Synthesis of fricative sounds using an aeroacoustic noise generation model
Daniel J. Sinder, Michael Krane, James L. Flanagan · The Journal of the Acoustical Society of America · 1998
Results from speech synthesis with an aeroacoustic fricative source model are presented. The source model, which is implemented in the time domain, is based on Howe’s reformulation of the acoustic analogy [Howe, J. Fluid Mech. 71, 625–673 (1975)]. It specifies the source as a function of flow conditions and the geometry of the vocal tract downstream from a constriction where a turbulent jet is formed. This is accomplished by combining elements of aeroacoustic theory and a model for jet behavior with a traditional model for wave propagation in the vocal tract. Using this formulation, the source strength, impedance, and spectrum can be determined through post-processing. The fricative model is implemented in an articulatory speech synthesizer. Evaluation of the resulting synthesis is emphasized. Comparisons are made with physical experiments in static configurations. Additionally, speech fricatives are compared with natural speech as well as synthetic speech from other models. These comparisons are made in the frequency domain and through listening tests in which fricatives in V-C-V sequences are judged for naturalness and intelligibility. [Research supported by NSF/ARPA IRI-9314946 and ARPA DAST 63-93-C-0064.]