BME PhD Prospectus Defense - Jenny Vojtech
- Starts: 9:00 am on Thursday, August 30, 2018
Title: “Acoustic and High-Speed Videoendoscopic Techniques to Improve Voice Assessment via Relative Fundamental Frequency” Committee: Cara Stepp, PhD – BU Biomedical Engineering/Speech, Language and Hearing Sciences (Advisor, Chair) Steve Colburn, PhD – BU Biomedical Engineering Melanie Matthies, PhD – BU Speech, Language and Hearing Sciences Matías Zañartu, PhD – Universidad Técnica Federico Santa María Electronic Engineering Abstract: Relative fundamental frequency (RFF) has shown promise as an acoustic tool for assessing laryngeal muscle tension. However, current semi-automated RFF estimation algorithms are insensitive to sample characteristics (e.g., overall severity of dysphonia, recording conditions of the room). Depending on which samples are used to train and test the algorithms, significant variations in algorithmic performance are observed when compared to manual RFF estimates. Moreover, current RFF algorithms make use of a set of acoustic features to identify the initiation and termination of vocal fold vibration within an acoustic signal. Yet, not only is it unclear which acoustic features best correspond to this vibratory boundary, but there has also been no physiological validation to confirm that vibration begins or ends at the identified point in time other than by manual approximation. The proposed project will employ acoustic and high-speed videoendoscopic (HSV) techniques to optimize current RFF estimation algorithms for the purpose of objectively quantifying the degree of laryngeal muscle tension. Specifically, sub- routines will be incorporated into the current algorithms in order to improve RFF estimation across the spectrum of dysphonia severity and signal acquisition quality within a large cohort of healthy and disordered (e.g., vocal hyperfunction, Parkinson’s disease) voices. Following, HSV images will be recorded concurrently with acoustic signals in healthy and disordered voices in order to identify and validate acoustic features that best coincide with the physiologically-detected boundary between voicing offset and onset. The resulting optimized RFF estimation algorithms will then be applied to examine the physiological basis of RFF by determining the relationship between vocal fold abductory kinematics and RFF estimates in healthy and disordered voices.
- 44 Cummington Mall, room 203