End-to-end Multichannel Speaker-Attributed ASR: Speaker Guided Decoder and Input Feature Analysis
End-to-end Multichannel Speaker-Attributed ASR: Speaker Guided Decoder and Input Feature Analysis
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.
Imran Ahamad Sheikh、Emmanuel Vincent、Can Cui、Mostafa Sadeghi
MULTISPEECHMULTISPEECHMULTISPEECHMULTISPEECH
通信无线通信计算技术、计算机技术
Imran Ahamad Sheikh,Emmanuel Vincent,Can Cui,Mostafa Sadeghi.End-to-end Multichannel Speaker-Attributed ASR: Speaker Guided Decoder and Input Feature Analysis[EB/OL].(2023-10-16)[2025-07-09].https://arxiv.org/abs/2310.10106.点此复制
评论