Real Acoustic Data Research Articles

Parameter estimation and sensitivities of a Bayesian source localization method in an urban environment

Acoustic signals propagating in urban environments are influenced by rough-surface scattering, multipath reflections, and diffraction. The performance of conventional source localization algorithms often suffers when these effects are present. Bayesian approaches, however, are particularly well suited to incorporating physics-based statistical models for the signal propagation. Previously, we found that the complex Wishart distribution, which describes fully saturated scattered signals across a network of receivers, can be readily employed in a Bayesian framework. In this presentation, a new source localization algorithm based on the Wishart distribution signal model is described and tested on real acoustic data. The experimental data were collected within a mock urban environment as part of a NATO urban acoustics-seismics experiment in Walenstadt, Switzerland. Four acoustic arrays recorded signatures from gunshots and ground vehicles. The present study uses the measured signals to investigate the sensitivity and relative importance of the model parameters, including source and noise amplitudes, frequency band, prior signal sampling, and sensor-sensor correlation behavior. Results are presented in the form of source location probability maps and localization error metrics. Interpretations of the study, including strengths and weaknesses of the new method, are discussed.

Optimal compactness of fractional Fourier domain characterizes frequency modulated signals

The Fourier transform (FT) is a mathematical tool widely used in signal processing applications; however, it presents limitations when dealing with non-stationary time series. By considering the fractional powers of the FT operator, a generalized version is obtained known as the fractional FT. This transformation allows free rotations of the time–frequency plane that can be exploited for processing frequency modulated signals. This work addresses the problem of characterizing noisy, multicomponent, and non-linear frequency modulated signals through a proper order of the fractional FT, whose kernel consists of a chirp with linear frequency modulation. The estimation of the optimal fractional FT order obeys a strategy that includes the quantification of the compactness of fractional Fourier domains and the search for the order that leads to the most compact domain. For this purpose, five compactness measures are assessed in combination with four different optimization algorithms. Numerical experiments are performed on synthetic signals, generated under distinct frequency modulation conditions, and on real acoustic signals. The results reveal that the spectral second moment and the spectral entropy provide robust and reliable measures of the compactness of the fractional Fourier domain. These metrics enable an effective computation of the optimal fractional order that describes the frequency modulation content of the underlying signal. The optimization algorithms assessed in this study yield similar estimations of the optimal fractional order, yet the coarse-to-fine algorithm is more efficient in terms of computation time, followed by the particle swarm optimization algorithm. Moreover, it is verified that the strategy can be adopted for extracting dynamical information of the frequency modulation content from synthetic signals with multiple linear and non-linear components and from real acoustic data, e.g., bat and bird recordings. The extensive assessment of the signal processing strategy based on the fractional FT outlined in this work provides relevant information for exploring further applications with time series captured when studying complex and non-stationary processes, such as biological, medical, or economic systems.

Bayesian detection and tracking of odontocetes in 3-D from their echolocation clicks.

Localization and tracking of marine animals can reveal key insights into their behaviors underwater that would otherwise remain unexplored. A promising nonintrusive approach to obtaining location information of marine animals is to process their bioacoustic signals, which are passively recorded using multiple hydrophones. In this paper, a data processing chain that automatically detects and tracks multiple odontocetes (toothed whales) in three dimensions (3-D) from their echolocation clicks recorded with volumetric hydrophone arrays is proposed. First, the time-difference-of-arrival (TDOA) measurements are extracted with a generalized cross-correlation that whitens the received acoustic signals based on the instrument noise statistics. Subsequently, odontocetes are tracked in the TDOA domain using a graph-based multi-target tracking (MTT) method to reject false TDOA measurements and close gaps of missed detections. The resulting TDOA estimates are then used by another graph-based MTT stage that estimates odontocete tracks in 3-D. The tracking capability of the proposed data processing chain is demonstrated on real acoustic data provided by two volumetric hydrophone arrays that recorded echolocation clicks from Cuvier's beaked whales (Ziphius cavirostris). Simulation results show that the presented MTT method using 3-D can outperform an existing approach that relies on manual annotation.

SINR: Deconvolving Circular SAS Images Using Implicit Neural Representations

Circular synthetic aperture sonars (CSAS) capture multiple observations of a scene to reconstruct high-resolution images. We can characterize resolution by modeling CSAS imaging as the convolution between a scene's underlying point scattering distribution and a system-dependent point spread function (PSF). The PSF is a function of the system bandwidth and determines a fixed degree of blurring on reconstructed imagery. In theory, deconvolution overcomes bandwidth limitations by reversing the PSF-induced blur and recovering the scene's scattering distribution. However, deconvolution is an ill-posed inverse problem and sensitive to noise. We propose an optimization method that leverages an implicit neural representation (INR) to deconvolve CSAS images. We highlight the performance of our SAS INR pipeline, which we call <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">SINR</i> , by implementing and comparing to existing deconvolution methods. Additionally, prior SAS deconvolution methods assume a spatially-invariant PSF, which we demonstrate yields subpar performance in practice. We provide theory and methods to account for a spatially-varying CSAS PSF, and demonstrate that doing so enables SINR to achieve superior deconvolution performance on simulated and real acoustic SAS data. We provide code <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/awreed/CSAS_Deconvolution_INR</uri> (click here) <xref ref-type="fn" rid="fn1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">¹</xref> <fn id="fn1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><label>¹</label> Code URL: <uri>https://github.com/awreed/CSAS_Deconvolution_INR</uri> </fn> to encourage reproducibility of research.

Assessment of Risks Induced by Countermining Unexploded Large-Charge Historical Ordnance in a Shallow Water Environment—Part I: Real Case Study

The goal of the work presented in a two-companion paper is to pave the way for reliably assessing the risks of damage to buildings on the shore, induced by the detonation of unexploded historical ordnance (UXO) of large weights in variable shallow water environments with a water depth less than 50 m. The risk assessment is quantified through the seismic magnitude on the Richter scale, induced by the detonation of charges of different weights (between 80- and 680-kg TNT-equivalent). This metric is investigated experimentally using a coupled seismo-acoustic approach within the framework of a UXO clearance (countermining) campaign in the Mediterranean Sea. Analysis of real acoustic and seismic data shows that, compared to a charge detonation in water, a similar detonation on the seabed generates seismic signals of lower frequencies and higher amplitudes that propagate in the seabed. The larger the charge weight, the higher the seismic amplitude. Besides the explosion-coast distance, the ground properties also affect the signals. The sediments favor a longer signal duration and the presence of late dispersive and very low-frequency signals with a large amplitude, whereas the rocky grounds better preserve the high-frequency energy propagation. For the local environment considered in this study, a charge detonation on the seafloor generates seismic events of higher magnitudes compared to a detonation in water. However, these magnitudes are likely low enough to prevent any large damage in the nearby inland infrastructures.

Residual physical modeling and semi-supervised source localization

Source localization in reverberant environments remains an open challenge which ML techniques have shown promise in addressing. Real-world acoustic environments contain irregular boundaries, scattering, and diffracting elements (e.g., furniture and uneven surfaces) which are impractical to model using acoustic propagation software. We develop a hybrid approach to acoustic source localization which utilizes both analytical, signal processing-based localization, and machine learning (ML) to provide a data-driven source localization model. In this approach, the output of an analytical source localization model (e.g., SRP-PHAT or MUSIC) is augmented by a neural network (NN) system. This augmented system consists of a variational autoencoder and localization network, which is designed to correct the analytic estimate. The networks are trained via semi-supervised learning, on both labeled and unlabeled inputs. The NNs in this approach parameterize stochastic functions, which facilitate a statistically principled learning approach via variational inference (VI). The stochasticity also helps quantify source location uncertainty under the VI approximate posterior. For simulated and real acoustic data, the hybrid approach generalizes better and is more efficient than ML or analytic approaches alone.

Model-based localization of deep-diving cetaceans using towed line array acoustic data.

Passive acoustic monitoring using a towed line array of hydrophones is a standard method for localizing cetaceans during line-transect cetacean abundance surveys. Perpendicular distances estimated between localized whales and the trackline are essential for abundance estimation using acoustic data. Uncertainties in the acoustic data from hydrophone movement, sound propagation effects, errors in the time of arrival differences, and whale depth are not accounted for by most two-dimensional localization methods. Consequently, location and distance estimates for deep-diving cetaceans may be biased, creating uncertainty in abundance estimates. Here, a model-based localization approach is applied to towed line array acoustic data that incorporates sound propagation effects, accounts for sources of error, and localizes in three dimensions. The whale's true distance, ship trajectory, and whale movement greatly affected localization results in simulations. The localization method was applied to real acoustic data from two separate sperm whales, resulting in three-dimensional distance and depth estimates with position bounds for each whale. By incorporating sources of error, this three-dimensional model-based approach provides a method to address and integrate the inherent uncertainties in towed array acoustic data for more robust localization.

Iterative Diagonal Unloading Beamforming for Multiple Acoustic Sources Localization Using Compact Sensor Arrays

Compact acoustic sensor arrays are nowadays popular devices for audio sensing due to minimum size and lightness characteristics that make it attractive for numerous applications such as human-computer interaction, acoustic scene analysis, teleconferencing systems, hearing aid, robotics, and bioacoustics. A compact array has a small inter-microphone distance that allows to reduce the spatial aliasing effect in the acoustic source localization problem, degrading however the spatial resolution. Thus, as a consequence, the capability of this class of sensors in the localization of multiple acoustic sources is rather limited. In this paper, we present an iterative diagonal unloading (IDU) beamforming suitable for the direction of arrivals (DOAs) estimation of concurrent multiple sources using compact microphone arrays. The method is based on the identification of the dominant signal, which corresponds to the largest peak in the pseudo-spectrum (i.e., the acoustic response map) computed by the DU beamformer, and on an iterative cancellation of the dominant signal from the covariance matrix to compute a new spatial pseudo-spectrum. All the pseudo-spectrum responses computed with the iterative procedure are summed to obtain a high resolution map of the acoustic scene. We analyze the DOAs estimation performance using compact uniform linear arrays (ULAs) and spherical microphone arrays (SMAs) with simulations and real acoustic data. The results show that the proposed IDU beamforming increases the spatial resolution allowing an effective localization of concurrent multiple sources.

A matched-field processing approach to ranging surface vessels using a single hydrophone and measured replica fields.

A matched-field processing (MFP) approach is proposed to provide instantaneous estimates of the horizontal range of a surface vessel moving in shallow water, using a single hydrophone located above the sea bottom. The field to be matched is the cepstrum of the acoustic signal received at the hydrophone. A set of replica fields (cepstra), each at a known horizontal range, is generated using the recorded acoustic data for a typical surface vessel transit in the area of interest. The instantaneous horizontal range of any surface vessel moving in the same area of interest is estimated by finding the replica field that best matches the observed field. The proposed method is tested using real acoustic data recorded from a bottom-mounted linear array of eight hydrophones, for six transits of a small ship and a single transit of a rigid-hulled inflatable boat in shallow water. In this experiment, the replica fields are generated using the acoustic data from the two hydrophones at both ends of the array for one of the small ship transits. The proposed MFP method is applied, in turn, to the acoustic data from each hydrophone for each vessel transit, and the results demonstrate the effectiveness of the method.

Direction-of-arrival tracking using a co-prime microphone array: A particle filter perspective

Direction-of-arrival (DOA) estimation using a linear co-prime microphone array has attracted tremendous attention in many acoustic applications. Spatial smoothing MUSIC (SS-MUSIC)-based methods and compressive sensing-based methods are two popular approaches for DOA estimation using co-prime linear arrays. However, these methods exhibit high computational complexity and suffer from performance degradation under low signal-to-noise ratio conditions. This paper proposes a particle filter (PF)-based DOA tracking algorithm using a linear co-prime array. The proposed PF employs the current spatial information obtained from the array measurement and temporal information obtained from a constant-velocity motion model to estimate the DOA values recursively, which is a more suitable approach for real-world applications. An improved likelihood function model derived from the SS-MUSIC pseudo-spectra is implemented in the PF-based algorithm. Theoretical analyses and simulation results and the results of real acoustic data processing experiments are presented to demonstrate the effectiveness of the proposed PF-based algorithm.

Diagonal Unloading Beamforming in the Spherical Harmonic Domain for Acoustic Source Localization in Reverberant Environments

Spherical microphone arrays allow the sound field analysis in three dimensions with the advantage of having the same resolution in all directions. By considering the frequency-independent character of the steering vectors in the spherical harmonic (SH) domain, we propose a very low-complexity SH diagonal unloading (DU) beamforming with a novel frequency smoothing power transform (FSPT) of the covariance matrices. We consider the direction of arrival (DOA) estimation problem of acoustic sources in reverberant conditions. The DU beamforming provides high resolution directional response since it exploits the subspace orthogonality property of the covariance matrix by the removal or the attenuation of the signal subspaces, obtained through the subtraction of an opportune diagonal matrix from the covariance matrix. The FSPT aims at smoothing the narrowband covariance matrices of the entire set of frequency domain components, and it pursues this goal by minimizing the narrowband error contributions due to reverberation in the broadband frequency smoothing covariance matrix. We analyze the DOA estimation performance using speech signals with simulations and real acoustic data in reverberant conditions. The results show that the proposed SH-DU-FSPT has a DOA estimation performance comparable to that of high resolution state-of-the-art methods with a significant reduction of the computational cost, since the steering directional responses are computed on the broadband frequency smoothing covariance matrix.

Acoustic localization of vehicular sources using distributed sensors.

This paper considers the problem of locating ground vehicles using their acoustic signatures recorded by unattended passive acoustic sensors. Acoustic signatures of the ground sources captured by different sensors within a cluster are used to generate direction of arrival (DoA) of the propagating wavefronts. Using the estimated DoAs of disparate distributed sensor node clusters, this paper introduced and compared several different existing target localization methods that provide the location and velocity estimates of a moving source. A robust source localization method is then proposed to account for large DoA errors and outliers which often occur in realistic settings. This method does not use any prior knowledge of the dynamical model of the moving source. The effectiveness and complexity of these methods are compared using synthesized and real acoustic signature data sets.

Exploiting platform motion for passive source localization with a co-prime sampled large aperture array.

Co-prime array geometries have received a great deal of attention due to their ability to discriminate O(MN) sources with only O(M + N) sensors. This has been demonstrated both theoretically and in simulation. However, there are many practical limitations that make it difficult to realize the enhanced degrees of freedom when applying co-prime geometries to real acoustic data taken on a horizontal line array. For instance, co-prime sampling leads to grating lobes that can obscure lower signal-to-noise-ratio acoustic signals making them difficult to detect. In this work, a synthetic aperture (SA) method is presented for filling in holes and increasing redundancy in the difference co-array by exploiting array motion. The SA method is applied to acoustic data collected off the Southeastern shore of Florida on a fixed large aperture horizontal array. Array motion is simulated by taking a co-prime sampled subarray and virtually moving it along the horizontal aperture of the fixed array. It is demonstrated that SA processing on real acoustic data results in reduced side-lobe and grating lobe levels compared to that of the physical co-prime aperture.

Exploiting CNNs for Improving Acoustic Source Localization in Noisy and Reverberant Conditions

This paper discusses the application of convolutional neural networks (CNNs) to minimum variance distortionless response localization schemes. We investigate the direction of arrival estimation problems in noisy and reverberant conditions using a uniform linear array (ULA). CNNs are used to process the multichannel data from the ULA and to improve the data fusion scheme, which is performed in the steered response power computation. CNNs improve the incoherent frequency fusion of the narrowband response power by weighting the components, reducing the deleterious effects of those components affected by artifacts due to noise and reverberation. The use of CNNs avoids the necessity of previously encoding the multichannel data into selected acoustic cues with the advantage to exploit its ability in recognizing geometrical pattern similarity. Experiments with both simulated and real acoustic data demonstrate the superior localization performance of the proposed SRP beamformer with respect to other state-of-the-art techniques.

The Underwater Acoustic Target Recognition Algorithm Based on Evidence Clustering

In underwater acoustic target recognition, the target signal is usually complex and the samples which are difficult to obtain also have some uncertain information. In order to effectively solve these problems, the evidence clustering recognition algorithm (TECRA) is presented. In this new method, the k-nearest neighbor are first determined by using the feature distance between the object and its neighbors in each class of the training set, and a reasonable initial basic belief assignments (bba's) for each target data are constructed by the improved k-nearest neighbor classification algorithm. Then the final global bba's of the target is obtained by optimizing the objective function of the algorithm. Finally the object can be recognized by the fusion result and the classification rule presented in the paper. Several experiments based on real underwater acoustic data sets are made to test the effectiveness of TECRA in comparison with some other methods. The results indicate that TECRA can effectively improve the recognition accuracy.

Passive acoustic measurement of bedload grain size distribution using self-generated noise

Abstract. Monitoring sediment transport processes in rivers is of particular interest to engineers and scientists to assess the stability of rivers and hydraulic structures. Various methods for sediment transport process description were proposed using conventional or surrogate measurement techniques. This paper addresses the topic of the passive acoustic monitoring of bedload transport in rivers and especially the estimation of the bedload grain size distribution from self-generated noise. It discusses the feasibility of linking the acoustic signal spectrum shape to bedload grain sizes involved in elastic impacts with the river bed treated as a massive slab. Bedload grain size distribution is estimated by a regularized algebraic inversion scheme fed with the power spectrum density of river noise estimated from one hydrophone. The inversion methodology relies upon a physical model that predicts the acoustic field generated by the collision between rigid bodies. Here we proposed an analytic model of the acoustic energy spectrum generated by the impacts between a sphere and a slab. The proposed model computes the power spectral density of bedload noise using a linear system of analytic energy spectra weighted by the grain size distribution. The algebraic system of equations is then solved by least square optimization and solution regularization methods. The result of inversion leads directly to the estimation of the bedload grain size distribution. The inversion method was applied to real acoustic data from passive acoustics experiments realized on the Isère River, in France. The inversion of in situ measured spectra reveals good estimations of grain size distribution, fairly close to what was estimated by physical sampling instruments. These results illustrate the potential of the hydrophone technique to be used as a standalone method that could ensure high spatial and temporal resolution measurements for sediment transport in rivers.

Impact of the Test Device on the Behavior of the Acoustic Emission Signals: Contribution of the Numerical Modeling to Signal Processing

In a context of nuclear safety experiment monitoring with the non destructive testing method of acoustic emission, we study the impact of the test device on the interpretation of the recorded physical signals by using spectral finite element modeling. The numerical results are validated by comparison with real acoustic emission data obtained from previous experiments. The results show that several parameters can have significant impacts on acoustic wave propagation and then on the interpretation of the physical signals. The potential position of the source mechanism, the positions of the receivers and the nature of the coolant fluid have to be taken into account in the definition a pre-processing strategy of the real acoustic emission signals. In order to show the relevance of such an approach, we use the results to propose an optimization of the positions of the acoustic emission sensors in order to reduce the estimation bias of the time-delay and then improve the localization of the source mechanisms.

Geophysical inversion by dictionary learning

Dictionary learning, a form of unsupervised machine learning, has recently been applied to ocean sound speed profile (SSP) data to obtain compact dictionaries of shape functions which explain SSPs using as few as one non-zero coefficient. In this presentation, the results of this analysis and potential applications of dictionary learning techniques to the inversion of real acoustic data are discussed. The estimation of true geophysical parameters from acoustic observations often is an ill-conditioned problem that is regularized by enforcing prior constraints such as sparsity or energy penalities, and by reducing the size of the parameter search. Traditionally, empirical orthogonal functions (EOFs) and overcomplete wavelet and curvelet dictionaries have been used to represent complex geophysical structures with few parameters. Using the K-SVD dictionary learning algorithm, the representation of ocean SSP data is significantly compressed relative to EOF analysis. The regularization performance of these learned dictionaries is evaluated against EOFs in the estimation of ocean sound speed structure from ocean acoustic observations and the limitations of such unsupervised methods are considered.

Source localization in underwater waveguides using machine learning

The problem of underwater acoustic source localization is solved using an artificial neural network (ANN) under the machine learning framework. Source localization in a waveguide is posed as a classification problem by training a feed-forward neural network (FFN) with one hidden layer. In this paper, the acoustic pressure signals received by a vertical linear array are preprocessed by constructing the normalized cross-spectral density matrices (CSDM), which are used as input of the FFN. Each unit of the output layer represents one possible range (Here, for simplicity, source range is the only parameter that needs to be determined). Different from model-based localization methods such as matched field processing (MFP), ANN as a data-driven method can learn features from real acoustic data, thereby bypassing the sound propagation modeling step completely. Simulations show that FFN achieves a good performance in determining the source ranges even with deficient training data samples and low signal-to-noise ratios (SNRs). The validity of FFN in source localization is further demonstrated with vertical array data from Noise09 experiment where the ship is located successfully using FFN.

Bayesian State-Space Modelling of Conventional Acoustic Tracking Provides Accurate Descriptors of Home Range Behavior in a Small-Bodied Coastal Fish Species.

State-space models (SSM) are increasingly applied in studies involving biotelemetry-generated positional data because they are able to estimate movement parameters from positions that are unobserved or have been observed with non-negligible observational error. Popular telemetry systems in marine coastal fish consist of arrays of omnidirectional acoustic receivers, which generate a multivariate time-series of detection events across the tracking period. Here we report a novel Bayesian fitting of a SSM application that couples mechanistic movement properties within a home range (a specific case of random walk weighted by an Ornstein-Uhlenbeck process) with a model of observational error typical for data obtained from acoustic receiver arrays. We explored the performance and accuracy of the approach through simulation modelling and extensive sensitivity analyses of the effects of various configurations of movement properties and time-steps among positions. Model results show an accurate and unbiased estimation of the movement parameters, and in most cases the simulated movement parameters were properly retrieved. Only in extreme situations (when fast swimming speeds are combined with pooling the number of detections over long time-steps) the model produced some bias that needs to be accounted for in field applications. Our method was subsequently applied to real acoustic tracking data collected from a small marine coastal fish species, the pearly razorfish, Xyrichtys novacula. The Bayesian SSM we present here constitutes an alternative for those used to the Bayesian way of reasoning. Our Bayesian SSM can be easily adapted and generalized to any species, thereby allowing studies in freely roaming animals on the ecological and evolutionary consequences of home ranges and territory establishment, both in fishes and in other taxa.