A Method For Distinguishing Tongue Surface Topology For Different Categories Of Speech Sound
The Journal Of The Acoustical Society Of America
An analytic method has been developed to assist in interpreting data from two-dimensional ultrasound tongue imaging by determining whether the position of the tongue surface significantly differs for two categories of speech sound and, if so, which region is most associated with the difference. Individual tongue traces are modeled as curves in a polar coordinate system around a reference point. Two categories of speech sound are compared as a function of the distance between the traces of the two categories at each arc angle. The regions of the arc with the most significant deviation (i.e., where the categories are best contrasted) are shown by the largest ratio of mean to standard deviation (the greatest number of standard deviations separating the mean tongue surfaces). Categories are most distinguished in regions with more than one standard deviation separating them. A ratio of the angular distance between a point of complete overlap (no contrast) and the maximum positive and negative differentiation points has potential as a speaker-agnostic measure of which regions of tongue position are important at the level of the linguistic variety. To demonstrate this method, its application to the vowel anteriority contrast in several Turkic languages is presented.
175th Meeting Of The Acoustical Society Of America
May 7-11, 2018
Jonathan North Washington and P. A. Washington.
"A Method For Distinguishing Tongue Surface Topology For Different Categories Of Speech Sound".
The Journal Of The Acoustical Society Of America.