A Comparison of Discrete Latent Variable Models for Speech Representation Learning
Audio and Speech Processing
2020-10-28 v1 Artificial Intelligence
Machine Learning
Sound
Abstract
Neural latent variable models enable the discovery of interesting structure in speech audio data. This paper presents a comparison of two different approaches which are broadly based on predicting future time-steps or auto-encoding the input signal. Our study compares the representations learned by vq-vae and vq-wav2vec in terms of sub-word unit discovery and phoneme recognition performance. Results show that future time-step prediction with vq-wav2vec achieves better performance. The best system achieves an error rate of 13.22 on the ZeroSpeech 2019 ABX phoneme discrimination challenge
Cite
@article{arxiv.2010.14230,
title = {A Comparison of Discrete Latent Variable Models for Speech Representation Learning},
author = {Henry Zhou and Alexei Baevski and Michael Auli},
journal= {arXiv preprint arXiv:2010.14230},
year = {2020}
}
Comments
7 pages, 4 figures