Practical, portable, and pain-free: Improving ultrasound image collection for endangered languages
Date:
Abstract: Many of the world’s smaller, endangered languages remain phonetically under-described. Instrumental studies of such languages have unique challenges, as there are a limited number of competent, willing participants, and many of them are elderly. It is often difficult to bring participants into a laboratory setting, but even in the field they may not be comfortable with invasive instrumental studies. Ultrasound can yield great results in field phonetics, due to its portable and non-invasive nature, but its images are inherently collected with respect to the probe’s position, and not with respect to common articulatory landmarks such as the palate. Positional correction of these images has traditionally involved larger, more complex, and more invasive methods such as electromagnetic articulography, and are not feasible in the field. In this study, an improved method of optical correction of ultrasound is demonstrated using only a smartphone camera and minimal markers on the ultrasound probe and speaker’s face. Processing of the collected imaging and positional correction is accomplished using custom scripts written in MATLAB. Challenges in implementation and analysis are also presented, as well as potential improvements for the future.