Posterior sampling algorithms for unsupervised speech enhancement with recurrent variational autoencoder
This work addresses computational bottlenecks in unsupervised speech enhancement, offering incremental improvements for applications requiring efficient real-time processing.
The paper tackles the high computational complexity of iterative variational expectation-maximization in unsupervised speech enhancement with recurrent variational autoencoders by introducing efficient sampling techniques based on Langevin dynamics and Metropolis-Hasting algorithms. The results show that these methods significantly outperform the variational approach in computational efficiency and overall performance, and demonstrate robust generalization compared to a supervised baseline in mismatched conditions.
In this paper, we address the unsupervised speech enhancement problem based on recurrent variational autoencoder (RVAE). This approach offers promising generalization performance over the supervised counterpart. Nevertheless, the involved iterative variational expectation-maximization (VEM) process at test time, which relies on a variational inference method, results in high computational complexity. To tackle this issue, we present efficient sampling techniques based on Langevin dynamics and Metropolis-Hasting algorithms, adapted to the EM-based speech enhancement with RVAE. By directly sampling from the intractable posterior distribution within the EM process, we circumvent the intricacies of variational inference. We conduct a series of experiments, comparing the proposed methods with VEM and a state-of-the-art supervised speech enhancement approach based on diffusion models. The results reveal that our sampling-based algorithms significantly outperform VEM, not only in terms of computational efficiency but also in overall performance. Furthermore, when compared to the supervised baseline, our methods showcase robust generalization performance in mismatched test conditions.