5th International Conference on Spoken Language Processing

Sydney, Australia
November 30 - December 4, 1998

Temporal Organization of Speech for Normal and Fast Rates

Geetha Krishnan, Wayne Ward

Carnegie Mellon University, USA

In this study predictors of speech rate that are sensitive to local and global rate changes, and relevant to different types of speakers, were examined. Two groups of subjects, normal and disfluent speakers (whose speech was clinically rated as "slow"), provided speech samples at normal and fast rates. Samples were segmented into interstress intervals (ISI) of varying length (i.e., varying number of syllables). The compressibility of components within ISIs of varying length provided information on local rate control strategies. The fast speech samples were useful for examining strategies used in global rate increases. Stressed vowels and intervowel intervals (IVI) showed similar trends in compression for both speakers, for local and global rate increases. We then investigated two measures of speech rate based on intervowel intervals: the ratio measure (IVI/ISI) and the average IVI. High correlation of average IVI with phone rate was found. Results of speech rate estimations are presented.

Full Paper

Bibliographic reference.  Krishnan, Geetha / Ward, Wayne (1998): "Temporal organization of speech for normal and fast rates", In ICSLP-1998, paper 0930.