BUT/JHU System Description for CHiME-8 NOTSOFAR-1 Challenge

BUT/JHU System Description for CHiME-8 NOTSOFAR-1 Challenge

Paper in proceedings outside WoS and Scopus

This paper presents our method for tackling the CHIME-8 chal- lenge's NOTSOFAR-1 task, which requires participants to per- form multi-speaker automatic speech recognition (ASR) using audio from distant microphone arrays. We modify the Pyan- note3 diarization pipeline, incorporating pre-trained WavLM as local EEND to adapt effectively to new domains, and we intro- duce two diarization-aware approaches to ASR by condition- ing Whisper on diarization outputs for target-speaker ASR. The first method, which we refer to as Query-Key Biasing, modi- fies Whisper's attention mechanism and positional embeddings with a learnable attention mask to exclude non-target speaker segments in the audio. The second method, called Frame- Level Diarization-Dependent Transformations, applies affine, diarization-dependent transformations with trainable parame- ters to the inputs of one or more transformer blocks. We also extend both the ASR and diarization systems to a multichannel setup by incorporating cross-channel communication into our models. Finally, we report the performance of these approaches on the NOTSOFAR-1 dataset.

This paper presents our method for tackling the CHIME-8 chal- lenge's NOTSOFAR-1 task, which requires participants to per- form multi-speaker automatic speech recognition (ASR) using audio from distant microphone arrays. We modify the Pyan- note3 diarization pipeline, incorporating pre-trained WavLM as local EEND to adapt effectively to new domains, and we intro- duce two diarization-aware approaches to ASR by condition- ing Whisper on diarization outputs for target-speaker ASR. The first method, which we refer to as Query-Key Biasing, modi- fies Whisper's attention mechanism and positional embeddings with a learnable attention mask to exclude non-target speaker segments in the audio. The second method, called Frame- Level Diarization-Dependent Transformations, applies affine, diarization-dependent transformations with trainable parame- ters to the inputs of one or more transformer blocks. We also extend both the ASR and diarization systems to a multichannel setup by incorporating cross-channel communication into our models. Finally, we report the performance of these approaches on the NOTSOFAR-1 dataset.

multi-talker speech recognition, CHiME-8, NOTSOFAR-1, target-speaker

multi-talker speech recognition, CHiME-8, NOTSOFAR-1, target-speaker

POLOK, A.; KLEMENT, D.; HAN, J.; SEDLÁČEK, Š.; YUSUF, B.; MACIEJEWSKI, M.; WIESNER, M.; BURGET, L.

06.09.2024

International Speech Communication Association

Kos Island

Proceedings of CHiME 2024 Workshop

18

22

5

https://www.isca-archive.org/chime_2024/polok24_chime.pdf

polok24_chime_at Interspeech