Listener characteristics and the perception of speech in noise

Abstract

Speech communication is often made difficult by the presence of background noise. Much research on the perception of noise-masked speech has focused on the masking of phonetic information by different types of noise (e.g., white noise, speech-shaped noise, temporally modulated noise, multi-talker babble). The present work focuses on the relationships between some cognitive characteristics of listeners and accuracy in the identification of noise-masked consonants. 37 listeners identified numerous tokens of each of 4 consonants (p, b, f, v) in CV syllables produced by 8 talkers (4 male, 4 female) masked by 10-talker babble. Listeners also completed a number of tasks designed to measure selective attention: two dichotic listening tasks and two non-speech discrimination tasks. On each trial of the dichotic listening tasks, one or the other ear was cued visually (i.e., “right ear” or “left ear”), after which the listener indicated the talker sex or the consonant in the target ear, depending on task. In the two non-speech tasks, listeners discriminated either the frequency or the duration of broadband target noise bursts embedded in temporally modulated background noise. Analyses indicate a positive relationship between noise-masked speech accuracy and performance on the dichotic consonant identification and complex non-speech discrimination tasks.