Tobias Raichle

Speech enhancement studies improving the quality of spoken language and finds applications as a front-end in automatic speech recognition, telecommunication or hearing aids. Speech Enhancement models perform well in practical deployments as long as the target data distribution matches the training distribution. However, if there is a distribution shift between the domains, a drop in performance can be observed.

Test-time adaptation (TTA) aims to mitigate this by unsupervised adaptation to the target domain. Whereas TTA has been an active field of research for image classification, it remains underexplored for other use-cases like SE. My research focuses on developing test-time adaptation methods for speech enhancement.