An Unsupervised Autoregressive Model for Speech Representation Learning

TL;DR

This paper introduces an unsupervised autoregressive neural model that learns versatile speech representations from unlabeled data, enhancing performance on tasks like phone classification and speaker verification.

Contribution

It presents a novel unsupervised autoregressive model that captures multi-level speech information without requiring labeled data, improving downstream task performance.

Findings

Enhanced phone classification accuracy

Improved speaker verification results

Layer-wise analysis reveals different speech information levels

Abstract

This paper proposes a novel unsupervised autoregressive neural model for learning generic speech representations. In contrast to other speech representation learning methods that aim to remove noise or speaker variabilities, ours is designed to preserve information for a wide range of downstream tasks. In addition, the proposed model does not require any phonetic or word boundary labels, allowing the model to benefit from large quantities of unlabeled data. Speech representations learned by our model significantly improve performance on both phone classification and speaker verification over the surface features and other supervised and unsupervised approaches. Further analysis shows that different levels of speech information are captured by our model at different layers. In particular, the lower layers tend to be more discriminative for speakers, while the upper layers provide more phonetic content.