Study compares generative vs discriminative speech enhancement methods

Arxiv CS.SD (Sound/Audio ML)

The paper evaluates generative and discriminative deep learning methods for speech enhancement, focusing on noise reduction tasks. Key aspects compared include robustness, computational complexity, and hallucination artifacts.

speech-enhancementdeep-learning

Arxiv CS.SD (Sound/Audio ML)·Shrishti Saha ShetuEmanu\"el A. P. HabetsAndreas Brendel

8 days ago