The historical threads of a computational model that transforms relative acoustic events into speech based on the sensitivity of the vocal tract

This presentation will begin with review of a model in which speech segments can be encoded by specifying relative acoustic events along a time axis that consist of directional changes of the vocal tract resonance frequencies called resonance deflection patterns (RDPs) [JASA, 146(4), 2522–2528]. These events are transformed via acoustic sensitivity functions, into time-varying modulations of the vocal tract shape. RDPs specifying /b, p/, /d, t/, and /g, k/, for example, would typically be coded as [−1 −1 −1], [−1 1 1], and [−1 1 −1], respectively, where the vectors indicate, from left to right, the targeted directional frequency shift of the first, second, and third resonances of the vocal tract. When combined with a voice source, the model can generate sentence-level speech. This model was motivated by research reported over the past several decades on understanding changes of the resonance frequencies of the vocal tract to changes of the vocal tract configuration through calculation of sensitivity functions. Some of this historical work will be highlighted as providing the foundational building blocks that allowed for development of the current form of the model.