Fast and efficient speech enhancement with variational autoencoders

TL;DR

This paper introduces a Langevin dynamics-based variational autoencoder method for speech enhancement that balances computational efficiency with high-quality results, outperforming existing approaches.

Contribution

It proposes a novel Langevin dynamics approach with total variation regularization for variational autoencoders in speech enhancement, improving efficiency and performance.

Findings

Outperforms existing speech enhancement methods

Balances computational efficiency with enhancement quality

Uses Langevin dynamics with temporal regularization

Abstract

Unsupervised speech enhancement based on variational autoencoders has shown promising performance compared with the commonly used supervised methods. This approach involves the use of a pre-trained deep speech prior along with a parametric noise model, where the noise parameters are learned from the noisy speech signal with an expectationmaximization (EM)-based method. The E-step involves an intractable latent posterior distribution. Existing algorithms to solve this step are either based on computationally heavy Monte Carlo Markov Chain sampling methods and variational inference, or inefficient optimization-based methods. In this paper, we propose a new approach based on Langevin dynamics that generates multiple sequences of samples and comes with a total variation-based regularization to incorporate temporal correlations of latent vectors. Our experiments demonstrate that the developed framework makes an effective compromise between computational efficiency and enhancement quality, and outperforms existing methods.