AbstractDepression significantly affects a large percentage of the population, with young adult females being one of the most at-risk demographics. Concurrently, there is a growing demand on healthcare, and with sufficient resources often unavailable to diagnose depression, new diagnostic methods are needed that are both cost-effective and accurate. The presence of depression is seen to significantly affect certain acoustic features of the human voice. Acoustic features have been found to exhibit subtle changes beyond the perception of the human auditory system when an individual has depression. With advances in speech processing, these subtle changes can be observed by machines. By measuring these changes, the human voice can be analysed to identify acoustic features that show a correlation with depression. The implementation of voice diagnosis would both reduce the burden on healthcare and ensure those with depression are diagnosed in a timely fashion, allowing them quicker access to treatment.
The research project presents an analysis of voice data from 17 biological females between the ages of 20-26 years old in higher education as a means to detect depression. Eight participants were considered healthy with no history of depression, whilst the other nine currently had depression. Participants performed two vocal tasks consisting of extending sounds for a period of time and reading back a passage of speech. Six acoustic features were then measured from the voice data to determine whether these features can be utilised as diagnostic indicators of depression. The main finding of this study demonstrated one of the acoustic features measured demonstrates significant differences when comparing depressed and healthy individuals.
|Date of Award
|6 Dec 2024
|Monty Adkins (Main Supervisor)