Committee: 

Dr. Anderson, Advisor   

Dr. Rozell, Chair

Dr. Moore

Abstract:

The objective of the proposed research is to provide patients in voice therapy with automated voice quality feedback outside of the clinical room through i-vector and deep learning approaches. Voice disorders affect a large portion of the population, especially impacting heavy voice users such as teachers or call-center workers. Voice therapy requires regular voice technique practice under the guidance of a voice therapist in weekly therapy sessions. Patients commonly experience difficulty reproducing the prescribed voice technique and voice quality independently between sessions without the guidance of a voice therapist. By adapting the i-vector approach and other deep learning algorithms, the proposed work effectively examines a speaker's different voice quality modes in unscripted and connected speech with high accuracy, demonstrating the potential of these methods to extend therapist judgment beyond the clinic walls.