D. H. Whalen
Haskins Labs., 270 Crown St., New Haven, CT 06511
Charles E. Hoequist
BNR, Research Triangle Park, NC 27709
Although synthesizers do not have any air flow, they are representing creatures that do. While it is quite common to have long stretches of speech without audible intake (especially with speakers trained in acting and broadcasting), synthesis might be somewhat less artificial with the intake of breath added. To see whether adding an audible intake would affect the perception of synthetic speech, breath intake (naturally produced) was introduced in a synthesis transcription task. Subjects received the breath-added stimuli on either the first half of the test or the second. The largest effect was a typical one of increased performance in the second half, due to increased experience with the synthesizer. However, those subjects who received the breath second had a much larger improvement than those who had it first, who were essentially the same in the second condition as the first. The results indicate that, as with almost anything that increases naturalness, allowing computers to breathe will improve the perception of synthetic speech. [Work supported by NIH Grant No. HD-01994.]