ACCURATE RECOVERY OF ARTICULATOR POSITIONS FROM ACOUSTICS : NEW CONCLUSIONS BASED ON HUMAN DATA

摘要: Vocal tract models are often used to study the problem of mapping from acoustic transfer function vocal area (inverse mapping). Unfortunately, results based on strongly affected by assumptions underlying models. In this study, acoustics (digitized speech samples) articulation (measurements positions receiver coils placed tongue, jaw, and lips) is examined using human data a single speaker: Simultaneous articulator measurements made for vowel-to-vowel transitions, /g/ closures, transitions into out closures. Articulator were measured an EMMA system track lips, tongue. Using these data, look-up tables created that allow be estimated signals. On set not making tables, correlations between actual coil around 94% root-mean-squared errors 2 mm common An error source evaluation shows estimating quantized gives typically less than 1 greater would obtained quantizing themselves. This agrees with extends previous studies showing studied, can accurately recover positions.