VoiceExtender: Short-utterance Text-independent Speaker Verification with Guided Diffusion Model

Yayun He, Zuheng Kang, Jianzong Wang, Junqing Peng, Jing Xiao
September 2023
Cite arXiv IEEE
The speaker embedding guidance diffusion model overview

Abstract

Speaker Verification (SV) performance gets worse as utterances get shorter. To this end, we propose a new architecture called VoiceExtender which has two implementations and provides a promising solution for improving SV performance when handling short-duration speech signals. We used two bootstrap diffusion models, the built-in speaker embedding (SE) and the external-SE guidance diffusion model, both of which utilize a diffusion model-based sample generator that leverages SE guidance to augment the speech features based on a short utterance. Extensive experimental results on the Voxceleb1 dataset show that our method outperforms the baseline, with relative improvements in Equal Error Rate (EER) of 46.1%, 35.7%, 10.4%, and 5.7% for the short accent conditions of 0.5, 1.0, 1.5, and 2.0 seconds, respectively.

Type
1
Publication
In IEEE Workshop on Automatic Speech Recognition and Understanding 2023
Click the Cite button above to demo the feature to enable visitors to import publication metadata into their reference management software.
Speech
Yayun He
Yayun He
Researcher
Zuheng Kang
Zuheng Kang
Researcher
Jianzong Wang
Jianzong Wang
Honorary Director

[email protected]