Spectral and temporal information and presentation mode effects on individual speaker identification and listening effort

Abstract

Identifying speakers of interest in an auditory scene is a fundamental task that facilitates effective communication. Little is known about the specific contributions of spectral and temporal detail required for identifying a specific speaker of interest by human listeners. This study investigated the relative contributions of spectral and temporal detail for identifying a speaker of interest and perceived effort in doing so. Results showed significant improvements in speaker identification and decreased effort ratings as spectral channels increased. Improved speaker identification performance with increased temporal filter cutoff from 20 Hz to 800 Hz was observed. These results have implications for speech signal processing by amplification devices and automated speaker recognition systems.