Effects of Context on Talker Identification

Publisher Website
Google Scholar
Abstract
The ability to identify talkers from monosyllables spoken in context was examined: Kersta's method of visually comparing spectrograms was employed. Ten observers were trained to identify five talkers from spectrograms of two words spoken in isolation. The experimental task then required the observers to identify the same talkers from the same words spoken in different contexts. The correct rates for the training task (78.4%) could not be reproduced in the experimental task (37.3%). The results were interpreted to indicate that different contexts decrease the identification ability of observers because: (a) the shorter stimulus durations of words in context decreases the amount of acoustic information available for matching, and (b) the different spectrographic portrayals introduced by different phonetic contexts outweighs any intratalker consistency.

This publication has 0 references indexed in Scilit: