Unsupervised Learning of Disentangled Speech Content and Style Representation

TL;DR

This paper introduces an unsupervised model that disentangles speech content and style, enabling effective speech recognition and speaker identification with minimal labeled data.

Contribution

The paper proposes a novel unsupervised speech representation learning model with separate local and global encoders for content and style, respectively.

Findings

Local latent variables encode speech content with low WER in ASR.

Global latent variables encode speaker style accurately.

Pre-trained global representations enable high-accuracy speaker recognition with minimal data.

Abstract

We present an approach for unsupervised learning of speech representation disentangling contents and styles. Our model consists of: (1) a local encoder that captures per-frame information; (2) a global encoder that captures per-utterance information; and (3) a conditional decoder that reconstructs speech given local and global latent variables. Our experiments show that (1) the local latent variables encode speech contents, as reconstructed speech can be recognized by ASR with low word error rates (WER), even with a different global encoding; (2) the global latent variables encode speaker style, as reconstructed speech shares speaker identity with the source utterance of the global encoding. Additionally, we demonstrate an useful application from our pre-trained model, where we can train a speaker recognition model from the global latent variables and achieve high accuracy by fine-tuning with as few data as one label per speaker.