Date of Defense

4-30-2021

Date of Graduation

5-2021

Department

Speech, Language and Hearing Science

First Advisor

Stephen Tasko

Second Advisor

Hope Gerlach

Abstract

Purpose: People who stutter tend to be more variable in their motor speech movements than people who do not stutter, and this variability has been measured using both kinematic and acoustic metrics. This study compared well-established kinematic metrics of variability and a novel acoustic approach that measures variability for different frequency bands across different speech tasks and conditions in adults who do and do not stutter.

Method: Two kinematic metrics and four acoustic metrics of variability were generated and compared. The kinematic metrics measured the variability exhibited in the tongue and lips, and the acoustic metrics measured the variability in the original speech signal, as well in 3 frequency bands that roughly corresponded to the formants F0/F1, F2, and F3. Variability was assessed as a function of diagnostic category, speech task, and speech condition (rate and loudness).

Results: Out of the four acoustic metrics, the wide-band variability metric had the highest correlation with the kinematic variability metrics across all speech tasks and conditions. Filtering the acoustic signal into frequency bands proved to add little additional information, and the correlation between the kinematic and acoustic metrics weakened as the acoustic metrics went from wide-band to low to mid to high.

Conclusions: The wide-band acoustic metric seems less sensitive to stuttering diagnosis than the kinematic metrics and the acoustic metrics in general provide a smaller range of values and variance than kinematic metrics. However, of all acoustic metrics, the wide-band predicted the highest percentage of variability in both kinematic metrics, and it seems to be the most effective proxy for kinematic analyses.

Access Setting

Honors Thesis-Restricted

Download

Share

COinS