Search

A class of metamaterials that exhibit mass-momentum coupling are known as Willis materials. The mass-momentum coupling arises from hidden degrees of freedom that are asymmetrical with respect to the overall material. Resonant Willis metamaterials incorporate resonators, e.g., a Helmholtz resonator, along with material asymmetry to effectively obtain Willis coupling over a narrow frequency band. By incorporating fractal geometry into the resonator shapes Willis coupling can be obtained over a broader frequency range. This is accomplished by self-similar resonator shapes, which translates into resonance frequencies that exhibit a power-law dependence. Furthermore, manipulation of the internal asymmetry allows for tuning of the Willis coupling. This study reports on the measurement and modeling of resonant Willis metamaterials based on fractal geometry. Measurements are obtained with a transmission impedance tube, and predictions based on a lumped-element model. With resonator volume shapes based on the Sierpiński triangle, Willis coupling is obtained over a frequency range of 500 to 1500 Hz. Furthermore, for a subset of the resonators, Willis coupling is tuned by inverting the asymmetric location of the resonator necks.

Rotorcraft emit unique acoustic signatures tied primarily to their blade pass frequencies. In this study, we examine the ability of artificial neural network machine learning methods to successfully classify several distinct types of rotorcraft based on their low-frequency acoustic signatures. Time series, power spectral density, and Welch power spectral density input types are examined, corresponding to convolutional neural networks for the first input type, and multi-layer perceptrons for the latter two input types. A K-fold cross-validation is used to initially assess the generalization performance of the various models, along with confusion matrices for a deeper dive into the model performance. Ablation studies are discussed for insights into model behavior. These insights are used to showcase further enhanced neural network models with increased robustness and accuracy.

Supersonic projectiles in air generate acoustical signatures that are fundamentally related to the projectile’s shape, size, and velocity. These characteristics influence various mechanisms involved in the generation, propagation, decay, and coalescence of acoustic waves. To understand the relationships between projectile shape, size, velocity, and the physical mechanisms involved, an experimental effort captured the acoustic field produced by a range of supersonic projectiles using both conventional pressure sensors and a schlieren imaging system. The results of this ongoing project will elucidate those fundamental mechanisms, enabling more sophisticated tools for detection, classification, localization, and tracking. This paper details the experimental setup, data collection, and preliminary analysis of a series of ballistic projectiles, both idealized and currently in use by the U.S. Military.

Monitoring noise levels over large areas is typically limited by the number of sensor nodes. While previous studies have been able to accurately estimate noise levels using densely distributed sensors in confined spaces, such as along roads in urban areas, estimating noise levels with relatively few sensors spaced over a large area remains a challenging problem. Point sampling of noise levels due to single noise events is also limited by the models used to estimate between sensor locations and their inherent assumptions. To address these limitations we propose nonlinear, adaptive, and robust tools to estimate levels between sensor nodes based on machine learning. Random forests, support vector machines, and Gaussian processes are explored along with conventional geostatistical methods such as ordinary kriging. These methods are trained and evaluated on data from both measurement and simulation of blast noise on military testing and training ranges. The performance of the methods is evaluated using cross validation and root-mean-square-error.

Domestic military installations generate high levels of noise due to testing and training which leads to annoyance and complaints from surrounding communities. This necessitates continuous noise monitoring to provide decision makers with the information they need to proactively manage their noise environment. Due to the diverse climates in which military testing and training are conducted (e.g., desert, tundra, and rainforest), monitoring equipment that can operate in a variety of environmental conditions with minimal maintenance and low power consumption is needed. Using existing technologies as a baseline, various iterations of a low-cost acoustic monitor were designed to meet these constraints while minimizing initial investment cost, improving the mean time between failures, and increasing overall system capability. This paper will describe the system developed to provide a rapid deployment option that is robust to extreme temperatures, humidity, and destructive wildlife. A review of operational logs collected during multiple deployments was used to evaluate system performance against benchtop and off-the-shelf solutions. This data demonstrate the reliability of the monitoring stations and the sustainability of their hardware.Domestic military installations generate high levels of noise due to testing and training which leads to annoyance and complaints from surrounding communities. This necessitates continuous noise monitoring to provide decision makers with the information they need to proactively manage their noise environment. Due to the diverse climates in which military testing and training are conducted (e.g., desert, tundra, and rainforest), monitoring equipment that can operate in a variety of environmental conditions with minimal maintenance and low power consumption is needed. Using existing technologies as a baseline, various iterations of a low-cost acoustic monitor were designed to meet these constraints while minimizing initial investment cost, improving the mean time between failures, and increasing overall system capability. This paper will describe the system developed to provide a rapid deployment option that is robust to extreme temperatures, humidity, and destructive wildlife. A review of operational logs ...

Continuously recording noise monitoring stations provide feedback of the noise environment at monitor locations. While this feedback is useful, it only provides information at a few point locations and, in many cases, it is of interest to know the noise level(s) at locations between and beyond noise monitoring locations. In this study, we test several machine learning models (e.g., random forests, support vector machines, and neural networks) for their ability to predict noise levels using a small number of noise monitors. Unlike the hybrid geostatistical-acoustical method proposed by Nykaza [Nykaza, Ph.D. thesis, 2013] which requires knowledge of the source location, the methods proposed in this study can be trained to estimate noise levels without knowledge of the source location. The accuracy of each model is assessed with the root-mean-square-error, and we discuss the practical implications of implementing such approaches in real-time systems.

A microphone’s history of calibration is a useful source of information, as an independent analysis of the data can reveal more than simply the metrological traceability. The U.S. Army Engineer Research and Development Center maintains over 13 years of calibration records from an inventory of over 280 Class 1 microphones, which were examined for historical validity and consistency. Analyzing trends in the laboratory reported sensitivities showed that 12.9% of active microphones exhibited four possible trends that could indicate potential microphone failure: sensitivity drift, sensitivity drift followed by instability, jump in sensitivity, and general instability. Many of the microphones that displayed these trends had uncertainties greater than a ± 1.0 dB tolerance inferred from ANSI S1.14 Class 1 microphone specifications, and an additional 10.7% of microphones showed errors in the calibration laboratory’s transcription process or testing procedures. Recommendations are put forth for best practices relat...

Classification of noise sources based on their acoustic signatures is becoming increasingly necessary due to the increasing prevalence of noise monitoring systems. The intractable amount of data that these systems capture necessitates robust and efficient classification algorithms. However, the process of building a classifier requires manually labeling recorded noise events, which is a time consuming task. In this paper we explore a method for reducing the human burden and making the process more efficient called active learning, or “listener-in-the-loop.” This method iteratively improves classifier performance by querying a human listener using optimally chosen observations based on certainty factors, which measure the degree of belief or disbelief based on probability for a particular classification. Classifier performance as well as algorithm efficiency, in terms of computational costs and amount of data required, will be discussed.

Classification algorithms are an essential component of continuously running environmental noise monitors. Without them, one does not know which noise sources are responsible for the levels recorded by the monitor. This is problematic given that continuously recording monitors may accumulate millions of triggered events and terabytes of data. In this study, we look at the utility of Bayesian classification methods. We compare the performance of these methods to some of the top performing environmental noise classifiers (e.g., support vector machines, random forest, and bagged trees), and discuss the advantages and disadvantages of the Bayesian approach. In particular, we compare the accuracy, number of observations needed to achieve an accurate classification, computation time, and feature importance.

Modern automated acoustic classifiers have been shown to perform remarkably well with human speech recognition and music genre classification. These problems are well defined; there is a deep understanding of the source signal, and the required robustness of the model can be decreased without significantly sacrificing accuracy. Unfortunately, this simplification creates models that are insufficient when tasked with classifying environmental noise, which is inherently more variable and difficult to constrain. To further close the gap between human and computer recognition, we must find feature extraction techniques that address the additional set of complexities involved with environmental noise. In this paper, we will explore sophisticated feature extraction techniques (e.g., convolutional auto-encoders and scattering networks), and discuss their effect when applied to acoustic classification.

With the advent of reliable and continuously operating noise monitoring systems, we are now faced with an unprecedented amount of noise monitor data. In the context of environmental noise monitoring, there is a need to automatically detect, separate, and classify all environmental noise sources. This is a complex task because sources can overlap, vary by location, and have an unbounded number of noise sources that a monitor device may record. In this study, we synthetically generate datasets that contain Gaussian noise and overlaps for several pre-labeled environmental noise monitoring datasets to examine how well deep learning methods (e.g., autoencoders) can separate environmental noise sources. In addition to examining performance, we also focus on understanding which signal features and separation metrics are useful to this problem.

With the increasing availability of low-cost and portable noise monitors, the amount of captured data can reasonably be predicted to grow exponentially. The financial and computational expense of portable data storage, processing hardware, and wireless or cellular upload bandwidth, however, necessitate data compression techniques to ensure feasible continuous operation. Sample rate reduction causes a loss of high frequency components while compressed formats (such as mp3) remove parts of the signal that are inaudible to humans, but may be important to statistical learning algorithms. The degree to which these data reduction techniques degrade the accuracy of metrics calculated from the signals (e.g. peak level, equivalent level, sound exposure level, etc.) and the performance of statistical learning classifiers that use the metrics and/or band levels is examined. Over 20,000 recordings from noise monitors located on military installations are systematically down-sampled and compressed to use in the analys...

One of the most difficult problems in noise control is the attenuation of low frequency noise. Typical solutions require barriers with high density and/or thickness. Membrane-type acoustic metamaterials are a novel type of engineered material capable of high low-frequency transmission loss despite their small thickness and light weight. These materials are ideally suited to applications with strict size and weight limitations such as aircraft, automobiles, and buildings. This paper reviews computationally efficient dynamic models for designing membrane-type acoustic metamaterials that have been proposed by the authors, based on the impedance-mobility approach. The computational efficiency of the impedance-mobility models compared to finite element methods enables implementation in design tools based on a graphical user interface and in optimization schemes. Genetic algorithms are used to optimize the unit cell design for a variety of noise reduction goals, including maximizing transmission loss for broadb...

Deep learning is becoming ubiquitous; it is the underlying and driving force behind many heavily embedded technologies in society (e.g., search engines, fraud detection warning systems, and social-media facial recognition algorithms). Over the past few years there has been a steady increase in the number of audio related applications of deep learning. Recently, Nykaza et al. presented a pedagogical approach to understanding how the hidden layers recreate, separate, and classify environmental noise signals. That work presented some feature extraction examples using simple pure tone, chord, and environmental noise datasets. In this paper, we build upon this recent analysis and expand the datasets to include more realistic representations of those datasets with the inclusion of noise and overlapping signals. Additionally, we consider other related architectures (e.g., variant-autoencoders, recurrent neural networks, and fixing hidden nodes/layers), explore their advantages/drawbacks, and provide insights on ...

Environmental noise can cause sleep disturbance, annoyance, complaints, and quite possibly adverse health effects. This is true for continuous noise sources such as steady road traffic noise, impulsive noise sources such as blasts or sonic booms, or sources that fall in-between such as intermittent train and aircraft noise. One way to manage environmental noise is to use noise-monitoring technology to provide both the noise-producers and noise-experiencers feedback on the actual noise environment. Traditional noise-monitoring systems, however, only provide this information at a few locations resulting in an incomplete picture of the noise environment over the entire regions of interest. In this paper, we discuss a framework for providing real-time feedback of the noise environment over a large area (e.g., 100 km2). We show all the steps that are needed to convert the raw noise-monitor data into noise maps and noise impact maps to help manage environmental noise. We discuss the complexity of the problem an...

Deep learning is becoming ubiquitous; it is the underlying and driving force behind many technologies we use everyday (e.g., search engines, fraud detection warning systems, and social-media facial recognition algorithms). Over the past few years, there has been a steady increase in the number of audio and acoustics related applications of deep learning. But what is exactly going on under the hood? In this paper, we focus on deep learning algorithms for unsupervised feature learning. We take a pedagogical approach to understanding how the hidden layers recreate, separate, and classify audio signals. We begin with a simple pure tone dataset, and systematically increase the complexity of this dataset in both frequency and time. We end the presentation with some feature extraction examples from real-world environmental recordings, and find that these features are easier to interpret given the understanding developed from the simpler tone datasets. The unsupervised feature learning techniques explored in this paper include: restricted Boltzmann machines (RBMs) and auto-encoders (AEs).

Unattended and continuously running environmental noise monitoring systems can capture an intractable amount of data. The signals captured can include a multitude of sources (e.g., wind noise and anthropogenic noise sources) in addition to the environmental noise sources of interest (e.g., aircraft, vehicles, trains, and military weapons). In this presentation, we explore the use of machine-learning methods to effectively isolate and identify environmental noise sources captured on such a noise monitoring system. Specifically, we consider the use of both unsupervised (e.g., principle components analysis, clustering methods, and deep belief networks) and supervised (e.g., logistic regression, support vector machines, and neural networks) pattern-learning methods to derive the features of interest and classify the signals based on the obtained features. The generalization performance of each method is assessed using a dataset of over 120,000 human classified signals, and the strengths and weaknesses of each approach are discussed.

Membrane-type acoustic metamaterials have been proven to exhibit high low-frequency transmission loss despite their small thickness and light weight. To date, analysis has focused primarily on experimental studies in plane-wave tubes and numerical modeling using finite element methods. These methods are inefficient when used for applications that require iterative changes to the structure of the material. In addition, high sound transmission loss with a single layer of such metamaterial only occurs in a narrow frequency range. To facilitate design and optimization of stacked membrane-type acoustic metamaterials, a computationally efficient dynamic model based on the impedance-mobility approach is proposed. Results are verified against a finite element model. Single and double layer transmission loss characteristics are compared. Wide-band high-transmission-loss acoustic metamaterials can be achieved by double layer membranes and using the proposed approach for optimization. The impedance-mobility approach is shown to be very efficient for modeling and optimization of such materials, compared against the conventional finite element approach.

NWU’s McDonald Theatre Auditorium is used for both musical and non-musical performances. The acoustics of the space were analyzed in order to determine whether the space could be modified to better fit its uses. The acoustic characteristics of the room were obtained from impulse responses using the methods established in ISO 3382-1 for measuring the acoustic parameters of a performance space. A total of 22 source/receiver pairs were used. The results indicate a need for increased reverberation in the mid to high frequency ranges of 500–8000 Hz. The experimental results were used to calibrate a virtual model of the space in ODEON acoustics software. Materials in the model were then successfully modified to increase reverberation time and eliminate unwanted flutter echoes to optimize the acoustics to better suit the intended purposes of the space.

Persons interested in pursuing graduate study in architectural acoustics are encouraged to consider joining the Architectural Engineering Program within the Durham School of Architectural Engineering and Construction at the University of Nebraska—Lincoln (UNL). Among the 21 ABET-accredited Architectural Engineering (AE) programs across the United States, the Durham School’s program is one of the few that offers graduate engineering degree programs (MAE, MS, and PhD) and one of only two that offers an area of concentration in architectural acoustics. Acoustics students in the Durham School benefit both from the multidisciplinary environment in an AE program and from our particularly strong ties to the building industry, since three of the largest architectural engineering companies in the United States are headquartered in Omaha, Nebraska. Descriptions will be given on the graduate-level acoustics courses, newly renovated acoustic lab facilities, the research interests and achievements of our acoustics faculty and students, and where our graduates are to date. Our group is also active in extracurricular activities, particularly through the University of Nebraska Acoustical Society of America Student Chapter. More information on the “Nebraska Acoustics Group” at the Durham School may be found online at http://nebraskaacousticsgroup.org/.

Sound power is a useful quantity in describing the strength of an acoustic source because its value is independent of distance. However, many standardized methods exist for the measurement of sound power and comparison between methods can give rise to discrepancies. An interlaboratory study was designed according to the ISO 5725 series to quantify the bias and reproducibility of three common sound power measurement methods in the HVACR industry: free field method, diffuse field method, and sound intensity method. A loudspeaker sound source was used to generate two test signals: a broadband signal with decreasing 5 dB slope per octave band, and the same broadband signal with discrete frequency tones at 58, 120, 300, and 600 Hz. The objective of the study is to quantify repeatability, reproducibility, laboratory bias, and measurement method bias, as well as investigate the influence of tones. The design of the interlaboratory study and preliminary results will be presented. The ISO 5725 methods used to inve...

Acoustic metrics are commonly expressed as single numbers in classroom acoustical designs, often neglecting the physical quantity's uncertainty due to the non-diffuse sound field in the seating area. A database of measured monaural and binaural room impulse responses (RIR) was previously gathered from a fully furnished mock-up classroom. Different wall and ceiling absorption configurations were used to alter the mid-frequency reverberation times (RT) in five scenarios between 0.4 and 1.1 s. The middle three RT scenarios were additionally created from two different material configurations. For each material configuration (eight in total), two furniture orientations were utilized. RIRs were measured at 9 to 10 receiver positions for each material/furniture configuration to document the spatial variation in the resulting sound field. Diffuseness has been calculated for each receiver position utilizing the measured RIRs by following Hanyu's (2013) method using normalized decay-canceled impulse responses. Variations in diffuseness and in the assorted acoustic metrics calculated from the measured RIRs are investigated across different receiver positions. These acoustic metrics, pertinent to classroom acoustical designs, include RT, speech transmission index, clarity, and interaural cross-correlation. Means to quantify uncertainty in these metrics due to spatial variation in the non-diffuse sound field will be discussed.

Measured absorption coefficients have some uncertainty that can be expressed in terms of the uncertainty associated with each measured input quantity. The uncertainties of input quantities contribute to the combined uncertainty of the absorption coefficients in varying degrees dependent on the underlying relationship between each parameter. In this study, the propagation of uncertainty of the two-microphone cross-spectral method for measuring normal-incidence absorption coefficients is analyzed according to the ISO/IEC Guide 98-3:2008 “Guide to the expression of uncertainty in measurement.” The results of an experimental investigation are explored to determine the chief sources of systematic error and the relationship between uncertainty of input quantities and uncertainty of intermediate calculations.

This investigation sought to measure the just noticeable difference (JND) for reverberation time (T30) using a rigorous psychophysical method, namely the transformed up-down adaptive method. In ISO 3382-1:2009, the JND for reverberation metrics is taken to be 5%, based on work by Seraphim (1958); however, others have suggested that the actual JND is higher. In this project, sound samples with varying T30 were auralized from impulse responses simulated in a realistically modeled performance space using ODEON. The model’s absorption coefficients were uniformly varied across all surfaces and frequencies to achieve the desired T30s. Three reference reverberation times were utilized (one, two, and three seconds), and eight T30 cases spaced at 4% intervals both above and below each of the three reference T30s were created. Auralizations using a 500 ms white noise burst were presented in a computer-based testing program running a three-interval one-up two-down forced choice method, presented in a sound booth over headphones with flat frequency response. The program randomly interleaved six staircase sequences, three of which ascended and three of which descended towards each reference T30. Results averaged across 30 participants will be presented. [Work supported by a UNL UCARE Grant and the ASA Robert W. Young Award.]

An inadequate educational system can be crippling to a developing country struggling to advance in the modern world. Poor classroom acoustics are a significant factor that can contribute to an inadequate education. Classroom acoustics is a mature field of research in the developed world; however, the classroom conditions and resources found in the developing world present new research challenges. A special project course was developed at Purdue University to give undergraduate students the opportunity to learn about and address classroom acoustics problems in underdeveloped regions. The goal of this special project course was to develop an evaluation and design manual that can be distributed to teachers and educational officials in those regions. The purpose of the manual is to help identify acoustic problems in classrooms, provide suggestions of simple noise control solutions, and outline best practices for the design of new classrooms. The manual specifically addresses problems related to excessive reverberation, poor signal‐to‐noise ratios, and noise transmission from outside the classroom. The ultimate intent of this project was to increase awareness of the unique classroom acoustic conditions in the developing world.