Author: Geng-Min Melody Lu
Lu, Geng-Min Melody, 2022 Acoustic analysis of snoring sounds towards testing relationships with other respiratory signals and health outcomes, Flinders University, College of Science and Engineering
Terms of Use: This electronic version is (or will be) made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. You may use this material for uses permitted under the Copyright Act 1968. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact copyright@flinders.edu.au with the details.
Snoring has been a prevalent sleeping disorder among the general population in Australia with around 20%-25% of Australians regularly snoring on most nights. Even though a lot of studies have been conducted to develop more accurate acoustic methods to identify and classify snoring signals, further investigation is still needed to advance snoring assessment methods. Moreover, the consensus for the objective definition of snoring is still lacking. This study aims to explore the information embedded in the acoustic signals of snoring and to examine the performance of a snoring detection algorithm. Another objective of this study is to advance the acoustic analytic techniques for future development that would allow for a more comprehensive differential diagnosis and evaluation of snoring and obstructive sleep apnoea in a home setting. A total of 2330 20-second audio segments from 6 participants were used for the acoustic analysis of snoring sounds. The performance of a snoring detection algorithm was evaluated by a confusion matrix, ROC curve, and other performance evaluation metrics. This study that indicated snoring happens in the lower frequency range at about 200 Hz. There were no noticeable differences between the snoring and non-snoring episodes at frequencies between around 200 Hz and 1000 Hz. For this snoring dataset, 0.62 was the best-performing cut-off relative power. Interestingly, the snoring and non-snoring power spectra appeared to fluctuate more in those participants who had a higher number of snoring events recorded. The results of this study may help in the development of an objective definition of snoring based on its acoustic characteristics. In addition, this work can be beneficial for the development of snoring detection algorithms. Future studies to support the conclusions of this study can be carried out by adding more human raters and participants.
Keywords: snoring, acoustic analysis, snoring detection, snoring classification, health outcomes
Subject: Engineering thesis
Thesis type: Masters
Completed: 2022
School: College of Science and Engineering
Supervisor: Kristy Hansen