New Preprint: Vowel Acoustics as Predictors of Speech Intelligibility in Dysarthria
Updated: Jul 30
Last week, my coauthors and I published a new preprint to OSF titled "Vowel Acoustics as Predictors of Speech Intelligibility in Dysarthria." This project examined the relationship between a handful of acoustic measures and speech intelligibility.
I've been dreaming of doing this project ever since I read Whitefield and Mehta's (2019) study examining various vowel space measures, including vowel space density, in speakers with Parkinson's disease. I instantly fell in love with this measure and wanted to explore using it in my research, which led to the development of this project.
My favorite part about this project was calculating the various vowel space measures (i.e., vowel space area, hull, and vowel space density) and visualizing them using ggplot2 in R. In particular, I thoroughly enjoyed making this plot shown below (making figures is my favorite part of research).
The most challenging part of this project was filtering the raw formant data obtained from Praat to reflect just the vowel formant data (i.e., removing consonant frequency data and formant tracking errors). In the project, we ultimately used filtering methods that were reported in the previous literature. However, when I visualized the filtered formant spaces for all of the speakers, I realized that this filtering method might have some flaws. For some speakers, it removed data that appeared to reflect vowel formant data. In other cases, it retained outlying data that were caused by formant tracking errors or consonant/fricative data. I believe that there may be a better, more nuanced way to filter the data. I hope to tackle this issue in some of my future work.