Streaming Audio-Visual Speech Recognition with Alignment Regularization

TL;DR

This paper introduces a streaming audio-visual speech recognition system using a hybrid CTC/attention model with conformer encoders, achieving state-of-the-art results on the LRS3 dataset without external data.

Contribution

It presents a novel alignment regularization technique for synchronized audio-visual encoding and a streaming recognition method with triggered attention.

Findings

Achieves 2.0% WER offline and 2.6% WER online on LRS3 dataset.

Introduces a streamable conformer-based AV encoder with alignment regularization.

Sets new state-of-the-art results without external training data.

Abstract

In this work, we propose a streaming AV-ASR system based on a hybrid connectionist temporal classification (CTC)/attention neural network architecture. The audio and the visual encoder neural networks are both based on the conformer architecture, which is made streamable using chunk-wise self-attention (CSA) and causal convolution. Streaming recognition with a decoder neural network is realized by using the triggered attention technique, which performs time-synchronous decoding with joint CTC/attention scoring. Additionally, we propose a novel alignment regularization technique that promotes synchronization of the audio and visual encoder, which in turn results in better word error rates (WERs) at all SNR levels for streaming and offline AV-ASR models. The proposed AV-ASR model achieves WERs of 2.0% and 2.6% on the Lip Reading Sentences 3 (LRS3) dataset in an offline and online setup, respectively, which both present state-of-the-art results when no external training data are used.