Microphone array geometry-independent multi-talker distant ASR: NTT system for DASR task of the CHiME-8 challenge

Download PDF
Open Webpage

Naoyuki Kamo, Naohiro Tawara, Atsushi Ando, Takatomo Kano, Hiroshi Sato, Rintaro Ikeshita, Takafumi Moriya, Shota Horiguchi, Kohei Matsuura, Atsunori Ogawa, Alexis Plaquet, Takanori Ashihara, Tsubasa Ochiai, Masato Mimura, Marc Delcroix, Tomohiro Nakatani, Taichi Asami, Shoko Araki

Published: 2026, Last Modified: 02 Feb 2026Comput. Speech Lang. 2026EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•A multi-talker ASR system achieving 63% relative macro tcpWER improvement over the CHiME-8 DASR task baseline.•A powerful diarization frontend combining EEND-VC, TS-VAD, and multi-channel speaker counting.•Speech enhancement using improved microphone selection and SP-MWF beamformer.•Four ASR backends exploiting speech foundation models (Whisper and WavLM).•An extensive experimental study and ablation analyzing each component of our system.