A speaker diarization system that adds a shared-and-soft mixture-of-experts layer to a memory-augmented sequence-to-sequence model reports lower error rates on CHiME-6, DiPCo, and Mixer 6, with some DIHARD-III claims contradicted by its own tables.
NTT Multi-Speaker ASR System for the DASR Task of CHiME-8 Challenge
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
abstract
We present a distant automatic speech recognition (DASR) system developed for the CHiME-8 DASR track. It consists of a diarization first pipeline. For diarization, we use end-to-end diarization with vector clustering (EEND-VC) followed by target speaker voice activity detection (TS-VAD) refinement. To deal with various numbers of speakers, we developed a new multi-channel speaker counting approach. We then apply guided source separation (GSS) with several improvements to the baseline system. Finally, we perform ASR using a combination of systems built from strong pre-trained models. Our proposed system achieves a macro tcpWER of 21.3 % on the dev set, which is a 57 % relative improvement over the baseline.
fields
cs.SD 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Exploring Speaker Diarization with Mixture of Experts
A speaker diarization system that adds a shared-and-soft mixture-of-experts layer to a memory-augmented sequence-to-sequence model reports lower error rates on CHiME-6, DiPCo, and Mixer 6, with some DIHARD-III claims contradicted by its own tables.