VC-AUG: Voice Conversion Based Data Augmentation for Text-Dependent Speaker Verification

In this paper, we focus on improving the performance of the text-dependent speaker verification system in the scenario of limited training data. The deep learning based text-dependent speaker verification system generally needs a large-scale text-dependent training data set which could be both labor and cost expensive, especially for customized new wake-up words. In recent studies, voice conversion systems that can generate high quality synthesized speech of seen and unseen speakers have been proposed. Inspired by those works, we adopt two different voice conversion methods as well as the very simple re-sampling approach to generate new text-dependent speech samples for data augmentation purposes. Experimental results show that the proposed method significantly improves the Equal Error Rate performance from 6.51% to 4.48% in the scenario of limited training data. In addition, we also explore the out-of-set and unseen speaker voice conversion based data augmentation.