End-to-end Multichannel Speaker-Attributed ASR: Speaker Guided Decoder and Input Feature Analysis - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2023

End-to-end Multichannel Speaker-Attributed ASR: Speaker Guided Decoder and Input Feature Analysis

Abstract

We present an end-to-end multichannel speaker-attributed automatic speech recognition (MC-SA-ASR) system that combines a Conformer-based encoder with multi-frame crosschannel attention and a speaker-attributed Transformer-based decoder. To the best of our knowledge, this is the first model that efficiently integrates ASR and speaker identification modules in a multichannel setting. On simulated mixtures of LibriSpeech data, our system reduces the word error rate (WER) by up to 12% and 16% relative compared to previously proposed single-channel and multichannel approaches, respectively. Furthermore, we investigate the impact of different input features, including multichannel magnitude and phase information, on the ASR performance. Finally, our experiments on the AMI corpus confirm the effectiveness of our system for real-world multichannel meeting transcription.
Fichier principal
Vignette du fichier
Template_Blind.pdf (505.26 Ko) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-04235774 , version 1 (11-10-2023)

Licence

Identifiers

Cite

Can Cui, Imran Ahamad Sheikh, Mostafa Sadeghi, Emmanuel Vincent. End-to-end Multichannel Speaker-Attributed ASR: Speaker Guided Decoder and Input Feature Analysis. 2023 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU 2023), Dec 2023, Taipei, Taiwan. ⟨10.1109/ASRU57964.2023.10389729⟩. ⟨hal-04235774⟩
189 View
99 Download

Altmetric

Share

Gmail Mastodon Facebook X LinkedIn More