Modelling the Noise Source in Voiced Fricatives

Shadle, C.H. (1995) Modelling the Noise Source in Voiced Fricatives

Abstract

The noise source in voiced fricatives has not received as much attention as that of unvoiced fricatives, in part because the voiced case, with two sound sources, is undoubtedly more complicated, and the unvoiced case cannot be considered to be solved. In this paper results from previous studies are considered together with data from three subjects to correct this imbalance. The classic model of voiced fricatives includes two sources of sound: a periodic volume-velocity source located at the glottal end of the tract, and a noise source located in the vicinity of the primary tract constriction. The amplitude of the noise source has long been assumed to be modulated by the voicing, although this effect is sometimes neglected. The noise source before modulation is presumed to be similar to that used in models of unvoiced fricatives: it consists of white or broadband noise, and its strength depends primarily on the pressure drop across the constriction. This latter characteristic means that in general it is weaker than the noise source in unvoiced fricatives, that is, it produces less noise because the pressure drop across it is lower. This has been attributed to the need to maintain a significant transglottal pressure drop in order to maintain voicing, which therefore reduces the pressure differential that can be maintained across the constriction. It has been noted, however, that different speakers use different strategies with regard to glottis-constriction coordination, and so the picture is somewhat more complex. Apart from coordination issues, characterization of the noise source is more complex in certain other respects. First, the geometry of the vocal tract downstream of the constriction has a significant effect on the noise source spectrum, in particular by offering an obstacle to the emerging jet at which noise is generated (Shadle, 1990). Within a voiced-voiceless pair, one could assume that the geometry and therefore the parameters controlling the noise source spectrum are the same. Some work has been done on characterizing the dependence of spectral amplitude and spectral tilt on pressure drop and constriction area for [f,s,]. Second, while variation in the flowrate through a constant-area constriction can be predicted to change spectral amplitude and tilt of the noise source, it is not clear how such modulation would be timed with respect to glottal vibration. Acoustic variations generated at the glottis will travel at the speed of sound to the constriction; hydrodynamic variations, which may be of similar strength, will convect at a slower rate that depends on vocal tract area and is therefore much more difficult to predict. There are therefore two distinct problems in characterizing the noise source in voiced fricatives: understanding the nature of the glottis-constriction coordination, and describing the effect of the modulation imposed by voicing. We focus on the latter in this paper by describing the results of an Fo -synchronous analysis of a voiced fricative, and comparing to results of mechanical model studies.

This record has no associated files available for download.

Contributors

Download statistics