Title:  Analysis of Affective States from Vocal Acoustics in Adults with Aphasia

Committee:

Dr. Elliot Moore, ECE, Chair , Advisor

Dr. Jacqueline Laures-Gore, GSU, Co-Advisor

Dr. Mark Clements, ECE

Dr. Mark Davenport, ECE

Dr. Pamela Bhatti, ECE

Dr. Bruce Walker, CoC

Abstract:

This research analyzed objective vocal acoustic measures of aphasic speech as they related to the detection or prediction of stress, depression, and emotional state in adults with aphasia. Assessing stress and depression in persons living with aphasia is a challenging task, often utilizing caregiver surveys or diagnostic tests which have not been modified for language difficulties. Cepstral Peak Prominence features were found to be most useful when creating an automatic classifier to detect depression. Further analysis of the feature sets suggested many of the features were moderately correlated to changes in short-term affective states. Additionally, the same feature sets used to attempt the detection and prediction of stress and depression were also used to successfully predict the presence of dysarthria in a cross-database analysis. Further work in this area could lead to automated tools to assist clinicians with their diagnoses of stress, depression, and other forms of affect in adults with aphasia.