On the Transferability of Whisper-based Representations for "In-the-Wild" Cross-Task Downstream Speech Applications

TL;DR

This paper investigates the transferability and robustness of Whisper's speech representations across various real-world tasks and noisy environments, demonstrating its potential for practical, cross-task speech applications.

Contribution

It is the first comprehensive study evaluating Whisper's representations beyond ASR, including robustness in noisy and reverberant conditions across multiple speech tasks.

Findings

Whisper achieves promising results across multiple tasks.

Whisper's representations are robust in noisy and reverberant environments.

Potential for real-world deployment of Whisper-based systems.

Abstract

Large self-supervised pre-trained speech models have achieved remarkable success across various speech-processing tasks. The self-supervised training of these models leads to universal speech representations that can be used for different downstream tasks, ranging from automatic speech recognition (ASR) to speaker identification. Recently, Whisper, a transformer-based model was proposed and trained on large amount of weakly supervised data for ASR; it outperformed several state-of-the-art self-supervised models. Given the superiority of Whisper for ASR, in this paper we explore the transferability of the representation for four other speech tasks in SUPERB benchmark. Moreover, we explore the robustness of Whisper representation for ``in the wild'' tasks where speech is corrupted by environment noise and room reverberation. Experimental results show Whisper achieves promising results across tasks and environmental conditions, thus showing potential for cross-task real-world deployment.