Distinguishing Dysarthric Speech: Vowel Acoustics and Measurement

Centralization of vowel formants has been associated with reduced intelligibility in healthy speakers and those with motor speech disorders [1-3]. Hence, estimations of speakers’ vowel space area (VSA) provide a promising avenue for assessing speech motor control. However, VSA measurements have traditionally demonstrated limited success in distinguishing healthy and disordered speech [1, 3]. One factor that may play a role is the formant measurement point. Research in speech motor control almost universally takes measurements of formant values from vowels’ temporal midpoint. While this provides measurement consistency, the influence of vowels’ phonetic context rarely results in symmetrical formant trajectories, and thus midpoint values exhibit considerable variation across word tokens [4]. There is also increasing evidence that a vowel’s temporal midpoint reflects a different stage of articulation in faster and slower speech [4]. For people with motor speech disorders, significant reductions in speech rate are common. It has been hypothesized that the extraction of formant values from a static time point could be responsible, in part, for the failure of studies to consistently reveal reductions in the VSA of this population – despite evidence of reduced lip, tongue and jaw movements [1]. This project explores the extent to which different formant measurement points are capable of distinguishing healthy from mild dysarthric speech using common vowel centralization metrics.