Generation and Modelling of Fricative Phonemes

Fricative Consonants

The unvoiced fricative phonemes stem from the hissing of a steady airstream through the mouth. Such sounds lack the combined glottal impulse and vocal tract vibration phenomenon of voiced sounds. Unvoiced phonemes are distinct from one another due to differences in the positions of the lips, teeth, and tongue rather than the presence of formant frequencies. [2]

The voiced fricative phonemes include both vocal tract formant resonances driven by glottal impulses and fricative hiss.

We modelled the hiss used to create fricatives as a zero-mean Gaussian white process of low variance. The glottal driver, when present, was modelled as a 120 Hz pulse train. We fitted an AR model to a sampled utterance of each of these phonemes to represent filtering due to the mouth and, in the case of the voiced fricatives, the vocal tract. We preconditioned our sampled voiced fricative utterances to remove the glottal impulse peak from the spectrum before modelling; such preconditioning was unnecessary for the unvoiced fricatives.


Timothy D. Dorney and Robert H. Sparr
Electrical and Computer Engineering Department, Rice University
April 1996