The temporal organization of synthetic speech of less than optimum quality should be adapted to the temporal course of speech perception. Considerations on the process of spoken word recognition make it likely that the difficulties encountered by listeners, due to degraded speech quality, can be reduced by regularly inserting speech pauses, even where such pauses are unnecessary in high-quality speech. This hypothesis is confirmed in a listening test comparing the intelligibility of Dutch natural and diphone-concatenated speech, both with and without grammatical speech pauses.
|Journal||IPO Annual Progress Report|
|Publication status||Published - 1983|