Abstract

Speech variation is of importance to workers in automated speech recognition and synthesis. This paper will discuss specific sources of variability that are of practical concern to these researchers. Quantitative evidence will be presented from American dialects of both English and French to demonstrate how vowel formant frequency variation is correlated with both the sonority of intrasyllabic following consonants and with prosodic information. Evidence will also be presented from analysis of vowel formant information from various speech ‘‘styles’’ and ‘‘registers’’ to demonstrate how researchers can appropriately integrate such variation into synthesis modules. Evidence for several sources of variation will be presented, and both theoretical and practical conclusions will be drawn from the evidence. [Work partially supported by NSF.]

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call