This paper describes the STCON system for the CHiME-8 Challenge Task 1 (DASR) aimed at distant automatic speech transcription and diarization with multiple recording devices. Our main attention was paid to carefully trained and tuned diarization pipeline and speaker counting. This allowed to significantly reduce diarization error rate (DER) and obtain more reliable segments for speech separation and recognition. To improve source separation, we designed a Guided Target speaker Extraction (G-TSE) model and used it in conjunction with the traditional Guided Source Separation (GSS) method. To train various parts of our pipeline, we investigated several data augmentation and generation techniques, which helped us to improve the overall system quality.
Cite
@article{arxiv.2410.13411,
title = {STCON System for the CHiME-8 Challenge},
author = {Anton Mitrofanov and Tatiana Prisyach and Tatiana Timofeeva and Sergei Novoselov and Maxim Korenevsky and Yuri Khokhlov and Artem Akulov and Alexander Anikin and Roman Khalili and Iurii Lezhenin and Aleksandr Melnikov and Dmitriy Miroshnichenko and Nikita Mamaev and Ilya Odegov and Olga Rudnitskaya and Aleksei Romanenko},
journal= {arXiv preprint arXiv:2410.13411},
year = {2024}
}