AI RESEARCH
CleanCodec: Efficient and Robust Speech Tokenization via Perceptually Guided Encoding
arXiv CS.CL
•
ArXi:2606.04418v1 Announce Type: cross Neural audio codecs are a key component of speech processing pipelines, compressing audio into discrete tokens for downstream modeling. However, existing codecs struggle to balance reconstruction quality with token efficiency, often encoding perceptually irrelevant information such as background noise and recording artifacts at the expense of linguistically and acoustically meaningful content.