Joint Speaker Counting, Speech Recognition, and Speaker Identification for Overlapped Speech of Any Number of Speakers

TL;DR

This paper introduces an end-to-end model that simultaneously performs speaker counting, speech recognition, and speaker identification on overlapped speech, improving accuracy over separate methods.

Contribution

The authors extend serialized output training with a speaker inventory and joint optimization, enabling unified processing of overlapped speech with multiple speakers.

Findings

Significantly improved speaker-attributed word error rate

Effective joint modeling of speech recognition and speaker ID

Outperforms baseline methods on LibriSpeech

Abstract

We propose an end-to-end speaker-attributed automatic speech recognition model that unifies speaker counting, speech recognition, and speaker identification on monaural overlapped speech. Our model is built on serialized output training (SOT) with attention-based encoder-decoder, a recently proposed method for recognizing overlapped speech comprising an arbitrary number of speakers. We extend SOT by introducing a speaker inventory as an auxiliary input to produce speaker labels as well as multi-speaker transcriptions. All model parameters are optimized by speaker-attributed maximum mutual information criterion, which represents a joint probability for overlapped speech recognition and speaker identification. Experiments on LibriSpeech corpus show that our proposed method achieves significantly better speaker-attributed word error rate than the baseline that separately performs overlapped speech recognition and speaker identification.