Quantifying Between-Speaker Variation in Ultrasound Tongue Imaging Data

Abstract

This article outlines a quantitative, between-group comparison of tongue shapes using ultrasound tongue imaging, one of the vocal tract imaging techniques widely used in articulatory phonetics research. This article first provides a brief overview of ultrasound tongue imaging, followed by a description of a cross-speaker normalisation method based on bite plane rotation. It then outlines ultrasound data recording and analysis workflow with a case study illustrating data analysis using Principal Component Analysis (PCA). This paper demonstrates that the bite plane rotation, coupled with statistical normalisation methods, allows for a reliable tongue-shape comparison by establishing a common coordinate system across speakers.