Anti-Spoofing Using Transfer Learning with Variational Information Bottleneck

View on arXiv ← Back to list

Authors: Youngsik Eom, Yeonghyeon Lee, Ji Sub Um, Hoirin Kim

Published: 2022-04-04 11:08:21+00:00

AI Summary

This paper proposes a transfer learning scheme for speech anti-spoofing using a pre-trained wav2vec 2.0 model and a variational information bottleneck (VIB). The approach improves the performance of distinguishing unseen spoofed and genuine speech, surpassing state-of-the-art systems, particularly in low-resource and cross-dataset settings.

Abstract

Recent advances in sophisticated synthetic speech generated from text-to-speech (TTS) or voice conversion (VC) systems cause threats to the existing automatic speaker verification (ASV) systems. Since such synthetic speech is generated from diverse algorithms, generalization ability with using limited training data is indispensable for a robust anti-spoofing system. In this work, we propose a transfer learning scheme based on the wav2vec 2.0 pretrained model with variational information bottleneck (VIB) for speech anti-spoofing task. Evaluation on the ASVspoof 2019 logical access (LA) database shows that our method improves the performance of distinguishing unseen spoofed and genuine speech, outperforming current state-of-the-art anti-spoofing systems. Furthermore, we show that the proposed system improves performance in low-resource and cross-dataset settings of anti-spoofing task significantly, demonstrating that our system is also robust in terms of data size and data distribution.

Key findings

The proposed system outperforms state-of-the-art anti-spoofing systems on the ASVspoof 2019 LA database. It demonstrates significant improvements in low-resource scenarios, maintaining high performance even with limited training data. The system also shows robustness across different datasets, indicating improved generalization ability.

Approach

The authors leverage a pre-trained wav2vec 2.0 model for speech embedding extraction. A variational information bottleneck (VIB) is then incorporated to regularize the latent representation, improving generalization by suppressing irrelevant information. A simple classifier is used for spoofing detection.

Datasets

ASVspoof 2019 logical access (LA) database, ASVspoof 2015 database, ASVspoof 2021 database

Model(s)

wav2vec 2.0, Multi-layer perceptron (MLP) with variational information bottleneck (VIB)

Author countries

Republic of Korea

← Previous