2609.31041

Total: 1

#1 Room Impulse Response Embeddings for Speech Enhancement in Noisy and Reverberant Environments [PDF] [Copy] [Kimi] [REL]

Authors: Adrian Meise, Reinhold Haeb-Umbach

We propose a self-supervised approach for learning room impulse response (RIR) representations from single-channel noisy-reverberant speech. It consists of first training on reverberant data, then on noisy-reverberant data, and finally with a teacher-student approach, where the student learns to replicate the teacher's embeddings when given a noisy version of the reverberant input. We assess their representational capabilities by estimating acoustic room parameters from them. Conditioning a discriminative speech enhancement model on the derived embeddings yields consistent gains across all evaluated metrics, including downstream word error rate, for both reverberant and noisy-reverberant speech.

Subjects: Audio and Speech Processing , Sound

Publish: 2026-09-25 09:33:19 UTC