Extraneous Noise Research Articles

Use of Vocal Signatures for the Inventory of Free‐flying Neotropical Bats1

ABSTRACTAn ongoing study is being conducted to test the efficacy of the Anabat II detector and analysis system in obtaining reliable vocal signatures for the identification of non–phyllostomid species of bats. We sampled a wide range of elevations and associated habitat types throughout Belize. Anabat provides an instantaneous output of echolocation call structure with a laptop computer. Select sequences can be saved directly to the hard drive, avoiding extraneous noise and sound distortion commonly associated with tape recorders. To date, 18 of the 37 species known or expected to occur in the study region were identified by recognizable differences in the time–frequency characteristics of echolocation calls. In general, each family is recognizable by call structure patterns and species readily separated by frequency range parameters. Species that commute or forage at high altitudes are not susceptible to capture but are conspicuous by acoustic sampling. Further work is needed to determine limitations of the equipment, establish better sampling procedures, and develop a comprehensive library of vocal signatures incorporating the range of variation inherent in each species. As this work progresses, we predict the addition of hitherto unknown species occurring within the study region.

Separation of an acoustic signal from noise via the use of point pressure and velocity measurements

The separation of an acoustic plane wave signal from noise via the use of pressure and velocity measurements at a point is investigated using a time domain and Fourier transform approach. The noise is assumed to originate from a compact source at a known location. A time domain analysis of the separation method for the deterministic signal problem is first presented. A spectral analysis which provides additional insight into the method is then presented. An analysis of the sensitivity of the signal separation method to processor errors including quantization errors and extraneous uncorrelated sensor noise is also presented. Numerical results for some simple examples illustrate the use of a digital signal processor for separating signal from narrow-band and broadband noise. The sensitivity of the signal separation method to processor errors including numerical differentiation and quantization errors is simply illustrated via the examples. In addition, the sensitivity of the separation method to the noise source/sensor separation distance is also presented.

A New Method in Acoustic Health Monitoring in Mechanical Seals©

Mechanical seals are a vital component of rotating equipment in the process and power generation industries. Due to the mechanical seal's wide range of application, mechanical seal failures constitute a large proportion of the overall mechanical failures resulting in unscheduled machinery shutdown and even catastrophic consequences. This paper outlines the development of an acoustic measurement procedure in detecting damage in face seals in a centrifugal pump. The acoustic measurements described allowed the elimination of extraneous background noise from nearbly operating machinery and other mechanical components without losing the main acoustic characteristics of the damaged seal. The background noise elimination significantly improved the accuracy of the analysis for the detection and identification of wear/damage in mechanical seals. The results are demonstrated on a set of damaged seals in a ANSI centrifugal pump in a laboratory environment. Presented at the 53rd Annual Meeting in Detroit, Michigan May 17–21, 1998

Directional microphone array for measurement in an anechoic wind tunnel

An anechoic wind tunnel in which full-scale acoustic measurements may be conducted on automobiles at speeds up to 200 km/h has been constructed. The wind tunnel uses an open jet which is collected and returned in the closed-loop wind tunnel. This design allows for microphones to be positioned outside of the air flow typically some 4 to 5 m from the vehicle. Thus problems associated with extraneous wind noise on the microphones are largely eliminated but this has resulted in the need to develop microphone arrays which have directional characteristics in order to identify noise sources on the vehicle. A wide range of microphone configurations was examined recognizing the various constraints imposed by cost, data collection facilities, frequency range, and the geometry of the wind tunnel and vehicle. The design that produced an optimum solution was based on an arced array of 10 microphones. The output from each was summed having first been weighted. The array could be moved along the vehicle. Data will be presented indicating the characteristics of this and other arrays as well as the cost of construction.

Experimental study on the accuracy of sound power determination of S. I. scanning method

In ISO/TC43/SC1/WG25, the precision method of sound power determination by sound intensity scanning method is being drafted as the third standard of ISO 9614 series. For this drafting work, some experimental studies were performed on the measurement accuracy, the methods for checking the measurement sound field (field indicators), and practical procedures of signal processing during the scanning. The sound power determination was made for artificial sound sources (a reference source and a sound source system consisted of two loudspeakers) and actual noise sources (a vacuum cleaner and an air compressor) by changing such measurement conditions as the scanning density, the scanning speed, and the strength of the extraneous noise. From the experimental results, the validities of the specifications in the draft are examined.

The use of near infrared reflectance spectroscopy (NIRS) on undried samples of grass silage to predict chemical composition and digestibility parameters

Near infrared reflectance spectroscopy (NIRS) has become increasingly used as a rapid, accurate method of evaluating some chemical constituents in cereals and dried animal forages. However relatively few studies have been carried out to evaluate the potential of NIRS to predict chemical constituents and digestibility parameters in undried forages. The predictive accuracy of NIRS relies heavily upon obtaining a calibration set which represents the variation in the main population, accurate laboratory analyses and the application of the best mathematical procedures. This study was undertaken to examine the potential of NIRS to accurately determine the chemical composition and a range of digestibility parameters of undried grass silage with the objective of characterising the feeding value of forage to the ruminant animal. A representative population of 136 grass silages covering a wide distribution in chemical and digestibility parameters formed the database for this investigation. A comprehensive chemical profile of the silages was determined and a range of digestibility parameters was measured through 72 wether sheep. Undried finely chopped silage samples were scanned at 2 nm intervals over the wavelength range 400–2500 nm and the optical data recorded as log 1/Reflectance (log 1/R). The spectral data were regressed against a range of chemical and digestibility parameters using modified partial least squares (MPLS) multivariate analysis in conjunction with first and second order derivatization, with and without three scatter correction procedures to reduce the effect of extraneous noise. Cross validation was used to avoid overfitting of the equations. The optimum calibrations were selected on the basis of minimizing the standard error of cross validation (SECV). The SD/SECV (standard deviation of the constituent data/standard error of cross validation) ratio was also calculated to evaluate the calibrations independent of their units (ratios>2.5 are acceptable for quality screening). In the case of the digestibility data a sub-calibration set of 90 samples and a validation set of 46 samples were selected on the basis of their spectral variation, and the optimum mathematical treatment for the 136 was applied. The results of this study show that NIRS predicted the main chemical parameters with a very high degree of accuracy (i.e., the correlation coefficient of cross validation ( R cv 2) ranged from 0.81–0.96) in all cases, except for buffering capacity which had a R cv 2 0.68 and a SD/SECV ratio of 2.19. The nitrogen fractions all formed calibrations with R cv 2 ranging from 0.87 to 0.94 for amino acid nitrogen and kjeldhal nitrogen, respectively and all SD/SECV ratios were greater than 2.5. The alcohol and volatile fatty acid calibrations produced slightly poorer calibrations with butyric and lactic acid and ethanol, giving the highest correlations. The in vivo digestibility parameters were predicted with considerable accuracy R cv 2 0.79–0.85 ( n=136) and the standard errors of prediction (SEP) for the validation set compared favourably with the SECV statistic for calibrations based on the 136 samples. This study found that NIRS calibrations based on undried grass silage spectra have the capability of predicting a wide range of chemical and digestibility parameters which would afford a rapid assessment of the forage for diet formulation.

24] Signal processing techniques for channel current analysis based on hidden markov models

Publisher Summary The patch-clamp technique can be used to record single-channel activity from a small patch of cell membrane. Although the noise from a small patch is much less than that from a whole-cell membrane, signals of interest are often obscured by the noise. Even if the signal frequently emerges from the noise, low-amplitude events such as small subconductance states can remain below the noise level and there may be little evidence of their presence. It is desirable therefore to have a method to measure and characterize not only relatively large ionic currents but also much smaller current fluctuations that are obscured by noise. Extracting the real signal from a limited set of imperfect measurements is a problem that commonly occurs in science, and techniques have been developed to overcome this difficulty. Following digitization, a single-channel record consists of a sequence of data points. Any movement between successive data points can be due to the signal of interest, extraneous noise, or both. The challenge is to remove the noise leaving the biological signal untouched. Some of the methods that have been used in this chapter to do this are linear filtering, nonlinear filtering, and transition detectors.

Crime automatic speaker verification and identification system

A new automatic speaker verification process was presented at the Eighth Annual International Conference on Signal Processing Applications and Technology (ICSPAT ’97), San Francisco, CA (I. I. Gorban, Robust Speaker Verification Algorithms). On the basis of these algorithms the crime automatic speaker verification and identification (CASVI) system has been created. The main feature of the system is that it is possible to verify and identify a speaker when the speech records are very corrupted and/or contain extraneous noise. Testing conducted under different conditions has shown that the CASVI system works well when the level of frequency distortions is lower than 35 dB and the level of the signal-to-noise ratio is lower than 12 dB. These results have been obtained for a new modification of the algorithms in which the adaptive threshold verification process is included and the identification algorithms are added. A description of the CASVI system, it’s algorithms, the methodology, and the results of the system testing will also be presented at the meeting.

Measuring Fluxes in Tidal Estuaries: Sensitivity to Instrumentation and Associated Data Analyses

This study examines the feasibility of quantifying net fluxes across a macrotidal estuary employing a range of available instrumentation. The observational programme in the Mersey Estuary used the following array of instruments to measure tidal currents: bottom mounted and towed ADCPs (acoustic Doppler current profiler); bottom and (floating) platform- mounted electromagnetic current meters; and moored rotary meters. In 1992, observations extending over a 15-day spring-neap tidal cycle included near-continuous half-hourly towed ADCP transects across the 1·5km wide Mersey Narrows. The ADCPs also recorded acoustic backscatter, used here to indicate concentrations of suspended particulate matter (SPM).Tidal analyses of data of this type are complicated by the effect of the change in cross-sectional area between low and high waters. Here, an appropriate vertical reference frame is suggested that optimizes the tidal analyses (especially where drying occurs) and allows both the comparison of moored and floating instruments, and subsequent synthesis into cross-sectionally integrated parameters. Formulation of a related cross-sectional grid scheme for analyses of the towed ADCP data is also described.After careful adoption of the above frameworks, coherent horizontal and vertical distributions were identified from analyses of the towed ADCP data for amplitude, phase and direction of N2, M2, S2, M4, MSf and Z0tidal ellipses. Encouraging comparisons were found between tidal current constituents derived from the various instrument arrays. Similarly, coherent cross-sectional distributions for both tidal and residual fluxes of water and SPM were obtained.Residual current time series recorded at fixed positions showed good cross-correlation, but limited correlation against local wind data. By contrast, the residual time series for the principal mode representing the towed ADCP current data was uncorrelated with these fixed position data. It is possible that extraneous noise inherent in the towed instruments is not effectively filtered in the tidal analysis, and contaminates the residual current time series. Although the spatial distributions of the various estimates of Z0for the net water flux are coherent, the values (mean over 15 days) are clearly unrepresentative. Thus, despite the scale of the observational programme, it is concluded that where net fluxes are less than a few percent of tidal oscillations, these can not be determined by direct measurements of the sort described, even with state-of-the-art instrumentation. This conclusion essentially confirms experience elsewhere. Alternative methods of determining net fluxes include interpretation of measurements (of the present cross-sectional kind) with the aid of numerical models, or for ‘conservative’ riverine tracers, calculation from longitudinal profiles.

Speech recognition employing keyword modeling and non-key word modeling

Speaker independent recognition of small vocabularies, spoken over the long distance telephone network, is achieved using two types of models, one type for defined vocabulary words (e.g., collect, calling-card, person, third-number and operator), and one type for extraneous input which ranges from non-speech sounds to groups of non-vocabulary words (e.g. `I want to make a collect call please`). For this type of key word spotting, modifications are made to a connected word speech recognition algorithm based on state-transitional (hidden Markov) models which allow it to recognize words from a pre-defined vocabulary list spoken in an unconstrained fashion. Statistical models of both the actual vocabulary words and the extraneous speech and background noises are created. A syntax-driven connected word recognition system is then used to find the best sequence of extraneous input and vocabulary word models for matching the actual input speech.

Turbulence and separation induced pressure fluctuations on a finite circular cylinder — application of a linear unsteady strip theory

Abstract In a turbulent stream, the fluctuating pressures on bluff bodies are mainly caused by the oncoming turbulence. Further unsteadiness is engendered by the wake turbulence and by the vortex shedding. Although the gross properties of the flow around a body and in particular the separated flow are affected by the intensity and the scale of the incident turbulence, these sources of pressure fluctuations are thought to be statistically independent. On this basis, the physical system of a turbulent flow around a body is approximated by an artificial black box model. This model corresponds to a constant parameter linear system with multiple stationary random inputs, i.e. turbulent velocity fluctuations far upstream the body. The model output is equal to an instantaneous surface pressure, which is subdivided into a turbulence induced contribution and an extraneous noise term to account for the residual fluctuations. By means of velocity/pressure cross spectral densities (i) significant velocity fluctuations, (ii) the associated admittance functions and (iii) the conditioned and residual pressure spectra are identified experimentally for the flow around a surface mounted circular cylinder in an atmospheric boundary layer flow. In a first approximation, the longitudinal u′(t) and lateral velocity fluctuation ν′(t) of a single stream filament are thought to be the only decisive inputs to the multivariate system (linear unsteady strip theory). The latter induces maximum spectral amplitudes in the mid-frequency range, whereas the longitudinal component achieves maximum effectiveness at the lowest frequencies. However, significant deviations between the (u, ν)-conditioned and the factual pressure spectrum remain. Depending on the circumferential angle and in a certain frequency range, vortex shedding contributes up to 45% to the total pressure RMS-values.

Sound power determination using sound intensity: A review of the technique and the ISO 9614/1 standard

The sound power emitted by a source with a highly directional behavior was determined using sound intensity measurements. The parameters varied were: number of measurement points, degree of reactivity of the measurement room, level of extraneous noise, shape of the measurement surface, and distance from source to measurement surface. Both the ISO field indicators and ISO criteria were examined. The relation between the changes in the ISO indicators and criteria and the accuracy obtained was studied. It was found that, for most of the cases, the requirements due to the values taken by ISO criteria did not have a correlation with the accuracy obtained in the measurement, especially those values of ISO criterion 2 (related to the statistical distribution of measured intensity). The statistical nature of this criterion is not the best way to indicate the accuracy of a sound power determination. Also, criterion 2 fails when there is an extraneous noise source. The influence of environment reactivity was found to be a limitation on the accuracy of the technique. The results obtained coincide with those obtained by various researchers around the world. [Work supported by the National Center of Metrology, México, and the Acoustics Laboratory of IRC-NRC, Canada.]

KIDS: A database of children’s speech

A database of children reading age- and reading-level-appropriate text aloud has been collected. This labeled data, to be distributed in the near future, was primarily intended to be used in CMU’s LISTEN tutor, which employs speech recognition to monitor children’s reading and then helps correct errors. The speaker population was therefore chosen to represent good and poor readers and to incorporate dialects of the speakers for whom the reading coach is intended. Phonemic balance could not be achieved (although it has been calculated) since the primary concern in recording children reading is to present sentences that can effectively be read by first through third graders. The text is a series of sentences that was adapted from text in the Weekly Reader series—most of the adaptation concerned the lack of the accompanying images. The text was chosen for its intrinsic interest and widespread use. Several trial recording sessions were used to develop a protocol that kept extraneous noises produced by the children at a minimum. This and other problems inherent in recording children reading will be discussed. Novel techniques developed for labeling this kind of speech will also be presented. [This work was funded by NSF Grant No. IRI-9528984.]

A Rhythmic Bias in Preverbal Speech Segmentation

Four studies using a variant of the conditioned head turning procedure, in which response latencies to extraneous noises occurring at different junctures within synthetic syllable strings served as the dependent variable, investigated 6- and 9-month-old infants’ representations of familiar and novel syllable pairs manifesting diverse rhythmic patterns.Familiarbisyllables tended to be perceived similarly by both age groups: These bisyllables were perceived as being cohesive, without regard to whether they manifested trochaic (longer–shorter) or iambic (shorter–longer) rhythm.Novelbisyllables were perceived differently by the two age groups. Six-month-olds appeared to perceive segmentally novel, but rhythmically familiar, bisyllables as cohesive, whereas they failed to perceive rhythmically novel, but segmentally familiar bisyllables in the same fashion. Similar patterns of results were obtained with trochaic and iambic bisyllables. Nine-month-olds, however, were differentially sensitive to different rhythmic patterns. Older infants appeared to perceive novel bisyllables as cohesive only when they manifested trochaic rhythm, whether the bisyllables were segmentally or rhythmically novel. Nine-month-olds’ behavior is consistent with the possibility that they have adopted a metrical strategy for segmentation, as A. Cutler (1990 and elsewhere) has argued is the case for novice and expert English speakers alike. The plausibility of such a strategy with respect to the nature of child-directed speech is discussed, as are possible consequences for acquisition of such a strategy.

Advances in feedforward neural networks: demystifying knowledge acquiring black boxes

We survey research of recent years on the supervised training of feedforward neural networks. The goal is to expose how the networks work, how to engineer them so they can learn data with less extraneous noise, how to train them efficiently, and how to assure that the training is valid. The scope covers gradient descent and polynomial line search, from backpropagation through conjugate gradients and quasi Newton methods. There is a consensus among researchers that adaptive step gains (learning rates) can stabilize and accelerate convergence and that a good starting weight set improves both the training speed and the learning quality. The training problem includes both the design of a network function and the fitting of the function to a set of input and output data points by computing a set of coefficient weights. The form of the function can be adjusted by adjoining new neurons and pruning existing ones and setting other parameters such as biases and exponential rates. Our exposition reveals several useful results that are readily implementable.

Automatic microphone mixer

First Page

A Modified Gauss‐Newton Method for Aquifer Parameter Identification

AbstractThe accuracy of parameter estimation procedures is evaluated for a modified Gauss‐Newton method applied to transient ground‐water flow. Three different approaches of evaluating the sensitivity coefficient matrix are examined, including influence coefficient, sensitivity equation, and variational approaches. The performance of each of the techniques is evaluated by applying a common synthetic data set. The latter two techniques are shown to perform with least sensitivity to starting parameters and extraneous sampling noise. Where either random or systematic noise is added to the time‐series data set, the resulting predictions become increasingly more sensitive to the form of the starting transmissivity vector. It is concluded that the modified Gauss‐Newton method is attractive because of its simplicity, high rate of convergence, and modest computational demands, especially when the number of the parameters to be identified is not large.

Corporate Incentives for Hedging and Hedge Accounting

This article explores the information effect of financial risk management. Financial hedging improves the informativeness of corporate earnings as a signal of management ability and project quality by eliminating extraneous noise. Managerial and shareholder incentives regarding information transmission may differ, however, leading to conflicts regarding an optimal hedging policy. We show that these incentives depend on the accounting information made available by the firm. Under some circumstances, if hedge transactions are not disclosed (i.e., firms report only aggregate earnings), managers hedge to achieve greater risk reduction than they would if full disclosure were required. In these cases, it is optimal for shareholders to request only aggregate accounting reports. Article published by Oxford University Press on behalf of the Society for Financial Studies in its journal, The Review of Financial Studies.

Discrete wavelet analysis of blade vortex interaction noise

The discrete wavelet transform, implemented in a multirate filter bank scheme using conjugate quadrature filters, is used to analyze blade vortex interaction (BVI) noise from helicopter rotors in different environments. The discrete wavelet transform successfully isolates the bursty, nonstationary nature of this impulsive noise source. The sub-band coding scheme is then used for signal reconstruction of only the portion of helicopter noise related to blade vortex interaction by filtering in the frequency and time domains. By eliminating extraneous noise sources, the wavelet analysis can produce a more direct comparison between BVI noise generated in flight and in other environments such as a wind tunnel.

Toward automatic segmentation and syllabification of dysarthric speech

A program has been developed which uses heuristic processing of the output of a neural network to locate syllable boundaries in connected speech. The neural network attempts a broad segmental (8 category) classification. This is followed by a program which locates syllable boundaries using phonological constraints and the pattern of network classifications over time. On test data consisting of 150 sentences (about 1800 syllables) from the TIMIT database, this program correctly identifies between 83% and 88% of the syllable boundaries (depending on the error criteria), including syllables bounded by sonorant consonants and syllables containing syllabic consonants. This program is now being applied to the identification of syllable boundaries in speech from talkers with cerebral palsy. In preliminary analyses of 20 sentences from ten dysarthric talkers (140 syllables) only 12 boundary omissions occurred. However, since dysarthric speech often contains hesitations, false starts, and extraneous noises, assessing boundary insertion errors is more difficult; about 50 boundary insertions occurred. Roughly 2/3 of these represented instances in which the talker inserted additional syllabic features into the speech. Further studies will examine methods for reducing the number of these ‘‘unintended’’ boundaries identified by the algorithm. [Work supported by the Nemours Foundation.]