Acoustic Beamforming Research Articles

Towards mobile gaze-directed beamforming: a novel neuro-technology for hearing loss.

Contemporary hearing aids are markedlylimited in their most important role: improving speech perception in dynamic "cocktail party" environments with multiple, competing talkers. Here we describe an open-source, mobile assistive hearing platform entitled "Cochlearity" which uses eye gaze to guide an acoustic beamformer, so a listener will hear best wherever they look. Cochlearity runs on Android and its eight-channel microphone array can be worn comfortably on the head, e.g. mounted on eyeglasses. In this preliminary report, we examine the efficacy of both a static (delay-and-sum) and an adaptive (MVDR) beamformer in the task of separating an "attended" voice from an "unattended" voice in a two-talker scenario. We show that the different beamformers have the potential to complement each other to improve target speech SNR (signal to noise ratio), across the range of speech power, with tolerably low latency.

Design of microphone phased arrays for acoustic beamforming

The design of a microphone array affects the performance of beamforming algorithms in the localization and evaluation of acoustic noise sources. This paper addresses a circular phased array design for near-field aeroacoustic measurements in a closed-test-section wind tunnel. Microphones were distributed in rings and occupied an equal aperture area—each ring could have a different number of microphones. The array performance was evaluated through the dynamic range and array resolution given by the beamwidth. The array designed for the exploration of the novel approach had 112 microphones and 950-mm aperture. In comparison with classical designs, also optimized for the same number of microphones and aperture, the approach provided the best array resolution and a high dynamic range level almost uniform over the frequency range of interest (800 to 20,000 Hz). Microphone shading was also assessed for improving the array performance, and the employment of only the outermost microphones (the innermost ones were shaded) reduced approximately 40% the array beamwidth.

Aerodynamic noise reduction using dual-jet planar air curtains

In this paper, dual-jet planar air curtains are demonstrated to be able to successfully remove aerodynamic noise radiated from tandem rods in a crossflow. Single-jet air curtains have shown significant promise as a low noise technology but can introduce additional noise sources such as lip and mixing noise. In this work, dual-jet air curtains are shown to address these obstacles, achieving the same shielding height with a significantly lower overall system noise. Providing high amplitude tonal and broadband noise, tandem rods are chosen as the test case and the fluid mechanics and acoustics associated with their shielding by air curtains are examined both experimentally and numerically. Particle image velocimetry and flow visualisation methods allow the flow fields to be examined, and a numerical analysis with computational fluid dynamics to be validated. Barely examined in the literature to date, the trajectory of planar jets in crossflow are studied here and compared to a basic semi-empirical model. The accurate identification of the trajectory allows a method for an optimum shielding height and dual jet velocity combination to be proposed and tested. Acoustic beamforming and 1/3 octave band analysis are performed for diagnostic noise source localisation and comparative evaluation. It is demonstrated that a judicious choice of flow parameters can result in the complete removal of the aerodynamic noise with almost no acoustic penalty.

Acoustic Characteristics of the Cabin of the Research Platform on the Airbag IL-PRC-600m

Abstract Hovercrafts are a universal means of transport intended for use on flat surfaces such as water, ice, snow, swamp, or sand. They are used in rescue operations and patrolling difficult areas inaccessible to other means of transport. The Institute of Aviation conducted acoustic measurements inside the cabin of the hovercraft to determine the source of the noise and the sound pressure exerted on the pilot and passengers. Assessment of the sources of noise in the cabin is made using the acoustic beamforming method. Assessment of the level of noise to which a pilot is exposed during the operation was prepared on the basis of a standard specifying the requirements and methods of determining occupational noise exposure [1]. The test results indicate a significant penetration of noise from the drivetrain into the cabin. It is recommended that a hovercraft pilot and operators use hearing protection in some specified conditions and during testing. Thus it is pointed out in the summary that additional soundproofing of the cabin is needed. The points of the greatest penetration of noise into the interior have been indicated.

On the influence of inlet elbow radius on recirculating backflow, whoosh noise and efficiency in turbocharger compressors

While the influence of inlet geometry on turbocharger compressor behaviour has usually been investigated in terms of performance, surge margin and efficiency, data is scarce regarding the impact of the inlet flow field onto the noise emission. In many applications where tight packaging is required, a 90° elbow is placed just upstream of the compressor inducer. This can create a distortion of the incoming flow that affects the turbocharger operation; a distortion that is related to the radius of the elbow. In this experimental investigation three 90° elbow inlets are tested, measuring the in-duct sound intensity through acoustic beamforming, the spectral signature of the noise, and the distortion of the high temperature backflow typical of partially stalled conditions by means of thermocouple arrays. Results show that a tighter elbow radius not only impacts efficiency but also increases inlet noise at conditions close to surge. Spectral analysis shows that this increase is mainly produced in the form of a medium frequency broadband noise usually known as ‘whoosh’ in the literature. On the other hand, effect on the outlet is less noticeable. Measurements of the recirculated backflow distortion in terms of circumferential skewness show good correlation with whoosh noise increase, indicating that flow distortion caused by tighter elbows at marginal surge conditions facilitates the transmission of whoosh noise oscillations to the inlet duct, worsening the acoustic behaviour of the system.

Drivetrain Noise of The Gyroplane I-28

Abstract Gyroplanes, as ultralight aircraft, are popular transport vehicles recently. Ultralight aircraft flights take place at a low altitude – their noise is not without effect on people and nature. The localization of the sources of noise and a possibility to decrease the noise of an gyroplane are described in this paper. The rules of design and exploitation of gyroplanes do not define the limits of emitted noise. Gyroplanes are not noisy aircraft vehicles but for their silencing the knowledge about the sources and frequency range of noise is necessary. The goal of the conducted measurement was to determine the gyro-plane noise properties and the noise measurement methods. The evaluation of the noise sources was made by acoustic beamforming and the directional emission with single microphones at various engine speeds. The supplement of these tests should be the rotor noise measurement but that investigation should be performed on a special stand, on which the rotor propulsion noise would not disturb the measurement.

Evaluating the effectiveness of subarrays in beamforming computations of complex jet noise source environments

Conventional acoustic beamforming methods, while powerful for localizing acoustic sources, require an uncorrelated monopole assumption, depend on the input array location, and are unreliable in a complex source environment, such as jet engine noise. The complexities arise from the large, partially correlated, extended source region, the multiple types of noise sources, such as directional and omnidirectional, and partially correlated sources that violate the integral uncorrelated monopole assumption of beamforming. The complexity of these sources requires advanced beamforming methods. The aim of this research is to determine what advanced beamforming methods tell us about complex, full-scale jet noise sources. To learn about these sources, an array spanning the entire length of a jet from an F-35 aircraft is split into multiple subarrays. Application of the beamforming algorithm to a subarray in the maximum sound region localizes the directional sources present in jet noise; a subarray to the side of the engine characterizes more omnidirectional sources; and a subarray farther forward yields the approximate origin of broadband shock-associated noise. Using these advanced beamforming methods, combined with the subarray processing, opens new insights in this complex environment. [Work supported by the Air Force Research Laboratory; data courtesy of F-35 JPO.]Conventional acoustic beamforming methods, while powerful for localizing acoustic sources, require an uncorrelated monopole assumption, depend on the input array location, and are unreliable in a complex source environment, such as jet engine noise. The complexities arise from the large, partially correlated, extended source region, the multiple types of noise sources, such as directional and omnidirectional, and partially correlated sources that violate the integral uncorrelated monopole assumption of beamforming. The complexity of these sources requires advanced beamforming methods. The aim of this research is to determine what advanced beamforming methods tell us about complex, full-scale jet noise sources. To learn about these sources, an array spanning the entire length of a jet from an F-35 aircraft is split into multiple subarrays. Application of the beamforming algorithm to a subarray in the maximum sound region localizes the directional sources present in jet noise; a subarray to the side of the ...

Robust data-independent object-related transfer function-based beamformer design for remote listening applications

Virtual reality offers users to experience remote acoustic scenes which are captured in the original environment by a microphone array carried by, e.g., a third person or a robot. To extract and preserve the spatiotemporal nature of specific sounds of interest in the original acoustic environments, data-independent acoustic beamforming appears to be a suitable technique. In this work, we present recent work on robust data-independent beamformer design which is applicable to unconstrained sensor topologies and allows for an intuitive control of the beamformer’s robustness. Combined with the concept of polynomial beamforming, flexible beam steering in real time is possible. Consequently, the beamformer design is well suited for remote listening applications. If the beamformer design is applied to a microphone array, which is integrated into a scatterer, object-related transfer functions (ORTFs) need to be incorporated into the beamformer design. We briefly discuss selected approaches to obtaining the required ORTFs in this work. Finally, we present exemplary results of our robust beamformer design, carried out for a 12-element microphone array integrated into the head of a humanoid robot.

A visually guided beamformer to aid listening in complex acoustic environments

The purpose of this paper is to provide an overview of recent work on the visually guided hearing aid (VGHA; Kidd et al. JASA 133, EL202). The latest prototype VGHA consists of a lightweight, head-mounted, 18-microphone array combined with PC-based software to create a highly-tuned acoustic beamformer steered in real time by an integrated binocular eye tracker. Recent work on the two main functional sub-components of the VGHA (beamforming and eye-gaze control) focused on the benefits that may be obtained by listeners with hearing loss (and in some cases by listeners with normal hearing) attempting to understand speech masked by one or more competing sounds (such as Gaussian noise or additional speakers). These benefits include improved signal-to-noise ratio under different masked conditions and faster steering (compared to fixed directional amplification steered by head turns). The VGHA may be simulated over headphones by PC-based implementations using HRTFs and an optional external eye-tracker. This has allowed testing a number of variations of beamformer properties and the evaluation of multiple types of sound processing strategies including the incorporation of natural binaural cues. Overall, the VGHA holds considerable promise for improving selective listening in a variety of complex listening conditions.

Spherical beamforming for spherical array with impedance surface

Spherical microphone array beamforming has been a popular research topic for recent years. Due to their isotropic beam in three dimensional spaces as well as a certain frequency range, the arrays are widely used in many applications such as sound field recording, acoustic beamforming, and noise source localisation. The body of a spherical array is usually considered perfectly rigid. A sound field captured by the sensors on spherical array can be decomposed into a series of spherical harmonics. In noise source localisation, the amplitude density of sound sources is estimated and illustrated by mean of colour maps. In this work, a rigid spherical array covered by fibrous materials is studied via numerical simulation and the performance of the spherical beamforming is discussed.

An ASIC implementation of a Dual Channel Acoustic Beamforming for MEMS microphone in 0.18㎛ CMOS technology

An ASIC implementation of a Dual Channel Acoustic Beamforming for MEMS microphone in 0.18㎛ CMOS technology

A Wearable Transcranial Doppler Ultrasound Phased Array System.

Practical deficiencies related to conventional transcranial Doppler (TCD) sonography have restricted its use and applicability. This work seeks to mitigate several such constraints through the development of a wearable, electronically steered TCD velocimetry system, which enables noninvasive measurement of cerebral blood flow velocity (CBFV) for monitoring applications with limited operator interaction. A highly-compact, discrete prototype system was designed and experimentally validated through flow phantom and preliminary human subject testing. The prototype system incorporates a custom two-dimensional transducer array and multi-channel transceiver electronics, thereby facilitating acoustic beamformation via phased array operation. Electronic steering of acoustic energy enables algorithmic system controls to map Doppler power throughout the tissue volume of interest and localize regions of maximal flow. Multi-focal reception permits dynamic vessel position tracking and simultaneous flow velocimetry over the time-course of monitoring. Experimental flow phantom testing yielded high correlation with concurrent flowmeter recordings across the expected range of physiological flow velocities. Doppler power mapping has been validated in both flow phantom and preliminary human subject testing, resulting in average vessel location mapping times <14 s. Dynamic vessel tracking has been realized in both flow phantom and preliminary human subject testing. A wearable prototype CBFV measurement system capable of autonomous vessel search and tracking has been presented. Although flow phantom and preliminary human validation show promise, further human subject testing is necessary to compare velocimetry data against existing commercial TCD systems. Additional human subject testing must also verify acceptable vessel search and tracking performance under a variety of subject populations and motion dynamics-such as head movement and ambulation.

Improving acoustic beamforming maps in a reverberant environment by modifying the cross-correlation matrix

A method to improve the quality of acoustic beamforming in reverberant environments is proposed in this paper. The processing is based on a filtering of the cross-correlation matrix of the microphone signals obtained using a microphone array. The main advantage of the proposed method is that it does not require information about the geometry of the reverberant environment and thus it can be applied to any configuration. The method is applied to the particular example of aeroacoustic testing in a hard-walled low-speed wind tunnel; however, the technique can be used in any reverberant environment. Two test cases demonstrate the technique. The first uses a speaker placed in the hard-walled working section with no wind tunnel flow. In the second test case, an airfoil is placed in a flow and acoustic beamforming maps are obtained. The acoustic maps have been improved, as the reflections observed in the conventional maps have been removed after application of the proposed method.

Enhancing Auditory Selective Attention Using a Visually Guided Hearing Aid.

Listeners with hearing loss, as well as many listeners with clinically normal hearing, often experience great difficulty segregating talkers in a multiple-talker sound field and selectively attending to the desired "target" talker while ignoring the speech from unwanted "masker" talkers and other sources of sound. This listening situation forms the classic "cocktail party problem" described by Cherry (1953) that has received a great deal of study over the past few decades. In this article, a new approach to improving sound source segregation and enhancing auditory selective attention is described. The conceptual design, current implementation, and results obtained to date are reviewed and discussed in this article. This approach, embodied in a prototype "visually guided hearing aid" (VGHA) currently used for research, employs acoustic beamforming steered by eye gaze as a means for improving the ability of listeners to segregate and attend to one sound source in the presence of competing sound sources. The results from several studies demonstrate that listeners with normal hearing are able to use an attention-based "spatial filter" operating primarily on binaural cues to selectively attend to one source among competing spatially distributed sources. Furthermore, listeners with sensorineural hearing loss generally are less able to use this spatial filter as effectively as are listeners with normal hearing especially in conditions high in "informational masking." The VGHA enhances auditory spatial attention for speech-on-speech masking and improves signal-to-noise ratio for conditions high in "energetic masking." Visual steering of the beamformer supports the coordinated actions of vision and audition in selective attention and facilitates following sound source transitions in complex listening situations. Both listeners with normal hearing and with sensorineural hearing loss may benefit from the acoustic beamforming implemented by the VGHA, especially for nearby sources in less reverberant sound fields. Moreover, guiding the beam using eye gaze can be an effective means of sound source enhancement for listening conditions where the target source changes frequently over time as often occurs during turn-taking in a conversation. http://cred.pubs.asha.org/article.aspx?articleid=2601621.

Average acoustic beamforming in car cabins: An automatic system for acoustic mapping over 3D surfaces

This paper describes a microphone array-based measurement systems for mapping noise sources in car cabins.The system exploits the average beamforming concept, a measuring/processing method that makes it possible to significantly reduce ghost sources caused by the presence of reflective surfaces and to improve the source localization task in closed environments.The average beamforming procedure consists in sampling a stationary acoustic field moving a microphone array of known geometry in different locations, so to perform a statistical processing out of the beamforming results obtained from each measurement position.Since the calculation is based on the relative distance between each microphone and the calculation grid, the identification of the array position inside the measurement environment is of primary importance. In an interior noise source localization application, which is quite typical of the automotive sector, this translates in the necessity of an accurate knowledge of both the microphone array locations and the cabin internal geometry, since the calculation grid is typically made to coincide with the cabin internal surfaces.The solution proposed in this paper involves a coordinate measurement system, based on the Microsoft Kinect (first release 2010) device, to measure both microphone array positions and cabin surface in the same reference frame. A calibration procedure was specifically performed on the Kinect system in order to transform this compact and cheap device in a coordinate measuring system.The performance of the whole strategy are discussed on the basis of tests performed in a real car environment in which a speaker was exploited as target source. Results provided demonstrate the effectiveness of the proposed method.

Adaptive acoustic energy delivery to near and far fields using foldable, tessellated star transducers

Methods of guiding acoustic energy arbitrarily through space have long relied on digital controls to meet performance needs. Yet, more recent attention to adaptive structures with unique spatial configurations has motivated mechanical signal processing (MSP) concepts that may not be subjected to the same functional and performance limitations as digital acoustic beamforming counterparts. The periodicity of repeatable structural reconfiguration enabled by origami-inspired tessellated architectures turns attention to foldable platforms as frameworks for MSP development. This research harnesses principles of MSP to study a tessellated, star-shaped acoustic transducer constituent that provides on-demand control of acoustic energy guiding via folding-induced shape reconfiguration. An analytical framework is established to probe the roles of mechanical and acoustic geometry on the far field directivity and near field focusing of sound energy. Following validation by experiments and verification by simulations, parametric studies are undertaken to uncover relations between constituent topology and acoustic energy delivery to arbitrary points in the free field. The adaptations enabled by folding of the star-shaped transducer reveal capability for restricting sound energy to angular regions in the far field while also introducing means to modulate sound energy by three orders-of-magnitude to locations near to the transducer surface. In addition, the modeling philosophy devised here provides a valuable approach to solve general sound radiation problems for foldable, tessellated acoustic transducer constituents of arbitrary geometry.

Online MVDR Beamformer Based on Complex Gaussian Mixture Model With Spatial Prior for Noise Robust ASR

This paper considers acoustic beamforming for noise robust automatic speech recognition. A beamformer attenuates background noise by enhancing sound components coming from a direction specified by a steering vector. Hence, accurate steering vector estimation is paramount for successful noise reduction. Recently, time-frequency masking has been proposed to estimate the steering vectors that are used for a beamformer. In particular, we have developed a new form of this approach, which uses a speech spectral model based on a complex Gaussian mixture model CGMM to estimate the time-frequency masks needed for steering vector estimation, and extended the CGMM-based beamformer to an online speech enhancement scenario. Our previous experiments showed that the proposed CGMM-based approach outperforms a recently proposed mask estimator based on a Watson mixture model and the baseline speech enhancement system of the CHiME-3 challenge. This paper provides additional experimental results for our online processing, which achieves performance comparable to that of batch processing with a suitable block-batch size. This online version reduces the CHiME-3 word error rate WER on the evaluation set from 8.37% to 8.06%. Moreover, in this paper, we introduce a probabilistic prior distribution for a spatial correlation matrix a CGMM parameter, which enables more stable steering vector estimation in the presence of interfering speakers. In practice, the performance of the proposed online beamformer degrades with observations that contain only noise or/and interference because of the failure of the CGMM parameter estimation. The introduced spatial prior enables the target speaker's parameter to avoid overfitting to noise or/and interference. Experimental results show that the spatial prior reduces the WER from 38.4% to 29.2% in a conversation recognition task compared with the CGMM-based approach without the prior, and outperforms a conventional online speech enhancement approach.

Establishment of a Beamforming Dataset on Basic Models of Low-Speed Axial Fan Blade Sections

The paper presents wind tunnel experiments, supplemented with phased array microphone measurements, on 2D basic models of low-speed axial fan blade sections: a flat plate, a cambered plate, and a RAF6-E airfoil. It aims at documenting the establishment of an acoustic beamforming dataset for the three profiles. The phased array microphone measurements offer spatially resolved information on the generated noise. The measurement setup enables the correlation of the streamwise evolution of the blade boundary layer with the associated noise characteristics. The dataset incorporates a wide range of incidence and Reynolds-numbers investigated. The present paper is confined to reporting on experimental results for arbitrarily selected representative incidences, Reynolds numbers, frequency bands, and profiles. The paper outlines a methodology for the evaluation and representation of the beamforming data in the following forms: source strength level based third-octave spectra obtained using background noise subtraction; maps presenting the loci of source strength level maxima; noise source maps for frequency bands of anticipated vortex shedding noise.

Computational Acoustic Beamforming for Noise Source Identification for Small Wind Turbines.

This paper develops a computational acoustic beamforming (CAB) methodology for identification of sources of small wind turbine noise. This methodology is validated using the case of the NACA 0012 airfoil trailing edge noise. For this validation case, the predicted acoustic maps were in excellent conformance with the results of the measurements obtained from the acoustic beamforming experiment. Following this validation study, the CAB methodology was applied to the identification of noise sources generated by a commercial small wind turbine. The simulated acoustic maps revealed that the blade tower interaction and the wind turbine nacelle were the two primary mechanisms for sound generation for this small wind turbine at frequencies between 100 and 630 Hz.

A generic neural acoustic beamforming architecture for robust multi-channel speech processing

Acoustic beamforming can greatly improve the performance of Automatic Speech Recognition(ASR) and speech enhancement systems when multiple channels are available. We recently proposed a way to support the model-based Generalized Eigenvalue beamforming operation with a powerful neural network for spectral mask estimation. The enhancement system has a number of desirable properties. In particular, neither assumptions need to be made about the nature of the acoustic transfer function (e.g., being anechonic), nor does the array configuration need to be known. While the system has been originally developed to enhance speech in noisy environments, we show in this article that it is also effective in suppressing reverberation, thus leading to a generic trainable multi-channel speech enhancement system for robust speech processing. To support this claim, we consider two distinct datasets: The CHiME 3challenge, which features challenging real-world noise distortions, and the Reverbchallenge, which focuses on distortions caused by reverberation. We evaluate the system both with respect to a speech enhancement and a recognition task. For the first task we propose a new way to cope with the distortions introduced by the Generalized Eigenvalue beamformer by renormalizing the target energy for each frequency bin, and measure its effectiveness in terms of the PESQ score. For the latter we feed the enhanced signal to a strong DNN back-end and achieve state-of-the-art ASR results on both datasets. We further experiment with different network architectures for spectral mask estimation: One small feed-forward network with only one hidden layer, one Convolutional Neural Network and one bi-directional Long Short-Term Memory network, showing that even a small network is capable of delivering significant performance improvements.