Analyzing Speaker Information in Self-Supervised Models to Improve Zero-Resource Speech Processing

TL;DR

This paper investigates how contrastive predictive coding (CPC) features encode speaker information and proposes normalization techniques to improve zero-resource speech processing, achieving state-of-the-art results in the ZeroSpeech2021 Challenge.

Contribution

It reveals that CPC features contain speaker information and introduces a normalization method to enhance acoustic unit discovery for zero-resource speech tasks.

Findings

Per-utterance mean of CPC features captures speaker info

Standardizing features removes speaker information

Normalization improves zero-resource speech task performance

Abstract

Contrastive predictive coding (CPC) aims to learn representations of speech by distinguishing future observations from a set of negative examples. Previous work has shown that linear classifiers trained on CPC features can accurately predict speaker and phone labels. However, it is unclear how the features actually capture speaker and phonetic information, and whether it is possible to normalize out the irrelevant details (depending on the downstream task). In this paper, we first show that the per-utterance mean of CPC features captures speaker information to a large extent. Concretely, we find that comparing means performs well on a speaker verification task. Next, probing experiments show that standardizing the features effectively removes speaker information. Based on this observation, we propose a speaker normalization step to improve acoustic unit discovery using K-means clustering of CPC features. Finally, we show that a language model trained on the resulting units achieves some of the best results in the ZeroSpeech2021~Challenge.