outperforms ROPA with respect to both the throughput and delay by around 9.65% and 11.36%, respectively.

The underwater soundscape of the North Sea.

As awareness on the impact of anthropogenic underwater noise on marine life grows, underwater noise measurement programs are needed to determine the current status of marine areas and monitor long-term trends. The Joint Monitoring Programme for Ambient Noise in the North Sea (JOMOPANS) collaborative project was funded by the EU Interreg to collect a unique dataset of underwater noise levels at 19 sites across the North Sea, spanning many different countries and covering the period from 2019 to 2020. The ambient noise from this dataset has been characterised and compared - setting a benchmark for future measurements in the North Sea area. By identifying clusters with similar sound characteristics in three broadband frequency bands (25-160 Hz, 0.2-1.6 kHz, and 2-10 kHz), geographical areas that are similarly affected by sound have been identified. The measured underwater sound levels show a persistent and spatially uniform correlation with wind speed at high frequencies (above 1 kHz) and a correlation with the distance from ships at mid and high frequencies (between 40 Hz and 4 kHz). Correlation with ocean current velocity at low frequencies (up to 200 Hz), which are susceptible to nonacoustic contamination by flow noise, was also evaluated. These correlations were evaluated and simplified linear scaling laws for wind and current speeds were derived. The presented dataset provides a baseline for underwater noise measurements in the North Sea and shows that spatial variability of the dominant sound sources must be considered to predict the impact of noise reduction measures.

Natural soundscapes of lowland river habitats and the potential threat of urban noise pollution to migratory fish.

Migratory fish populations have experienced great declines, and considerable effort have been put into reducing stressors, such as chemical pollution and physical barriers. However, the importance of natural sounds as an information source and potential problems caused by noise pollution remain largely unexplored. The spatial distribution of sound sources and variation in propagation characteristics could provide migratory fish with acoustic cues about habitat suitability, predator presence, food availability and conspecific presence. We here investigated the relationship between natural soundscapes and local river conditions and we explored the presence of human-related sounds in these natural soundscapes. We found that 1a) natural river sound profiles vary with river scale and cross-sectional position, and that 1b) depth, width, water velocity, and distance from shore were all significant factors in explaining local soundscape variation. We also found 2a) audible human activities in almost all our underwater recordings and urban and suburban river parts had elevated sound levels relative to rural river parts. Furthermore, 2b) daytime levels were louder than night time sound levels, and bridges and nearby road traffic were much more prominent with diurnal and weekly patterns of anthropogenic noise in the river systems. We believe our data show high potential for natural soundscapes of low-land river habitat to serve as important environmental cues to migratory fish. However, anthropogenic noise may be particularly problematic due to the omnipresence, and relatively loud levels relative to the modest dynamic range of the natural sound sources, in these slow-flowing freshwater systems.

Observation of Ultra-High-Q Resonators in the Ultrasound via Bound States in the Continuum.

The confinement of waves in open systems represents a fundamental phenomenon extensively explored across various branches of wave physics. Recently, significant attention is directed toward bound states in the continuum (BIC), a class of modes that are trapped but do not decay in an otherwise unbounded continuum. Here, the theoretical investigation and experimental demonstration of the existence of quasi-bound states in the continuum (QBIC) for ultrasonic waves are achieved by leveraging an elastic Fabry-Pérot metasurface resonator. Several intriguing properties of the ultrasound quasi-bound states in the continuum that are robust to parameter scanning are unveiled, and experimental evidence of a remarkable Q-factor of 350 at ≈1 MHz frequency, far exceeding the state-of-the-art using a fully acoustic underwater system is presented. The findings contribute novel insights into the understanding of BIC for acoustic waves, offering a new paradigm for the design of efficient, ultra-high Q-factor ultrasound devices.

Predicting the contribution of climate change on North Atlantic underwater sound propagation.

Since the industrial revolution, oceans have become substantially noisier. The noise increase is mainly caused by increased shipping, resource exploration, and infrastructure development affecting marine life at multiple levels, including behavior and physiology. Together with increasing anthropogenic noise, climate change is altering the thermal structure of the oceans, which in turn might affect noise propagation. During this century, we are witnessing an increase in seawater temperature and a decrease in ocean pH. Ocean acidification will decrease sound absorption at low frequencies (<10 kHz), enhancing long-range sound propagation. At the same time, temperature changes can modify the sound speed profile, leading to the creation or disappearance of sound ducts in which sound can propagate over large distances. The worldwide effect of climate change was explored for the winter and summer seasons using the (2018 to 2022) and (2094 to 2098, projected) atmospheric and seawater temperature, salinity, pH and wind speed as input. Using numerical modelling, we here explore the impact of climate change on underwater sound propagation. The future climate variables were taken from a Community Earth System Model v2 (CESM2) simulations forced under the concentration-driven SSP2-4.5 and SSP5-8.5 scenarios. The sound modeling results show, for future climate change scenarios, a global increase of sound speed at different depths (5, 125, 300, and 640 m) except for the North Atlantic Ocean and the Norwegian Sea, where in the upper 125 m sound speed will decrease by as much as 40 m s-1. This decrease in sound speed results in a new sub-surface duct in the upper 200 m of the water column allowing ship noise to propagate over large distances (>500 km). In the case of the Northeast Atlantic Ocean, this sub-surface duct will only be present during winter, leading to similar total mean square pressure level (SPLtot) values in the summer for both (2018 to 2022) and (2094 to 2098). We observed a strong and similar correlation for the two climate change scenarios, with an increase of the top 200 m SPLtot and a slowdown of Atlantic Meridional Overturning Circulation (AMOC) leading to an increase of SPLtot at the end of the century by 7 dB.