Galjaard, Ellemijn (2023) A Self-Supervised Approach to Speech Enhancement in Noisy Climbing Gym Environments. Master thesis, Voice Technology (VT).
|
PDF
MA 5397774 EHW Galjaard.pdf Download (3MB) | Preview |
Abstract
Existing speech enhancement models struggle to generalize to diverse acoustic environments with unfamiliar noise types. The acoustic environment of a climbing gym presents a particularly interesting challenge to speech enhancement models due to high levels of complex ambient noise. Therefore, this study investigates the effectiveness of a self-supervised speech enhancement model in removing climbing gym noise from speech signals. In order to achieve this goal, a range of different experiments are conducted which consider various factors that could have an influence on the model’s effectiveness, such as variations in training data and the inclusion of the audio signal’s phase information during model training. Despite the inconclusive results obtained, this study provides valuable insights into the complexities of speech enhancement tasks. Furthermore, it identifies potential areas for future research that can contribute to developing more effective speech enhancement algorithms for challenging noisy environments.
| Item Type: | Thesis (Master) |
|---|---|
| Name supervisor: | Nayak, S. |
| Date Deposited: | 12 Sep 2023 11:08 |
| Last Modified: | 12 Sep 2023 11:08 |
| URI: | https://campus-fryslan.studenttheses.ub.rug.nl/id/eprint/365 |
Actions (login required)
 |
View Item |