Domain Adaptation and Autoencoder-Based Unsupervised Speech Enhancement

As a category of transfer learning, domain adaptation plays an important role in generalizing the model trained in one task and applying it to other similar tasks or settings. In speech enhancement, a well-trained acoustic model can be exploited to obtain the speech signal in the context of other languages, speakers, and environments. Recent domain-adaptation research was developed more effectively with various neural networks and high-level abstract features. However, the related studies are more likely to transfer the well-trained model from a rich and more diverse domain to a limited and similar domain. Therefore, in this article, the domain-adaptation method is proposed in unsupervised speech enhancement for the opposite circumstance, that is, transferring to a larger and richer domain. On the one hand, the importance-weighting (IW) approach is exploited with a variance-constrained autoencoder to reduce the shift of shared weights between the source and target domains. On the other hand, in order to train the classifier with the worst-case weights and minimize the risk, the minimax method is proposed. Both the proposed IW and minimax methods are evaluated from the VOICE BANK and IEEE datasets to the TIMIT dataset. The experimental results show that the proposed methods outperform the state-of-the-art approaches.

I. Introduction

In Recent years, machine learning research has been developed and exploited in speech enhancement. In order to solve the tasks in real-world applications, such as hearing aids, machine translation, and robotics, various techniques, including deep learning, reinforcement learning (RL), and transfer learning, have been extensively utilized in the past decade [1]. As the main concept of deep learning, it refers to the hidden layers and neural units of various network models that have been applied in supervised and unsupervised problems. It has significantly improved speech-enhancement performance because of the regression model [2]–[4]. The main target for RL algorithms is to decide a direction or action in different environments, for maximizing the sum of a cumulative reward [5]. Due to this unique principle, the RL has been exploited in computer game design and dialogue management [6].