Localization of a Speech Target in Nondirectional and Directional Noise as a Function of Sensation Level

Abstract

Abstract : Both the detectability and localizability of a target signal must be considered when the signal is presented with multiple other sounds in an environment. When target and noise sources are separated spatially, target speech detection, recognition, and intelligibility improve substantially. Less information is known, however, about the effect of spatial separation on a listener's localization performance. In the present study, we examine the accuracy of localization judgments in various multiple sound source environments. Localization performance was examined as a function of sensation level (SL) in an attempt to account for changes in signal audibility resulting from changes in directional position of the sound source. A target speech signal (S) was presented at 0-, 6-, 12-, and 18-dB SL in reference to hearing threshold masked by 65-dB A-weighted nondirectional (NDN) or directional (DN) speech-spectrum noise. The listeners consisted of 40 young adults with normal hearing. At the beginning of the study, directional detection thresholds (DDTs) were established for all listeners in each target-noise condition (5 NDN conditions and 35 DN conditions). We calculated the appropriate SLs needed for the localization task from the DDTs of individual listeners. All listeners demonstrated decreases in localization error ( ) and increases in localization precision (%) as SL increased. Changes in overall localization performance as a function of SL were similar in NDN and DN. When target-noise conditions were grouped according to the location of S, localization performance was (a) most accurate when S originated from 0 and +90 azimuth, (b) relatively accurate when S was positioned at +45 azimuth, and (c) poor when S originated from rear-horizontal plane positions (+135 and 180 azimuth).