Synth2Aug: Cross-domain speaker recognition with TTS synthesized speech

TL;DR

This paper explores using multi-speaker TTS to generate synthetic speech data for improving cross-domain speaker recognition, especially with limited training speakers, and evaluates the impact of transcript content on synthesis effectiveness.

Contribution

It introduces a TTS-based data augmentation approach for speaker recognition and analyzes the influence of transcript types on synthesis quality and recognition performance.

Findings

TTS synthesis enhances cross-domain speaker recognition accuracy.

Matching transcript content to target domain improves synthesis effectiveness.

Using transcripts with large vocabulary can be a viable alternative when matching is not possible.

Abstract

In recent years, Text-To-Speech (TTS) has been used as a data augmentation technique for speech recognition to help complement inadequacies in the training data. Correspondingly, we investigate the use of a multi-speaker TTS system to synthesize speech in support of speaker recognition. In this study we focus the analysis on tasks where a relatively small number of speakers is available for training. We observe on our datasets that TTS synthesized speech improves cross-domain speaker recognition performance and can be combined effectively with multi-style training. Additionally, we explore the effectiveness of different types of text transcripts used for TTS synthesis. Results suggest that matching the textual content of the target domain is a good practice, and if that is not feasible, a transcript with a sufficiently large vocabulary is recommended.