Spijkerman, Marjolein (2022) Using voice conversion and time-stretching to enhance the quality of dysarthric speech for automatic speech recognition. Master thesis, Voice Technology (VT).
|
PDF
Msc 3219305 M Spijkerman.pdf Download (544kB) | Preview |
Abstract
Dysarthria is a motor-speech disorder that is caused by damage to the central or peripheral nervous system. It affects the muscular control over the speech mechanism. This can cause communication issues. To help people to engage in conversation and to use assistive devices a voice conversion system is used to enhance the quality of the dysarthric speech. Non-parallel voice conversion methods are applied in the form of a MaskCycleGAN-based model, this is combined with time-stretching methods to account for the slowness of spastic dysarthric speech. The results indicate that applying the voice conversion and the time-stretching does improve the performance, however the actual performance is still quite low.
| Item Type: | Thesis (Master) |
|---|---|
| Name supervisor: | Verkhodanova, V. |
| Date Deposited: | 09 Sep 2022 08:50 |
| Last Modified: | 09 Sep 2022 08:50 |
| URI: | https://campus-fryslan.studenttheses.ub.rug.nl/id/eprint/222 |
Actions (login required)
 |
View Item |