Streaming Multi-speaker ASR with RNN-T
This work provides a solution for real-time multi-speaker ASR, which is critical for improving the performance and usability of voice assistants in environments with multiple speakers. It is an incremental improvement over existing non-streaming methods.
This paper addresses the challenge of multi-speaker automatic speech recognition (ASR) in streaming scenarios, which is crucial for voice assistants. The authors developed an RNN-T based system that achieves a Word Error Rate (WER) of 10.2% on simulated 2-speaker LibriSpeech data, matching the performance of non-streaming state-of-the-art models.
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.