Search

This work examines the use of generative adversarial networks for reconstructing sound fields from experimental data. It is investigated whether generative models, which learn the underlying statistics of a given signal or process, can improve the spatio-temporal reconstruction of a sound field by extending its bandwidth. The problem is significant as acoustic array processing is naturally band limited by the spatial sampling of the sound field (due to the difficulty to satisfy the Nyquist criterion in space domain at high frequencies). In this study, the reconstruction of spatial room impulse responses in a conventional room is tested based on three different generative adversarial models. The results indicate that the models can improve the reconstruction, mostly by recovering some of the sound field energy that would otherwise be lost at high frequencies. There is an encouraging outlook in the use of statistical learning models to overcome the bandwidth limitations of acoustic sensor arrays. The approach can be of interest in other areas, such as computational acoustics, to alleviate the classical computational burden at high frequencies.

Near-field acoustic holography (NAH) is a very powerful and widely used technique for the study of complex acoustic radiators. NAH enables to quickly understand how a complex source radiates into the medium. The technique is particularly suitable at low frequencies. At high frequencies, a dense transducer interspacing is required, and the measurement microphones can disturb the studied sound field when their size is comparable to the acoustic wavelength. In this study we examine the use of acousto-optic sensing in NAH. Acousto-optic sensing uses light beams as the sensing element, making it possible to acquire remote and non-invasive measurements without introducing extraneous objects in the vicinity of the source. The pressure, particle velocity and intensity fields, as well as the sound power radiated by a complex source, are determined from measurements in the near-field with an optical interferometer. The presented results demonstrate the potential of optical sensing to non-intrusively characterize sound fields, particularly at high frequencies.

This study examines the edge diffraction effect when a sound wave impinges and reflects off finite porous absorbers, flush-mounted in an infinite hard baffle. A theoretical analysis of the diffraction is given by taking a two-dimensional spatial Fourier transform of a plane wave impinging on a finite absorber. Numerical experiments are also presented to simulate the sound field above infinite and finite locally reactive absorbers and the measurement with an array of pressure sensors. In such cases, a regularized solution is used to separate the incident and reflected plane wave components, in the wave-number domain, including both propagating and evanescent waves. The properties of the wave-number spectrum are associated either with the specular reflection or with the diffracted components, caused by the interaction of the sound wave with the finite absorber. From the regularized solution, it is possible to reconstruct the surface impedance and the absorption coefficient of the sample. The influence of Gaussian noise on such measurements is also investigated. The use of propagating and evanescent waves on the sound field model led to an estimation of the absorption coefficient that depends just slightly on the size of the sample, which is a desired feature for in situ measurement methods.

Kevlar-walled aeroacoustic wind tunnels have become important research facilities to evaluate low-noise wind turbine designs. In this paper a semi-analytical Green’s function for the transmission of sound through Kevlar walls and shear layers was developed in order to provide accurate acoustic measurements at high flow speeds. The semi-analytical Green’s function was evaluated experimentally with a monopole sound source. It was possible to find the correct source location when the Green’s function was used to formulate the steering vectors for the beamforming technique. The corrected steering vectors provided a much better source level estimation than if a free-field Green’s function was used to formulated the steering vectors, but there was a systematic overcorrection at low frequencies and an undercorrection at high frequencies. The analytical transmission loss was very sensitive to the convection velocity of the vorticity shed at the membrane apertures, which in turn depended on the aerodynamic loading of the Kevlar walls. It is likely that further fine-tuning of the formula of the convection velocity will reduce the systematic error. The experimental setup to validate the Green’s function could be improved if aerodynamic loading was applied to the walls.

Spatial sound field interpolation relies on suitable models to both conform to available measurements and predict the sound field in the domain of interest. A suitable model can be difficult to determine when the spatial domain of interest is large compared to the wavelength or when spherical and planar wavefronts are present or the sound field is complex, as in the near-field. To span such complex sound fields, the global reconstruction task can be partitioned into local subdomain problems. Previous studies have shown that partitioning approaches rely on sufficient measurements within each domain, due to the higher number of model coefficients. This study proposes a joint analysis of all local subdomains, while enforcing self-similarity between neighbouring partitions. More specifically, the coefficients of local plane wave representations are sought to have spatially smooth magnitudes. A convolutional model of the sound field in terms of plane wave filters is formulated and the inverse reconstruction problem is solved via the alternating direction method of multipliers. The experiments on simulated and measured sound fields suggest, that the proposed method both retains the flexibility of local models to conform to complex sound fields and also preserves the global structure to reconstruct from fewer measurements., Comment: Improve description of figures and figure labels, add code reference

The characterization of Room Impulse Responses (RIR) over an extended region in a room by means of measurements requires dense spatial with many microphones. This can often become intractable and time consuming in practice. Well established reconstruction methods such as plane wave regression show that the sound field in a room can be reconstructed from sparsely distributed measurements. However, these reconstructions usually rely on assuming physical sparsity (i.e. few waves compose the sound field) or trait in the measured sound field, making the models less generalizable and problem specific. In this paper we introduce a method to reconstruct a sound field in an enclosure with the use of a Generative Adversarial Network (GAN), which s new variants of the data distributions that it is trained upon. The goal of the proposed GAN model is to estimate the underlying distribution of plane waves in any source free region, and map these distributions from a stochastic, latent representation. A GAN is trained on a large number of synthesized sound fields represented by a random wave field and then tested on both simulated and real data sets, of lightly damped and reverberant rooms.

Sampling spatio-temporal acoustic fields is a challenging problem since it demands a large number of sensors. Typically, to characterize the pressure field inside an enclosure, the number of measurements required increases linearly with frequency and cubically with volume, becoming an intractable problem for rooms of moderate size even at low and mid frequencies. Sparse representation techniques, such as Compressed Sensing, rely on the sparsity of natural signals in certain representation domain to drastically reduce the number of measurements needed to sample such signals. In this study, we optimize the placement of sensors inside an enclosure in order to reduce the measurements required for a given reconstruction accuracy. The proposed methodology selects a sparse set of sensor positions from predefined grid via the QR factorization of the sensing matrix. Numerical results show an effective reduction in the required number of measurements when their positions are optimized, in contrast to standard random positioning. Unlike the majority of existing approaches, we study the placement problem for wide-band acoustic fields.

This study examines the use of Gaussian process (GP) regression for sound field reconstruction. GPs enable the reconstruction of a sound field from a limited set of observations based on the use of a covariance function (a kernel) that models the spatial correlation between points in the sound field. Significantly, the approach makes it possible to quantify the uncertainty on the reconstruction in a closed form. In this study, the relation between reconstruction based on GPs and classical reconstruction methods based on linear regression is examined from an acoustical perspective. Several kernels are analyzed for their potential in sound field reconstruction, and a hierarchical Bayesian parameterization is introduced, which enables the construction of a plane wave kernel of variable sparsity. The performance of the kernels is numerically studied and compared to classical reconstruction methods based on linear regression. The results demonstrate the benefits of using GPs in sound field analysis. The hierarchical parameterization shows the overall best performance, adequately reconstructing fundamentally different sound fields. The approach appears to be particularly powerful when prior knowledge of the sound field would not be available.

Clinical diagnostic applications of ear-canal reflectance have been researched extensively in the literature, however, the measurement uncertainty associated with the conventional measurement technique using an insert ear probe is unknown in human ear canals. Ear-canal reflectance measured using an ear probe is affected by multiple sources of error, including incorrect estimates of the ear-canal cross-sectional area and oblique ear-probe insertions. In this paper, ear-canal reflectance measurements are reproduced in an occluded-ear simulator and in 54 adult ear canals using two different measurement techniques: a conventional ear probe and a two-microphone probe that enables the separation of reverse- and forward-propagating plane waves. The two-microphone probe is inserted directly into test subjects' ear canals, and the two-microphone method is distinguished by not requiring the ear-canal cross-sectional area to calculate the ear-canal reflectance. The results show a reasonable agreement between the two measurement techniques. The paper further examines the influence of oblique ear-probe insertions and the compensation for such oblique insertions, which results in an improved agreement between the two measurement techniques.

Sound field analysis methods make it possible to characterize and reconstruct a sound field from a limited set of observations. Classical approaches rely on the use of analytical basis functions to model the sound field throughout the observed domain. When the complexity of the sound field is high, for example, in a room at mid and high frequencies, propagating wave representations can be suboptimal due to model discrepancy. We examine the use of local representations to alleviate this model discrepancy and explore data-driven approaches to obtain suitable models. Specifically, local representations are used to reconstruct the sound field over a large spatial aperture in a room. The performance of local models is compared against conventional plane wave reconstructions and the use of data-driven local functions is examined. Dictionary learning and principal component analysis are used to obtain functions from extensive spatial measurements in an empty room. The results indicate that local partitioning models conform to fields of high spatial complexity. Dictionary learning generalizes across different rooms and frequencies-conferring potential for modelling complex sound fields based on their local and statistical properties.

Acousto-optic sensing methods enable the measurement of sound, using the interaction between sound and light as the sensing principle. Specifically, acousto-optic tomography enables the reconstruction of sound fields across space and time. However, existing reconstruction methods have constrained the applicability of the principle to controlled laboratory conditions, due to various measurement restrictions. In the present study, we make use of a new spatial reconstruction approach that makes it possible to reconstruct a sound field from a set of arbitrary projections, enabling the in situ capture of sound fields in conventional spaces, such as rooms and enclosures. We examine how the proposed approach enables to measure a sound field in a room, and how the sensing principle can be used to reconstruct the sound pressure field, particle velocity vector and the flow ofacoustic intensity in a three-dimensional space. Thus, a complete volumetric reconstruction of the sound field is obtained from sets of arbitrary acousto-optic projection measurements.

The Reflections series takes a look back on historical articles from The Journal of the Acoustical Society of America that have had a significant impact on the science and practice of acoustics.

The characterization of sound fields over space is fundamental in acoustics. Yet, the volumetric acquisition of sound fields using conventional electromechanical transducers is challenging due to the high sampling requirements needed in a three-dimensional domain. In this study, we introduce an acousto-optic sensing method to remotely sample acoustic fields in a volume using a sparse set of noninvasive optical measurements. Optical methods that use light as a sensing element have drawn significant attention recently, since they allow for remote, noninvasive sampling of acoustic fields with fine spatial resolution. The practical deployment of optical sensing in acoustics, however, has been hindered by the limited acquisition of data available, and the unsuitability of existing methods to reconstruct sound fields in scarce-data regimes. The proposed methodology relies on projecting the measured data on a basis that satisfies the wave equation, thus alleviating the measurement requirements associated with traditional reconstruction methods. The proposed approach leads to accuracy improvements of 20 dB relative to current methods, and a 90% reduction of measured data. We demonstrate the approach experimentally by reconstructing in situ the three-dimensional sound field inside a room from remote measurements of the acousto-optic interaction. These results are key to advancing the use of acousto-optical sensing methods, which are currently limited to simplified domains. The work presented in this paper brings about the possibility to solve so far intractable problems, such as the volumetric and remote acquisition of complex three-dimensional acoustic fields.

In sound field reproduction and sound field control systems, the acoustic transfer functions between a set of sources and an extended reproduction area need to be accurately estimated in order to achieve good performance. This implies that large amounts of measurements should be performed if the area is large compared to the wavelengths of interest. In this paper, a method for reconstructing these transfer functions in highly damped conditions is proposed by using only a small number of measurements in the reproduction area. The source radiation is modeled with the spherical harmonics basis and its amplitude coefficients are fitted with Bayesian inference. This approach is validated in a sound field control experiment where a set of 12 control loudspeakers attenuate the sound pressure level generated by a set of six primary loudspeakers in a quiet zone while minimizing their radiation into a listening zone. The performance of the approach is studied by analyzing the sound field reconstruction and the sound field control performance. It is shown that it is possible to get-with few measurements and the source radiation model-results similar to those achieved using a dense grid of transfer function measurements.

Sound zones are valuable in scenarios where multiple people are present in the same room but want to listen to individual audio content without wearing headphones. The purpose of sound zone methods is to minimize the acoustic leakage between the zones by controlling multiple loudspeakers. This requires knowledge of how the loudspeakers interact with the room and radiate sound to the zones. That interaction is characterized by the transfer functions between the loudspeakers and microphones sampling the sound field in the zones. In this paper, the effect on the acoustic separation due to inherent noise in in situ transfer function measurements is investigated. The attainable separation is analyzed in the frequency range 20-300 Hz by means of the eigenfunctions of a rectangular room. The concept of observable degrees of freedom is introduced to indicate the number of active eigenfunctions, which are different within the zones at a given frequency. Likewise, controllable degrees of freedom indicate whether each source can excite the active eigenfunctions independently. It is argued that high separation can be achieved when the observable degrees of freedom are fewer than the controllable, and the target sound field can be described by the observable degrees of freedom. However, to attain this high separation it is a requirement that the details in the transfer functions associated with these degrees of freedom can be resolved in the presence of the measurement noise. For both simulated and experimental conditions the transfer functions are estimated using Bayesian inference and the uncertainty in the estimates is used to automatically regularize the sound field control. This regularization is seen to improve the performance when the measurement noise is correlated between the microphones and have little effect when the noise is uncorrelated.

This study examines the estimation of the surface impedance of an absorber with microphone arrays. Two array geometries are compared—a rigid spherical array and a double layer planar array. The impedance is estimated via reconstructing the sound field (pressure and particle velocity) on the absorber's surface, using a plane wave expansion. The comparison is carried out by studying the numerical properties of the two arrays as well as through experimental tests.This study examines the estimation of the surface impedance of an absorber with microphone arrays. Two array geometries are compared—a rigid spherical array and a double layer planar array. The impedance is estimated via reconstructing the sound field (pressure and particle velocity) on the absorber's surface, using a plane wave expansion. The comparison is carried out by studying the numerical properties of the two arrays as well as through experimental tests.

Measured values of acoustic absorption obtained from standardized reverberation-chamber measurements often differ across laboratories. These discrepancies arise due to non-isotropic sound incidence on the absorbing specimen, diffraction at the sample edges, and differences in the chambers' shapes and dimensions. The present study examines an experimental method for characterizing the distribution of sound incidence on the specimen in the steady state. The methodology relies on a plane wave decomposition (i.e., estimation of the wavenumber spectrum) to determine the magnitude of the sound waves arriving from definite directions onto the absorbing sample. Based on this decomposition, the sound pressure, particle velocity, and sound intensity can be reconstructed in the vicinity of the absorbing specimen. One can distinguish between the incident and reflected components of the sound field, making it possible to characterize the incident energy flows. Measurements with a programmable robotic arm are conducted in a reverberation chamber in two damping conditions (empty and with absorption on the floor). The quantitative accuracy of the method is examined via an estimation of the sample's angle-dependent absorption coefficient, showing good agreement with theoretical predictions. It is anticipated that the proposed method will be of value in explaining the deviations encountered across standardized laboratories.

Characterizing audible sound fields over space is at the core of many applications in acoustics and audio technology, including active sound field control, spatial audio, and the experimental analysis of sound radiation. For this purpose, arrays of microphones are commonly deployed across the sound field, with an inter-microphone spacing that depends on the highest frequency studied. Capturing sound fields at mid and high frequencies is nonetheless challenging, as the number of transducers required becomes impractically large and the scattering of sound introduced by the array is often significant. Acousto-optic sensing enables the remote and non-invasive characterization of acoustic fields, and thus it represents an attractive alternative to conventional electro-mechanical transducers in several applications. The phase-shift that laser beams experience as they travel through a pressure field are measured via optical interferometry. The optical measurements are then used to reconstruct the field that originated such phase variations. In this study, we project the measured data into a set of functions suitable to represent sound fields over space. The projection makes it possible to improve the reconstruction accuracy and extrapolate the reconstruction outside the measured region. We demonstrate the method by reconstructing the three-dimensional sound field inside a room using optical data.

Gaussian processes (GP) have been used to predict acoustic fields by interpolating under-sampled field observations. Using GP interpolation to predict fields is advantageous due to its ability to denoise measurements, and for its prediction of likely field outcomes given a certain field coherence, or in GP terminology, a kernel. While there are many design options for a coherence function, in this study we examine using the radial basis function kernel, the physically based plane wave kernel, and a composition of plane wave kernels representing a certain angular interval of directions. The composite kernel is relevant in ocean acoustics where it is often the case that arrivals can only be within a narrow direction of arrival. We demonstrate that an array sampled with spacing larger than a half wavelength can benefit from GP interpolation, giving less root mean squared error.

Sound source localization is crucial for communication and sound scene analysis. This study uses direction-of-arrival estimates of multiple ad hoc distributed microphone arrays to localize sound sources in a room. An affine mapping between the independent array estimates and the source coordinates is derived from a set of calibration points. Experiments show that the affine model is sufficient to locate a source and can be calibrated to physical dimensions. A projection of the local array estimates increases localization accuracy, particularly further away from the calibrated region. Localization tests in three dimensions compare the affine approach to a nonlinear neural network.

Sparse arrays are special geometrical arrangements of sensors which overcome some of the drawbacks associated with dense uniform arrays and require fewer sensors. For direction finding applications, sparse arrays with the same number of sensors can resolve more sources while providing higher resolution than a dense uniform array. This has been verified numerically and with real data for one-dimensional microphone arrays. In this study the use of nested and co-prime arrays is examined with sparse Bayesian learning (SBL), which is a compressive sensing algorithm, for estimating sparse vectors and support. SBL is an iterative parameter estimation method and can process multiple snapshots as well as multiple frequency data within its Bayesian framework. A multi-frequency variant of SBL is proposed, which accounts for non-flat frequency spectra of the sources. Experimental validation of azimuth and elevation [two-dimensional (2D)] direction-of-arrival (DOA)estimation are provided using sparse arrays and real data acquired in an anechoic chamber with a rectangular array. Both co-prime and nested arrays are obtained by sampling this rectangular array. The SBL method is compared with conventional beamforming and multiple signal classification for 2D DOA estimation of experimental data.

The acquisition of the spatio-temporal characteristics of a sound field over a large volume of space is experimentally challenging, as a large number of transducers is required to sample the sound field. Sound field reconstruction methods are a resourceful approach, as they enable the interpolation and extrapolation of the sound field from a limited number of observed data. In this study we examine the spatio-temporal and spatio-spectral reconstruction of the sound field in a room from distributed measurements of the sound pressure. Specifically, a variational Gaussian process regression model is formulated, using time-domain anisotropic kernels to reconstruct the direct sound and early reflections, and frequency-domain isotropic kernels for reconstructing the late reverberant field. The proposed methodology is compared experimentally to classical regression models based on plane wave decompositions, which are widely used in sound field reconstruction in enclosures due to their simplicity and accuracy.

International audience; Reverberation-room measurements of absorption coefficients are subject to high uncertainties and irreproducibility across different laboratories. One reason is the lack of a sufficiently isotropic sound field both in steady-state and during the sound field decay. Recently, two different methods - based on the decomposition of the sound field into plane waves - have been proposed for the estimation and analysis of the sound field isotropy in reverberation rooms [Berzborn, 2018; Gover, 2004; Nolan, 2018]. This study compares the two methods based on numerically calculated plane wave sound fields as well as experimental results obtained from array measurements in a reverberation room in different configurations. The room configurations comprise setups with and without hanging panel diffusers and with and without an absorbing sample on the floor. Both methods are compared regarding their suitability in the analysis of reverberation rooms and the required angular resolution for the plane wave decomposition.

A method of evaluating sound field isotropy in decaying reverberant sound fields is presented. The proposed method extends the experimental framework outlined in [J. Acoust. Soc. Am. 143(4), 2514-2526 (2018)] and analyzes the decaying sound field in a reverberation room. Spatio-temporal measurements of the sound field are obtained, and a wavenumber decomposition is performed as a function of time, which serves to examine the directional properties of the sound field and its angular symmetry. Experimental results are obtained in a reverberation room in four different configurations (the empty room, with an absorber on the floor, with panel diffusers, and without them). The results demonstrate how isotropy tends to increase or decrease as a function of time, depending on the disposition of the diffusing and absorbing elements. Diffusers are found to effectively redirect the energy in the room, although they do not succeed in generating a uniform incidence on the sample. The proposed approach makes it possible to analyze the specific processes occurring in a reverberation chamber and can provide valuable insights in the process of standardization to verify the directional properties found in each reverberation room.

Sound field reconstruction from finite measurement arrays provides a means to interpolate and extrapolate acoustic quantities that describe the field. By assuming a linear projection on a basis that follows a principled source propagation, one can recover accurate estimates of the aforementioned sound fields. However, the recovery of the basis coefficients relies on explicit models of which measurement noise and data incompleteness can profoundly affect the uncertainty of the solution. This work aims to estimate the distribution of the underlying pressure conditioned on the observations of the measured pressure in a room. A framework for approximate inference is adapted for sound field reconstruction by applying generative flow-based models and invertible neural network architectures. In particular, we use conditional normalising flows for fast conditional posterior estimation and uncertainty quantification. The model's evaluation is carried out using experimental data measured with a spherical array and compared to hierarchical Bayes with Markov Chain Monte-Carlo sampling.

The calibration of ear-probe microphones can increase the precision of calibrating stimulus levels in situ and of measuring acoustic responses from the ear. This paper proposes a methodology to measure the sensitivity of an ear-probe microphone, requiring only an acoustic coupler and a calibrated reference microphone. The input impedance of the coupler is measured, enabled by a preliminary acoustic Thevenin calibration of the ear probe, and the plane-wave transfer impedance of the coupler is calculated analytically. Using these two quantities, the pressure transfer function between the reference microphone and the ear-probe microphone is estimated. This enables estimating the sensitivity of the ear-probe microphone. The proposed and an existing method were compared, resulting in substantially similar ear-probe microphone sensitivities. The proposed method is practically feasible in producing reliable measurements of sound pressure in the ear canal and calibrating stimulus levels in a clinical setting.

Capturing the impulse or frequency response functions within extended regions of a room requires an unfeasible number of measurements. In this study, a method to reconstruct the response at arbitrary points based on compressive sensing (CS) is examined. The sound field is expanded into plane waves and their amplitudes are estimated via CS, obtaining a spatially sparse representation of the sound field. The validity of the CS assumptions are discussed, namely, the assumption of the wave field spatial sparsity (which depends strongly on the properties of the specific room), and the coherence of the sensing matrix due to different spatial sampling schemes. An experimental study is presented in order to analyze the accuracy of the reconstruction. Measurements with a scanning robotic arm make it possible to circumvent uncertainty due to positioning and transducer mismatch, and examine the accuracy of the reconstruction over extended regions of space. The results indicate that near perfect reconstructions are possible at low frequencies, even from a limited set of measurements. In addition, the study shows that it is possible to reconstruct damped room responses with reasonable accuracy well into the mid-frequency range.

Sparse reconstruction methods, such as Compressive Sensing, are powerful methods in acoustic array processing, as they make wideband reconstruction possible. However, when addressing sound fields that are not necessarily sparse (e.g., in acoustic near-fields, reflective environments, extended sources, etc.), the methods can lead to a poor reconstruction of the sound field. This study examines the use of sparse analysis priors to promote block-sparse solutions. In particular, a Fused Total Generalized Variation (F-TGV) method is developed, to analyze the sound field in the near-field of acoustic sources. The method promotes sparsity both on the spatial derivatives of the solution and on the solution itself, thus seeking solutions where the non-zero coefficients are grouped together. The performance of the method is examined numerically and experimentally, and compared with established methods. The results indicate that the F-TGV method is suitable to examine both compact and spatially extended sources. The method is promising for its generality, robustness to noise, and the capability to provide a wideband reconstruction of sound fields that are not necessarily sparse.

This study proposes an experimental method for evaluating isotropy in enclosures, based on an analysis of the wavenumber spectrum in the spherical harmonics domain. The wavenumber spectrum, which results from expanding an arbitrary sound field into a plane-wave basis, is used to characterize the spatial properties of the observed sound field. Subsequently, the obtained wavenumber spectrum is expanded into a series of spherical harmonics, and the moments from this spherical expansion are used to characterize the isotropy of the wave field. The analytical framework is presented. The method is examined numerically and experimentally, based on array measurements in four chambers: two anechoic chambers (one with a single source and another with an array of 52 sources), a reverberation chamber, and the same reverberation chamber with a sample of absorbing material on the floor. The results indicate that the proposed methodology is suitable for assessing the isotropy of a sound field.

We present a method for acoustic source localization in reverberant environments based on semi-supervised machine learning (ML) with deep generative models. Source localization in the presence of reverberation remains a major challenge, which recent ML techniques have shown promise in addressing. Despite often large data volumes, the number of labels available for supervised learning in reverberant environments is usually small. In semi-supervised learning, ML systems are trained using many examples with only few labels, with the goal of exploiting the natural structure of the data. We use variational autoencoders (VAEs), which are generative neural networks (NNs) that rely on explicit probabilistic representations, to model the latent distribution of reverberant acoustic data. VAEs consist of an encoder NN, which maps complex input distributions to simpler parametric distributions (e.g., Gaussian), and a decoder NN which approximates the training examples. The VAE is trained to generate the phase of relative transfer functions (RTFs) between two microphones in reverberant environments, in parallel with a DOA classifier, on both labeled and unlabeled RTF samples. The performance this VAE-based approach is compared with conventional and ML-based localization in simulated and real-world scenarios.

International audience; In this paper we propose the use of sparse reconstruction methods to characterize the pass- by noise of rail-bound transportation vehicles. European trains are homologated with the TSI- NOISE legislation, which specifies the quantification of the vehicle noise alone; that is, irrespective of the track on which the vehicle is running. However, some tracks contribute significantly to the pass-by noise in mid-frequencies, leading to geographically dependent noise emissions of the same vehicle. To remove this dependency from the homologation, the track and vehicle contributions must be separated. This paper proposes the use of sparse regularization methods to distinguish between vehicle and rail noise. The approach is based on measurements with a microphone array near the track, and the use of dictionaries of plane waves and moving equivalent sources. Compared with reference results predicted with the TWINS software, we show that both contributions can be separated more accurately than present state-of-the-art methods.

The feasibility and the performance of controlling low frequency sound of loudspeaker systems under varying atmospheric conditions is examined experimentally. In the experiment, a control subwoofer array is canceling the sound of a primary subwoofer array over long distances (∼100 m) and in large areas (∼320 m2) using the pressure-matching method. To avoid the measurement of the sound field over the entire control area, a sound propagation model is introduced that is fitted in situ to model the radiation properties of the loudspeakers and the variation of the speed of sound. The results show that the control system reduces the sound pressure levels by up to 15–20 dB over the subwoofers' frequency range. However, the reduction can vary considerably depending on the specific atmospheric condition. The model-based approach reduces the number of required measurements and achieves similar reduction performance to the control based on direct measurements with considerably fewer microphone locations while also being more robust. Additionally, the sound propagation model enables the reduction of acoustic energy in virtual control zones that are far away from the microphone location. The investigated methodology has a direct application in the mitigation of sound from outdoor concerts.

The acoustic properties of surfaces are commonly evaluated using samples of finite size, which generate edge diffraction effects that are often disregarded. This study makes use of sound scattering theory to characterize such finite samples. In a given sound field, the samples can be described by a unique complex directivity function called the far-field pattern. Numerical results show that the far-field pattern contains extensive information on the tested samples, including sound absorption and surface scattering, as well as scattering due to finiteness. In this paper, a method is introduced to estimate the far-field pattern of a finite sample. The method relies on measurements of the sound pressure and acoustic particle velocity in the near-field of the sample, and it makes use of the Helmholtz integral equation. The proposed technique is examined in an anechoic room where the sound field near the test sample is scanned with a three-dimensional sound intensity probe. The estimated far-field pattern is compared with numerical predictions up to 1 kHz.

Measurements of the ear-canal reflectance using an ear probe require estimating the characteristic impedance of the ear canal in situ. However, an oblique insertion of the ear probe into a uniform waveguide prevents accurately estimating its characteristic impedance using existing time-domain methods. This is caused by the non-uniformity immediately in front of the ear probe when inserted at an oblique angle, resembling a short horn loading, and introduces errors into the ear-canal reflectance. This paper gives an overview of the influence of oblique ear-probe insertions and shows how they can be detected and quantified by estimating the characteristic impedance using multiple truncation frequencies, i.e., limiting the utilized frequency range. Additionally, a method is proposed to compensate for the effects on reflectance of an oblique ear-probe insertion into a uniform waveguide. The incident impedance of the horn loading is estimated, i.e., were the uniform waveguide anechoic, which replaces the characteristic impedance when calculating reflectance. The method can compensate for an oblique ear-probe insertion into a uniform occluded-ear simulator and decrease the dependency of reflectance on insertion depth in an ear canal. However, more research is required to further assess the method in ear canals.

This study examines the spatial and directional properties of net energy flows in a reverberation chamber. Based on measurements with a spherical array, a method is proposed to estimate the flows of acoustic energy in the volume surrounding the array. The proposed method is used to examine the steady state, early decay, and late decay of the sound field in a reverberation room (both empty and with an absorber on the floor). The results show that the approach is successful in characterizing the spatio-spectral and spatio-temporal properties of power flows in reverberant sound fields, constituting a valuable analysis tool.

The reconstruction of sound fields in a room from a limited set of measurements is a central problem in acoustics, with relevant applications in e.g. acoustic analysis, audio, or sound field control. Conventional approaches rely on measuring the room impulse response (RIR) at several locations in the room, and fitting a wave model that enables to estimate the field at other locations—via solving an inverse problem. Previous studies have shown that the reconstruction of RIRs strongly depends on the regularization method used to solve the problem, and that enforcing sparsity is beneficial for the reconstruction of the early part of the RIR. This work studies Gaussian processes with time-dependent regularization in order to exploit the temporal properties of RIRs. The inverse problem is solved in the time domain, where hierarchical Bayes is applied in order to explicitly promote solutions where waves are more likely to propagate as time evolves. The model aims at reconstructing both the early part, and the late part of the RIRs. The performance of the proposed model is studied with experimental measurements, and compared to classical reconstruction methods with l1 and l2-norm regularization schemes

This study examines a near-field acoustic holography method consisting of a sparse formulation of the equivalent source method, based on the compressive sensing (CS) framework. The method, denoted Compressive-Equivalent Source Method (C-ESM), encourages spatially sparse solutions (based on the superposition of few waves) that are accurate when the acoustic sources are spatially localized. The importance of obtaining a non-redundant representation, i.e., a sensing matrix with low column coherence, and the inherent ill-conditioning of near-field reconstruction problems is addressed. Numerical and experimental results on a classical guitar and on a highly reactive dipole-like source are presented. C-ESM is valid beyond the conventional sampling limits, making wide-band reconstruction possible. Spatially extended sources can also be addressed with C-ESM, although in this case the obtained solution does not recover the spatial extent of the source.