Abstract
The influence of masking noise on the ability of listeners to distinguish among speakers was measured for noise-to-signal ratios of − 20 dB to + 10 dB. The prerecorded stimulus ensemble contained 15 utterances: three test sentences, each spoken by five speakers. Each trial consisted of two different test sentences, spoken by the same or different speakers with equal probability. Masker level was the same for the two utterances in a trial. The subjects' task was to judge whether the two utterances were spoken by the same or by different speakers. Five listeners unfamiliar with the speakers took part in the experiment. Percentage of correct responses decreased from 80% for a noise-to-signal ratio of − 20 dB to 65% for a noise-to-signal ratio of + 10 dB (chance performance = 50%), while response time increased monotonically. Multidimensional scaling techniques were used to determine those speakers for whom discrimination performance was most sensitive to masking noise. Continuing experiments with other signal degradations are outlined.
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call