The objective of this thesis is to characterize and identify the representation of four short English vowels in the frequency following response (FFR) of 22 normal-hearing adult subjects. The results of two studies are presented, with some analysis.
The result of the first study indicates how the FFR signal of four short vowels can be used to identity different subjects. Meanwhile, a rigorous test was conducted to test and verify the quality and consistency of responses from each subject between test and retest, in order to provide strong and representative features for subject identification.
The second study utilized machine learning and deep learning classification algorithms to exploit features extracted from the FFRs, in both time and frequency domains, to accurately identify subjects from their responses. We used three kinds of classifiers with respect to three aspects of the features, yielding a highest classification accuracy of 86.36%.
The results of the studies provide positive and important implications for establishing a biometric authentication system using speech-evoked FFRs.
| Identifer | oai:union.ndltd.org:uottawa.ca/oai:ruor.uottawa.ca:10393/40552 |
| Date | 27 May 2020 |
| Creators | Sun, Rui |
| Contributors | Bouchard, Martin, Dajani, Hilmi |
| Publisher | Université d'Ottawa / University of Ottawa |
| Source Sets | Université d’Ottawa |
| Language | English |
| Detected Language | English |
| Type | Thesis |
| Format | application/pdf |
Page generated in 0.0021 seconds