Dr. McGowan has been active in developing the Haskins Laboratories articulatory model for a number of years, with the aim of recovering the articulatory trajectories of the speaker's vocal tract from the acoustic signal. The method involves analyzing the signal to devise a hypothesis of how the vocal tract must have behaved in order to produce that signal, and then resynthesizing the signal using those articulatory parameters. A genetic algorithm is used as an optimization algorithm to recover the speaker's behavior performs fitness proportionate selection, mating and mutation, iterating for 60 generations. The resulting articulatory hypothesis can be compared with articulatory data from the speaker, and the synthesized speech signal can be compared to the original acoustic signal. The system was developed on the basis of formant frequencies of sonorant speech sounds like vowels, since the relationship between articulatory shape and acoustic outcome is particularly well worked out for these acoustic parameters. The proposed work will extend the articulatory model to the more challenging problem of non- sonorant consonants, which include abrupt discontinuities introduced into the signal by tighter constrictions in the vocal tract and more sudden releases of those constrictions. It will also focus on modelling the particular vocal tracts of several speakers in the extensive articulatory database developed by John Westbury and his colleagues using the x-ray microbeam facility at the University of Wisconsin; simultaneous pellet tracks and acoustic recordings are available for a wide variety of consonants and vowels.
| McGowan, Rebecca W; McGowan, Richard S; Denny, Margaret et al. (2014) A longitudinal study of very young children's vowel production. J Speech Lang Hear Res 57:1-15 |
| Jackson, Michel T-T; McGowan, Richard S (2008) Predicting midsagittal pharyngeal dimensions from measures of anterior tongue position in Swedish vowels: statistical considerations. J Acoust Soc Am 123:336-46 |
