# Transfer Learning from Audio-Visual Grounding to Speech Recognition

## TL;DR

This paper introduces a transfer learning approach that leverages audio-visual grounding models to extract phonetic features for speech recognition without using textual transcripts, demonstrating robustness and domain invariance.

## Contribution

It proposes a novel transfer learning scenario using grounding models trained on audio-visual data to improve speech recognition, without relying on speech transcripts during training.

## Key findings

- Layer proximity to input retains more phonetic information.
- Deeper layers show greater invariance to domain shifts.
- Grounding models trained without speech data can generalize to speech recognition.

## Abstract

Transfer learning aims to reduce the amount of data required to excel at a new task by re-using the knowledge acquired from learning other related tasks. This paper proposes a novel transfer learning scenario, which distills robust phonetic features from grounding models that are trained to tell whether a pair of image and speech are semantically correlated, without using any textual transcripts. As semantics of speech are largely determined by its lexical content, grounding models learn to preserve phonetic information while disregarding uncorrelated factors, such as speaker and channel. To study the properties of features distilled from different layers, we use them as input separately to train multiple speech recognition models. Empirical results demonstrate that layers closer to input retain more phonetic information, while following layers exhibit greater invariance to domain shift. Moreover, while most previous studies include training data for speech recognition for feature extractor training, our grounding models are not trained on any of those data, indicating more universal applicability to new domains.