Modelling the Noise Source in Voiced Fricatives

The classic model of voiced fricatives includes two sound sources, a periodic glottal source and a random noise source at the constriction. The amplitude of the noise source has long been assumed to be modulated by the voicing (Fan, 1960; Stevens, 1971; Flanagan, 1972), but the amount of this change has not been studied systematically, and the possibility of a change in spectral shape has been ignored. Spectral analysis of a voiced fricative that is F0-synchronized by the use of a simultaneous laryngograph channel shows that modulation of the noise source by voicing causes not only amplitude changes but also changes in the spectral shape. These changes, consisting of a 5-dB variation in amplitude above 10kHz in [v], are consistent with changes observed in time-averaged spectra of sustained fricatives at different effort levels, and with spectra of mechanical models having different constriction shapes. The amount of spectral change due to such modulation is likely to vary with strength and place of the noise source, and thus will depend on the particular glottis-constriction coordination employed by the speaker. Extension of this work to more fricatives and more subjects is planned