Streaming Multi-speaker ASR with RNN-T

Abstract

Recent research shows end-to-end ASR systems can recognize overlapped speech from multiple speakers. However, all published works have assumed no latency constraints during inference, which does not hold for most voice assistant interactions. This work focuses on multi-speaker speech recognition based on a recurrent neural network transducer (RNN-T) that has been shown to provide high recognition accuracy at a low latency online recognition regime. We investigate two approaches to multi-speaker model training of the RNN-T: deterministic output-target assignment and permutation invariant training. We show that guiding separation with speaker order labels in the former case enhances the high-level speaker tracking capability of RNN-T. Apart from that, with multistyle training on single-and multi-speaker utterances, the resulting models gain robustness against ambiguous numbers of speakers during inference. Our best model achieves a WER of 10.2% on simulated 2-speaker LibriSpeech data, which is competitive with the previously reported state-of-the-art nonstreaming model (10.3%), while the proposed model could be directly applied for streaming applications.

Large-Scale Pre-Training of End-to-End Multi-Talker ASR for Meeting Transcription with Single Distant Microphone

“…To overcome this suboptimality, there has been a series of studies for multi-talker ASR that directly transcribes multiple utterances from overlapped speech. One popular approach is using a neural network that has multiple output layers, each of which recognizes one speaker [16,17,18,19,20,21,22]. Permutation invariant training (PIT) [23] is usually used to train such a multiple-output models.…”

“…While promising results were shown, all the previous studies were limited to either simulated data [16,17,18,19,20,21,22,24,25,26,27] or small-scale real data [7,9,8,10]. It is due to the difficulty in collecting real meeting recordings with precise transcriptions at large-scale.…”