A Closer Look at Audio-Visual Multi-Person Speech Recognition and Active Speaker Selection

TL;DR

This paper investigates audio-visual multi-person speech recognition and active speaker selection, demonstrating that an end-to-end attention-based model can match larger two-step systems in accuracy across noisy conditions.

Contribution

It provides a detailed analysis of the attention mechanism's role in speaker selection and compares end-to-end models with traditional systems using extensive YouTube data.

Findings

Attention indirectly learns face-audio association.

End-to-end model performs as well as larger two-step systems.

Model maintains accuracy across various noise levels.

Abstract

Audio-visual automatic speech recognition is a promising approach to robust ASR under noisy conditions. However, up until recently it had been traditionally studied in isolation assuming the video of a single speaking face matches the audio, and selecting the active speaker at inference time when multiple people are on screen was put aside as a separate problem. As an alternative, recent work has proposed to address the two problems simultaneously with an attention mechanism, baking the speaker selection problem directly into a fully differentiable model. One interesting finding was that the attention indirectly learns the association between the audio and the speaking face even though this correspondence is never explicitly provided at training time. In the present work we further investigate this connection and examine the interplay between the two problems. With experiments involving over 50 thousand hours of public YouTube videos as training data, we first evaluate the accuracy of the attention layer on an active speaker selection task. Secondly, we show under closer scrutiny that an end-to-end model performs at least as well as a considerably larger two-step system that utilizes a hard decision boundary under various noise conditions and number of parallel face tracks.