摘要:

One class of phonetic context effects involves durational contrasts: A signal can be perceived as a stop consonant or a glide, depending on the length of the adjacent segment. For example, the syllable [wa] is perceived as [ba]if the steady‐state frequencies corresponding to the vowel are lengthened. Thus the vocalic information affects reported phonetic judgments of the prior consonantal segment. The neural model presented processes consonant–vowel transitions and steady‐state vowel signals by transient and sustained channels. This model predicts the experimental effects of vowel duration, CV transition rate, and frequency extent in identifying [ba] and [wa]. The product of a leaky integrator’s output with the quotient of two linear expressions in the stimulus parameters generates the model’s output. With a single small set of parameters, the model predicts over 99% of the variance in two data sets [J. L. Miller and A. M. Liberman, Percept. Psychophys.25(6), 457–465 (1979); Schwabetal., Percept. Psychophys.29(2), 121–128 (1981)]. It outperforms a number of simple alternative models. Implications for other phonetic distinctions and physiological analogs to the model in monkey cortex are described. [Work supported by AFOSR and ONR.]

ISSN:0001-4966    

DOI:10.1121/1.419312

出版商:Acoustical Society of America    

年代:1997

数据来源: AIP

摘要:

Vowel reduction as a function of speaking rate within individuals has been well established (Fourakis, 1991; van Bergem, 1995; Weismeretal., 1995). This phenomenon is challenged by the finding that differences exist in the control of speech rate for both slow and fast talker groups (see Tsao, 1995). This study will therefore examine vowel reduction in the speech of individuals separated on the rate continuum (i.e., slow versus average versus fast). The hypothesis that the habitual vowel spaces and the rate‐induced degree of vowel reduction is different for habitually slow versus fast talkers will be explored. Subjects will produce a reading passage, and vowel reduction will be measured based on the target frequencies forF1,F2, andF3 for the four corner vowels (e.g., /a/, /i/, /ae/, /u/). A linear ergression function will be used to demonstrate the relationship between vowel reduction of the 60 subjects (15 slow, 15 fast, and 30 average) and their habitual speech rate along the rate continuum. [Work supported by University of Nebraska—Kearney and UW—Madison Speech Acoustics Laboratory.]

ISSN:0001-4966    

DOI:10.1121/1.419205

出版商:Acoustical Society of America    

年代:1997

数据来源: AIP

摘要:

Using covariance structure anaysis, two linear models of vowel duration were tested. The first model, the Direct Effects Only model, tested the direct effects on vowel duration of intrinsic duration, following consonant voicing, stress, position in utterance, change in fundamental frequency, and final lengthening. The second model, the Direct and Indirect Effects model, tested all of the above direct effects on duration, plus the indirect effects of stress and phrase final position on vowel duration via their effects on fundamental frequency change. Model goodness of fit was estimated by three fit indices: the Normed Fit Index (NFI), the Nonnormed Fit Index (NNFI), and the Comparative Fit Index (CFI). The Direct Effects Only model yielded the following fit indices: NFI=0.810, NNFI=0.788, and CFI=0.821. The Direct and Indirect Effects model yielded slightly higher fit indices: NFI=0.822, NNFI=0.802, and CFI=0.833. Position in syllable, position in phrase, intrinsic duration, phrase final position, stress, and change in frequency were significant predictors of duration in both models. In the Direct and Indirect Effects model, phrase final position had an indirect effect on vowel duration via a significant effect on change in frequency; stress did not.

ISSN:0001-4966    

DOI:10.1121/1.419209

出版商:Acoustical Society of America    

年代:1997

数据来源: AIP

摘要:

Data from the acoustic surface reverberation experiment (ASREX) [Williamsetal., J. Acoust. Soc. Am.97, 3404 (1995)], is analyzed using multivariate Fourier and wavelet techniques to investigate relationships between environmental conditions and observed acoustic surface backscatter at 200, 400, and 800‐Hz frequencies. Environmental variables observed included wind speed and direction, wave height, air and sea temperatures, current speed and direction, ambient noise, and various measurements of bubble activity characterized by depth‐integrated air volume and e‐folding depth. The relative contribution of the environmental parameters to observed backscattering strengths at the various frequencies will be discussed. The analysis techniques will be described, compared, and contrasted. [Work supported by ONR.]

ISSN:0001-4966    

DOI:10.1121/1.419339

出版商:Acoustical Society of America    

年代:1997

数据来源: AIP

摘要:

Stochastic models and algorithms are developed for representing and simulating marine mammal spatial distributions, collective motions, and vocalization occurrence times. A general class of Markov‐process models of collective motion is first identified. Within this class, particular representations are constructed that incorporate interaction mechanisms of the type developed in distributed behavior modeling [C. Reynolds, Comput. Graph.21, 25 (1987)]. Collective motion is represented by a Markov process with a stationary probability density defined in terms of an ‘‘energy’’ function that expresses the propensity for clustering and polarization within groups of individuals. Computer‐generated sample functions of the Markov processes, based on variants of the Metropolis algorithm, provide for simulation of a variety of collective motions. These range from nearly independent motions of solitary individuals to motions in highly clustered and polarized configurations simulating motions of marine mammals in pods. Occurrence times of vocalizations are represented by Markov processes and associated simulation algorithms. Simulations of patterned click sequences (codas) [Watkins and Schevill, J. Acoust. Soc. Am.62, 1485 (1977)] produced by sperm whales (Physetercatodon) are presented as an example. Potential applications to sonar systems and marine mammal acoustic surveys are briefly discussed. [Work supported by ONR and NRL.]

ISSN:0001-4966    

DOI:10.1121/1.419343

出版商:Acoustical Society of America    

年代:1997

数据来源: AIP

摘要:

Analytical methods are used to express the dependence of low‐frequency bistatic reverberation on the power and direction spectra of the time‐dependent sea surface. The 3‐D cross‐spectral density (CSD) of the reverberation is obtained, to second order in the small‐waveheight approximation, as a sum of dc and sideband terms. For numerical simulations, this theoretical formulation is applied to a model ocean consisting of a homogeneous half‐space. The sensitivity of the CSD to the power spectrum of the source, the power and direction spectra of the surface, and bistatic source/receiver geometry are assessed for both the zero‐Doppler components and the Bragg–Doppler sidebands. [Work supported by ONR.]

ISSN:0001-4966    

DOI:10.1121/1.419347

出版商:Acoustical Society of America    

年代:1997

数据来源: AIP

摘要:

This study investigated whether the relative amount of speech information contained in specific frequency regions of the speech spectrum differs between implantees and normal‐hearing subjects, or between implantees having different speech perception abilities. In addition, the relationship between the amount of speech information implantees perceived in each frequency region and their ability to discriminate adjacent electrodes was investigated. Fifteen subjects using the Nucleus CI‐22M implant (SPECTRA‐22 processor/SPEAK strategy), and 16 normal‐hearing listeners participated. Articulation index procedures were used to measure the amount of speech information perceived in different frequency bands. Results showed that, compared to the normal‐hearing listeners, the implantees obtained less speech information in all frequency regions, but more information in the higher, compared to the lower, frequencies. Correlation analysis showed that electrode discrimination and perceivable speech information were correlated in the lower frequency regions, but not in the highest region, suggesting that the perception of speech information in the lower regions requires the ability to discriminate electrodes, while the perception of high‐frequency information does not. These findings predict that speech perception may be improved by changing the assignment of frequencies to electrode positions to provide better discrimination of low‐ to mid‐frequency speech information.

ISSN:0001-4966    

DOI:10.1121/1.419355

出版商:Acoustical Society of America    

年代:1997

数据来源: AIP

摘要:

Three measures of auditory‐visual (AV) integration using both congruent and discrepant speech stimuli were evaluated in 30 hearing‐impaired subjects. The first measure was based on Braida’s model of pre‐labeling integration [Q. J. Exp. Psych.43, 647–677 (1991)] which makes predictions of AV consonant recognition from separate auditory (A) and visual (V) consonant confusion matrices. The ratio of observed‐to‐predicted AV score was used as an index of integration, with unity representing optimal integration. The second measure used discrepant A and V consonant–vowel stimuli to determine susceptibility to ‘‘McGurk’’ effects. Susceptibility was defined as the difference between A and AV consonant recognition, with greater differences reflecting greater susceptibility. The third measure used temporally misaligned AV speech stimuli in a sentence recognition task to determine individual sensitivity to AV asynchrony. Auditory onset delays between 0‐320 ms were studied and individual sensitivity to AV asynchrony was defined in terms of decrements in sentence intelligibility relative to the synchronous condition (0‐ms delay). Relations among these three potential measures of AV integration, and between integration measures and a separate measure of relative AV sentence benefit, will be discussed. [Work supported by NIH Grant No. DC00792 and the Department of Clinical Investigation.]

ISSN:0001-4966    

DOI:10.1121/1.419357

出版商:Acoustical Society of America    

年代:1997

数据来源: AIP

摘要:

Normally hearing people have much lower speech reception thresholds (SRTs) in a background of a single talker than in the background of speech‐shaped noise, whereas hearing‐impaired people do not. The hearing impaired appear to be less able than normal to take advantage of temporal and spectral ‘‘dips’’ in interfering speech. SRTs were measured in background sounds that varied to the extent that they contained such dips. The subjects tested were: (a) young with normal hearing; (b) elderly with near‐normal hearing; (c) young with moderate to severe cochlear hearing loss and (d) elderly with moderate to severe cochlear hearing loss. In a background that contained both spectral and temporal dips, the hearing‐impaired and elderly with near‐normal hearing performed much more poorly than normals. The signal‐to‐background ratio required for 50% intelligibility was about 19 dB higher for elderly hearing impaired than for young normals. Young hearing‐impaired subjects showed a slightly smaller deficit, but still a substantial one. Linear amplification combined with appropriate frequency‐response shaping (NAL amplification) only partially compensated for these deficits. It is proposed that noise with spectral and temporal dips provides a potentially useful way of evaluating the effects of signal processing such as frequency‐selective amplification and compression.

ISSN:0001-4966    

DOI:10.1121/1.419363

出版商:Acoustical Society of America    

年代:1997

数据来源: AIP

摘要:

An acoustic dataset collected during the 1983 Florida Straits Experiment [H. A. DeFerrari and H. B. Nguyen, J. Acoust. Soc. Am.79, 299 (1986)] is modeled using the normal‐mode acoustic model SWAMP (shallow water acoustic modal propagation by M. F. Werby and N. A. Sidorovskaia) and parabolic acoustic model RAM (range‐dependent acoustic model by M. Collins). The goal of the computer simulation is to get systematic studies of the influence of range‐independent and range‐dependent environments on the received pulse structure observed in the experiment. In the experiment the pulse, with central frequency 460 Hz, was sent through three different shallow‐water acoustic channels with surface and bottom ducts. It was found that the surface duct generated an important temporally stable effect for the range‐independent environment. At a range 42 km, six surface trapped precursors were observed. The range‐dependent propagation channel implied that the energy transfers between modes made the effect less pronounced when only three weak precursors could be seen. The obtained systematic studies can be important in constructing inverse tomography techniques, in predicting sea‐bed characteristics, and in better understanding the influence of the ocean ducts on sound propagation.

ISSN:0001-4966    

DOI:10.1121/1.419311

出版商:Acoustical Society of America    

年代:1997

数据来源: AIP