Unsupervised Speech Decomposition via Triple Information Bottleneck

TL;DR

This paper introduces SpeechSplit, an unsupervised method that decomposes speech into content, timbre, pitch, and rhythm components using information bottlenecks, enabling style transfer without explicit annotations.

Contribution

SpeechSplit is among the first algorithms to blindly decompose speech into four components and perform style transfer without text labels, advancing speech representation disentanglement.

Findings

Successfully separates speech into four components without labels

Enables style transfer for timbre, pitch, and rhythm

Achieves competitive results in speech decomposition tasks

Abstract

Speech information can be roughly decomposed into four components: language content, timbre, pitch, and rhythm. Obtaining disentangled representations of these components is useful in many speech analysis and generation applications. Recently, state-of-the-art voice conversion systems have led to speech representations that can disentangle speaker-dependent and independent information. However, these systems can only disentangle timbre, while information about pitch, rhythm and content is still mixed together. Further disentangling the remaining speech components is an under-determined problem in the absence of explicit annotations for each component, which are difficult and expensive to obtain. In this paper, we propose SpeechSplit, which can blindly decompose speech into its four components by introducing three carefully designed information bottlenecks. SpeechSplit is among the first algorithms that can separately perform style transfer on timbre, pitch and rhythm without text labels. Our code is publicly available at https://github.com/auspicious3000/SpeechSplit.