Regenerating the spectral shapes of [s] and [∫] from a limited set of articulatory parameters

Abstract

This work was aimed at exploring articulatory-acoustic relationships in the production of French fricatives. More precisely, an attempt was made to find out whether the spectral shapes of [s] and 10 can be regenerated from the x and y coordinates of three electromagnetic transducers affixed to the tongue in the midsagittal plane. The corpus was composed of the two fricatives Is/ and Is/ combined with the vowels /al and Ii/ in sequences of the type lVs~/ and /V@VI, and was read by one male native speaker of French. The spectrum regeneration was based on a statistical procedure which consisted of estimating the factors explaining the main part of the acoustic variance from the position of the transducers, by means of multiple linear regression. The articulatory-acoustic correlations were high and allowed us to regenerate the fricative spectra with a good accuracy. The way in which the acoustic parameters varied as a function of the articulatory ones in the statistical model was in good agreement with data reported in previous works. The results support the idea that the tongue has relatively few degrees of freedom in the production of [s] and 10.

Crosslinguistic acoustic categorization of sibilants independent of phonological status

“…The acoustic properties of coronal fricatives have been investigated in a large number of studies, including Hughes and Halle (1956), Heinz and Stevens (1961), Strevens (1960), Bladon and Seitz (1986), Clark and Bladon (1986), Behrens and Blumstein (1988), Forrest, Weismer, Milenkovic andDougall (1988), Nittrouer, Studdert-Kennedy andMcGowan (1989) and Nguyen, Hoole, and Marchal (1994). Here we will briefly discuss the main points covered in these papers.…”

Dimensionality reduction of electropalatographic data using latent variable models

“…This is a matter currently being investigated (e.g. see (Stone, 1991;Nguyen et al, 1994Nguyen et al, , 1996 and references therein); a dimension of 5±10 is usually suggested. Alternatively, one can try to infer this intrinsic dimension directly from the data set by standard statistical techniques of model selection.…”

“…Potentially the tongue has many degrees of freedom owing to its lack of skeleton (Stone, 1991); however, a number of studies suggest that tongue movements in speech may be appropriately modelled using a few elementary articulatory parameters (e.g., (Nguyen et al, 1994(Nguyen et al, , 1996 and references therein). Spatial redundancy arises due to the limited number of possible tongue con®gurations ± only a tiny fraction of the potential 2 62 % 4X6 Â 10 18 EPG patterns can be produced by the articulatory system.…”

Modeling tongue-palate contact patterns in the production of speech

“…Using a statistical method, Harshman et al (1977) have demonstrated that the tongue sagittal shape can be reduced by means of a factor analysis to only two main components, namely the degree of raising/lowering of the front and of the back of the tongue, respectively (see also Maeda (1990), for a similar approach). In a recent study on the production of fricatives (Nguyen et al, 1994), high correlations between the x and y coordinates of three tongue points and the acoustic spectrum also tended to indicate that the tongue is controlled by a limited number of parameters. This apparent limitation in the variety of actual tongue shapes and movements has been attributed to di erent factors.…”