Speaker Representation Learning: Theories, Applications and Practice

Abstract

This tutorial aims to provide an in-depth explanation of the basic theories and mainstream methods of speaker representation learning, which covers conventional supervised schemes, self-supervised schemes and strategies based on pre-trained models. At the same time, we will also discuss key research topics such as the robustness and efficiency of speaker representation learning. Further, we will introduce the application and customization strategies of speaker representation in multiple related tasks, including speaker identification, speaker logging, speech generation, and target speaker extraction. Finally, we will take the wespeaker toolkit as an example and introduce it from a more practical perspective that is closer to the actual needs of the industry.