Conformer-Based Self-Supervised Learning for Non-Speech Audio Tasks

TL;DR

This paper introduces a conformer-based self-supervised learning approach for non-speech audio tasks, significantly reducing labeled data needs and achieving state-of-the-art results on AudioSet.

Contribution

It combines wav2vec 2.0 with conformer architectures for effective self-supervised learning on non-speech audio, a less-explored area.

Findings

Achieves a 0.415 mAP on AudioSet, setting a new state-of-the-art.

Reduces labeled data requirement by two-thirds.

Surpasses or matches supervised pre-training performance on multiple tasks.

Abstract

Representation learning from unlabeled data has been of major interest in artificial intelligence research. While self-supervised speech representation learning has been popular in the speech research community, very few works have comprehensively analyzed audio representation learning for non-speech audio tasks. In this paper, we propose a self-supervised audio representation learning method and apply it to a variety of downstream non-speech audio tasks. We combine the well-known wav2vec 2.0 framework, which has shown success in self-supervised learning for speech tasks, with parameter-efficient conformer architectures. Our self-supervised pre-training can reduce the need for labeled data by two-thirds. On the AudioSet benchmark, we achieve a mean average precision (mAP) score of 0.415, which is a new state-of-the-art on this dataset through audio-only self-supervised learning. Our fine-tuned conformers also surpass or match the performance of previous systems pre-trained in a supervised way on several downstream tasks. We further discuss the important design considerations for both pre-training and fine-tuning.