Fast acoustic likelihood computation using low-rank matrix approximation

Acoustic modeling using mixtures of multivariate Gaussians is the prevalent approach for many speech processing problems. Computing likelihoods against a large set of Gaussians is required as a part of many speech processing systems and it is the computationally dominant phase for LVCSR systems. We express the likelihood computation as a multiplication of matrices representing augmented feature vectors and Gaussian parameters. The computational gain of this approach over traditional methods is by exploiting the structure of these matrices and efficient implementation of their multiplication.In particular, we explore direct low-rank approximation of the Gaussian parameter matrix and indirect derivation of low-rank factors of the Gaussian parameter matrix by optimum approximation of the likelihood matrix. We show that both the methods lead to similar speedups but the latter leads to far lesser impact on the recognition accuracy. Experiments on a 1138 word vocabulary RM1 task using Sphinx 3.7 system show that, for a typical case the matrix multiplication approach leads to overall speedup of 46%. Both the low-rank approximation methods increase the speedup to around 60%, with the former method increasing the word error rate (WER) from 3.2% to 6.6%, while the latter increases the WER from 3.2% to 3.5%.

Three-dimensional localization of multiple acoustic sources in shallow ocean with non-Gaussian noise

In this paper, a low-complexity algorithm SAGE-USL is presented for 3-dimensional (3-D) localization of multiple acoustic sources in a shallow ocean with non-Gaussian ambient noise, using a vertical and a horizontal linear array of sensors. In the proposed method, noise is modeled as a Gaussian mixture. Initial estimates of the unknown parameters (source coordinates, signal waveforms and noise parameters) are obtained by known/conventional methods, and a generalized expectation maximization algorithm is used to update the initial estimates iteratively. Simulation results indicate that convergence is reached in a small number of (<= 10) iterations. Initialization requires one 2-D search and one 1-D search, and the iterative updates require a sequence of 1-D searches. Therefore the computational complexity of the SAGE-USL algorithm is lower than that of conventional techniques such as 3-D MUSIC by several orders of magnitude. We also derive the Cramer-Rao Bound (CRB) for 3-D localization of multiple sources in a range-independent ocean. Simulation results are presented to show that the root-mean-square localization errors of SAGE-USL are close to the corresponding CRBs and significantly lower than those of 3-D MUSIC. (C) 2014 Elsevier Inc. All rights reserved.

A systematic asymptotic approach to determine the dispersion characteristics of structural-acoustic waveguides with arbitrary fluid loading

Structural-acoustic waveguides of two different geometries are considered: a 2-D rectangular and a circular cylindrical geometry. The objective is to obtain asymptotic expansions of the fluid-structure coupled wavenumbers. The required asymptotic parameters are derived in a systematic way, in contrast to the usual intuitive methods used in such problems. The systematic way involves analyzing the phase change of a wave incident on a single boundary of the waveguide. Then, the coupled wavenumber expansions are derived using these asymptotic parameters. The phase change is also used to qualitatively demarcate the dispersion diagram as dominantly structure-originated, fluid originated or fully coupled. In contrast to intuitively obtained asymptotic parameters, this approach does not involve any restriction on the material and geometry of the structure. The derived closed-form solutions are compared with the numerical solutions and a good match is obtained. (C) 2016 Elsevier Ltd. All rights reserved.

ZnO film based surface acoustic wave micro-pump

In this study, a micro-pump unit based on surface acoustic wave (SAW) on piezoelectric ZnO film is designed and fabricated as a micro-fluidic device. It employs a mechanical wave, which is generated electrically using an aluminum interdigital transducer (IDT), and propagates on the surface of the ZnO film. The ZnO film was used in this study because it has a high electromechanical coefficient and an excellent bonding with various substrate materials, in particular silicon. The sputtering parameters for ZnO film deposition have been optimized, and the ZnO films with different thickness from 1 micron to 5.5 microns were prepared. The film properties have been characterized using different methods, such as scanning electron microscopy, X-ray diffraction and atomic force microscopy. Aluminum IDT with a finger width and spacing of 8 microns was patterned on the ZnO film using a lift-off process. The frequency generated was measured using a network analyzer, and it varies from 130 MHz to 180 MHz as a function of film thickness. A signal generator was used to generate the frequency with a power amplifier to amplify the signal, which was then applied to aluminum IDT to generate the surface acoustic wave. If a liquid droplet exists on the surface carrying the acoustic wave, the energy and the momentum of the SAW will be coupled into the fluid, causing the liquid to vibrate and move on film surface. The strength of this movement is determined by the applied voltage and frequency. The volume of the liquid drop loaded on the SAW device in this study is of several hundreds of nanoliters. The movement of the liquid inside the droplet and also on the ZnO film surface can be demonstrated. The performance of ZnO SAW device was characterized as a function of film thickness. © 2007 IOP Publishing Ltd.

Methods for enhanced electrical transduction and characterization of micromechanical resonators

This paper details the design and enhanced electrical transduction of a bulk acoustic mode resonator fabricated in a commercial foundry MEMS process utilizing 2.5 μm gaps. The I-V characteristics of electrically addressed silicon resonators are often dominated by capacitive parasitics, inherent to hybrid technologies. This paper benchmarks a variety of drive and detection principles for electrostatically driven square-extensional mode resonators operating in air via analytical models accompanied by measurements of fabricated devices with the primary aim of enhancing the ratio of the motional to feedthrough current at nominal operating voltages. In view of ultimately enhancing the motional to feedthrough current ratio, a new detection technique that combines second harmonic capacitive actuation and piezoresistive detection is presented herein. This new method is shown to outperform previously reported methods utilizing voltages as low as ±3 V in air, providing a promising solution for low voltage CMOS-MEMS integration. To elucidate the basis of this improvement in signal output from measured devices, an approximate analytical model for piezoresistive sensing specific to the resonator topology reported here is also developed and presented. © 2010 Elsevier B.V. All rights reserved.

Enhanced transduction methods for electrostatically driven MEMS resonators

Electrically addressed silicon bulk acoustic wave microresonators offer high Q solutions for applications in sensing and signal processing. However, the electrically transduced motional signal is often swamped by parasitic feedthrough in hybrid technologies. With the aim of enhancing the ratio of the motional to feedthrough current at nominal operating voltages, this paper benchmarks a variety of drive and detection principles for electrostatically driven square-extensional mode resonators operating in air and in a foundry MEMS process utilizing 2μm gaps. A new detection technique, combining second harmonic capacitive actuation and piezoresistive detection, outperforms previously reported methods utilizing voltages as low as ± 3V in air providing a promising solution for low voltage CMOS-MEMS integration. ©2009 IEEE.

Acoustic signatures of the seafloor: Tools for predicting grouper habitat

Groupers are important components of commercial and recreational fisheries. Current methods of diver-based grouper census surveys could potentially benefit from development of remotely sensed methods of seabed classification. The goal of the present study was to determine if areas of high grouper abundance have characteristic acoustic signatures. A commercial acoustic seabed mapping system, QTC View Series V, was used to survey an area near Carysfort Reef, Florida Keys. Acoustic data were clustered using QTC IMPACT software, resulting in three main acoustic classes covering 94% of the area surveyed. Diver-based data indicate that one of the acoustic classes corresponded to hard substrate and the other two represented sediment. A new measurement of seabed heterogeneity, designated acoustic variability, was also computed from the acoustic survey data in order to more fully characterize the acoustic response (i.e., the signature) of the seafloor. When compared with diver-based grouper census data, both acoustic classification and acoustic variability were significantly different at sites with and without groupers. Sites with groupers were characterized by hard bottom substrate and high acoustic variability. Thus, the acoustic signature of a site, as measured by acoustic classification or acoustic variability, is a potentially useful tool for stratifying diver sampling effort for grouper census.

Estimating species and size composition of rockfishes to verify targets in acoustic surveys of untrawlable areas

Rockfish (Sebastes spp.) biomass is difficult to assess with standard bottom trawl or acoustic surveys because of their propensity to aggregate near the seafloor in highrelief areas that are inaccessible to sampling by trawling. We compared the ability of a remotely operated vehicle (ROV), a modified bottom trawl, and a stereo drop camera system (SDC) to identify rockfish species and estimate their size composition. The ability to discriminate species was highest for the bottom trawl and lowest for the SDC. Mean lengths and size distributions varied among the gear types, although a larger number of length measurements could be collected with the bottom trawl and SDC than with the ROV. Dusky (S. variabilis), harlequin (S. variegatus), and northern rockfish (S. polyspinis), and Pacific ocean perch (S. alutus) were the species observed in greatest abundance. Only dusky and northern rockfish regularly occurred in trawlable areas, whereas these two species and many more occurred in untrawlable areas. The SDC was able to resolve the height of fish off the seafloor, and some of the rockfish species were observed only near the seafloor in the acoustic dead zone. This finding is important, in that fish found exclusively in the acoustic dead zone cannot be assessed acoustically. For these species, methods such as bottom trawls, long-lines, or optical surveys using line transect or area swept methods will be the only adequate means to estimate the abundance of these fishes. Our results suggest that the selection of appropriate methods for verifying targets will depend on the habitat types and species complexes to be examined.

Combustion rumble prediction with integrated computational-fluid-dynamics/ low-order-model methods

The pressure oscillation within combustion chambers of aeroengines and industrial gas turbines is a major technical challenge to the development of high-performance and low-emission propulsion systems. In this paper, an approach integrating computational fluid dynamics and one-dimensional linear stability analysis is developed to predict the modes of oscillation in a combustor and their frequencies and growth rates. Linear acoustic theory was used to describe the acoustic waves propagating upstream and downstream of the combustion zone, which enables the computational fluid dynamics calculation to be efficiently concentrated on the combustion zone. A combustion oscillation was found to occur with its predicted frequency in agreement with experimental measurements. Furthermore, results from the computational fluid dynamics calculation provide the flame transfer function to describe unsteady heat release rate. Departures from ideal one-dimensional flows are described by shape factors. Combined with this information, low-order models can work out the possible oscillation modes and their initial growth rates. The approach developed here can be used in more general situations for the analysis of combustion oscillations. Copyright © 2012 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

Nonlinear phenomena in thermoacoustics: A comparison between single-mode and multi-mode methods

Nonlinear analysis of thermoacoustic instability is essential for prediction of frequencies and amplitudes of limit cycles. In frequency domain analyses, a quasi-linear transfer function between acoustic velocity and heat release rate perturbations, called the flame describing function (FDF), is obtained from a flame model or experiments. The FDF is a function of the frequency and amplitude of velocity perturbations but only contains the heat release response at the forcing frequency. While the gain and phase of the FDF provide insight into the nonlinear dynamics of the system, the accuracy of its predictions remains to be verified for different types of nonlinearity. In time domain analyses, the governing equations of the fully coupled problem are solved to find the time evolution of the system. One method is to discretize the governing equations using a suitable basis, such as the natural acoustic modes of the system. The number of modes used in the discretization alters the accuracy of the solution. In our previous work we have shown that predictions using the FDF are almost exactly the same as those obtained from the time-domain using only one mode for the discretization. We call this the single-mode method. In this paper we compare results from the single-mode and multi-mode methods, applied to a thermoacoustic system of a premixed flame in a tube. For some cases, the results differ greatly in both amplitude as well as frequency content. This study shows that the contribution from higher and subharmonics to the nonlinear dynamics can be significant and must be considered for an accurate and comprehensive analysis of thermoacoustic systems. Hence multi-mode simulations are necessary, and the single-mode method or the FDF may be insufficient to capture some of the complex nonlinear behaviour in fhermoacoustics.

A six sensor method for measuring acoustic properties in ducts.

An accurate description of sound propagation in a duct is important to obtain the sound power radiating from a source in both near and far fields. A technique has been developed and applied to decompose higher-order modes of sound emitted into a duct. Traditional experiments and theory based on two-sensor methods are limited to the plane-wave contribution to the sound field at low frequency. Due to the increase in independent measurements required, a computational method has been developed to simulate sensitivities of real measurements (e.g., noise) and optimize the set-up. An experimental rig has been constructed to decompose the first two modes using six independent measurements from surface, flush-mounted microphones. Experiments were initially performed using a loudspeaker as the source for validation. Subsequently, the sound emitted by a mixed-flow fan has been investigated and compared to measurements made in accordance with the internationally standardized in-duct fan measurement method. This method utilizes large anechoic terminations and a procedure involving averaging over measurements in space and time to account for the contribution from higher-order modes. The new method does not require either of these added complications and gives detail about the underlying modal content of the emitted sound.

Comparison of stationary acoustic monitoring and visual observation of finless porpoises

The detection performance regarding stationary acoustic monitoring of Yangtze finless porpoises Neophocaena phocaenoides asiaeorientalis was compared to visual observations. Three stereo acoustic data loggers (A-tag) were placed at different locations near the confluence of Poyang Lake and the Yangtze River, China. The presence and number of porpoises were determined acoustically and visually during each 1-min time bin. On average, porpoises were acoustically detected 81.7 +/- 9.7% of the entire effective observation time, while the presence of animals was confirmed visually 12.7 +/- 11.0% of the entire time. Acoustic monitoring indicated areas of high and low porpoise densities that were consistent with visual observations. The direction of porpoise movement was monitored using stereo beams, which agreed with visual observations at all monitoring locations. Acoustic and visual methods could determine group sizes up to five and ten individuals, respectively. While the acoustic monitoring method had the advantage of high detection probability, it tended to underestimate group size due to the limited resolution of sound source bearing angles. The stationary acoustic monitoring method proved to be a practical and useful alternative to visual observations, especially in areas of low porpoise density for long-term monitoring.

Comparison of acoustic velocity perturbation measurements using PIV vs. two-microphone technique

Understanding combustion instabilities requires accurate measurements of the acoustic velocity perturbation into injectors. This is often accomplished via the use of the two microphone technique, as this only requires two pressure transducers. However, measurements of the actual velocities emerging from the injectors are not often taken, leaving questions regarding the assumptions about the acoustic velocity. A comparison of velocity measured at downstream of the injector with that of two-microphone technique can show the accuracy and limitations of two-microphone technique. In this paper, velocity measurements are taken using both particle image velocimetry (PIV) and the two-microphone technique in a high pressure facility designed for aeroengine injector measurements. The flow is excited using an area modulation device installed on the choked end of the combustion chamber, with PIV measurements enabled by optical access downstream of the injector through a quartz tube and windows. Acoustic velocity perturbations at the injector are determined by considering the Fourier transformed pressure fluctuations for two microphones installed at a known distance upstream of the injector. PIV measurements are realized by seeding the air flow with micrometric water particles under 2.5 bar pressure at ambient temperature. Phase locked velocity fields are realized by synchronizing the acquisition of PIV images with the revolution of the acoustic modulator using the pressure signal measured at the face of injector. The mean velocity fluctuation is calculated as the difference between maximum and minimum velocities, normalized by the mean velocity of the unforced case. Those values are compared with the peak-to-peak velocity fluctuation amplitude calculated by the two-microphone technique. Although the ranges of velocity fluctuations for both techniques are similar, the variation of fluctuation with forcing frequencies diverges significantly with frequency. The differences can be attributed to several limitations associated with of both techniques, such as the quality of the signal, the signal/noise ratio, the accuracy of PIV measurements and the assumption of isentropic flow of the particle velocity from the plenum through the injector. We conclude that two-microphone methods can be used as a reference value for the velocity fluctuation in low order applications such as flame transfer functions, but not for drawing conclusions regarding the absolute velocity fluctuations in the injector. Copyright © 2013 by ASME.

Measurement of respiratory acoustic impedance in children by a modified forced oscillation method

An introduction to a modified forced oscillation method, square-wave excitation technique, including fundamentals and methods, as used in respiratory function examination. On the basis of experimental results and theoretical predictions, we suggest that Respiratory Acoustic Impedance (RAI) measurement by spectral analysis can significantly improve estimation of contribution to RAI from different part of respiratory tract. The outcome is of considerable interest in the study of lung disease, such as COPD and asthma in young children.