Abstract:
In adverse listening conditions, talkers can increase their intelligibility by speaking clearly. While producing clear speech, however, talkers often significantly reduce their speaking rate. A recent study [J. C. Krause and L. D. Braida, J. Acoust. Soc. Am. 98, 2982(A) (1995)] showed that talkers can be trained to produce a form of clear speech at normal conversational rates. This finding suggests that acoustical factors other than reduced speaking rate are responsible for the high intelligibility of clear speech. To gain insight into these factors, the acoustical properties of conversational and clear speech were analyzed to determine phonological and phonetic differences between the two speaking modes. These differences were interpreted in terms of error patterns made by normal-hearing listeners identifying key words in the presence of wideband noise. Additional intelligibility tests investigated other degradations in order to explore the robustness of the high intelligibility of clear speech produced at conversational rates. Native and non-native listeners were employed for degradations consisting of additive noise, high-pass and low-pass filtering, and reverberation. [Work supported by NIH.]