The Mean-ing of Vowels: Stability and Convergence of Vowel Formants in Read Speech
This study aimed to determine the minimum amount of read speech required for stable vowel formant measurements in speakers. Using audiobook data, the research utilizes a pipeline of forced alignment, formant extraction, outlier filtering, and convergence detection to estimate stable, speaker-specific vowel values.


LREC 2026 Conference Version
A shorter conference version of this thesis work was accepted to LREC 2026 and presented as a poster.