Perceptual attention in monitoring natural and synthetic speech
Howard C. Nusbaum, Steven L. Greenspan, David B. Pisoni · The Journal of the Acoustical Society of America · 1985
The role of voice distinctiveness and phonetic discriminability in perception of natural and synthetic speech was investigated. Subjects were instructed to monitor sequences of CV syllables for a specified target syllable in several conditions: (1) targets and distractors produced by the same human talker (N/N); (2) targets produced by a synthetic talker and distractors produced by a human talker (S/N); and (3) targets produced by a synthetic talker and distractors produced by both the same synthetic talker and a natural talker (S/N + S). Results indicate that highly intelligible synthetic targets are detected faster mixed with natural distractors, than are natural targets mixed with natural distractors. However, when subjects are required to discriminate between synthetic targets and synthetic distractors, performance is much worse than for natural targets and natural distractors. The distinctive mechanical sound of synthetic speech only appears to aid perception when there is just a single synthetic message among natural messages. When listeners must discriminate among synthetic messages, performance is significantly worse than when they must discriminate among natural messages. [Work supported by AFOSR and NIH.]