Authors

J. W. Mao; A. Vosoughi;L. H. Carney

Comments

Authors: contact us about adding a copy of your work at STARS@ucf.edu

Abbreviated Journal Title

J. Acoust. Soc. Am.

Keywords

AUDITORY-SYSTEM; PHENOMENOLOGICAL MODEL; AMPLITUDE-MODULATION; REPRODUCIBLE NOISE; QUANTITATIVE MODEL; DICHOTIC DETECTION; MASKED; THRESHOLD; NARROW-BAND; DISCRIMINATION; MASKING; Acoustics; Audiology & Speech-Language Pathology

Abstract

Tone-in-noise detection has been studied for decades; however, it is not completely understood what cue or cues are used by listeners for this task. Model predictions based on energy in the critical band are generally more successful than those based on temporal cues, except when the energy cue is not available. Nevertheless, neither energy nor temporal cues can explain the predictable variance for all listeners. In this study, it was hypothesized that better predictions of listeners' detection performance could be obtained using a nonlinear combination of energy and temporal cues, even when the energy cue was not available. The combination of different cues was achieved using the logarithmic likelihood-ratio test (LRT), an optimal detector in signal detection theory. A nonlinear LRT-based combination of cues was proposed, given that the cues have Gaussian distributions and the covariance matrices of cue values from noise-alone and tone-plus-noise conditions are different. Predictions of listeners' detection performance for three different sets of reproducible noises were computed with the proposed model. Results showed that predictions for hit rates approached the predictable variance for all three datasets, even when an energy cue was not available.

Journal Title

Journal of the Acoustical Society of America

Volume

134

Issue/Number

1

Publication Date

1-1-2013

Document Type

Article

Language

English

First Page

396

Last Page

406

WOS Identifier

WOS:000321908500059

ISSN

0001-4966

Download

Share

COinS