Spatial Reconstructed Local Attention Res2Net with F0 Subband for Fake Speech Detection

View on arXiv ← Back to list

Authors: Cunhang Fan, Jun Xue, Jianhua Tao, Jiangyan Yi, Chenglong Wang, Chengshi Zheng, Zhao Lv

Published: 2023-08-19 08:31:04+00:00

AI Summary

This paper proposes a novel fake speech detection method using an F0 subband and a spatial reconstructed local attention Res2Net (SR-LA Res2Net) architecture. The method leverages the discriminative information in the fundamental frequency (F0) subband, effectively modeled by SR-LA Res2Net to achieve state-of-the-art performance on the ASVspoof 2019 LA dataset.

Abstract

The rhythm of bonafide speech is often difficult to replicate, which causes that the fundamental frequency (F0) of synthetic speech is significantly different from that of real speech. It is expected that the F0 feature contains the discriminative information for the fake speech detection (FSD) task. In this paper, we propose a novel F0 subband for FSD. In addition, to effectively model the F0 subband so as to improve the performance of FSD, the spatial reconstructed local attention Res2Net (SR-LA Res2Net) is proposed. Specifically, Res2Net is used as a backbone network to obtain multiscale information, and enhanced with a spatial reconstruction mechanism to avoid losing important information when the channel group is constantly superimposed. In addition, local attention is designed to make the model focus on the local information of the F0 subband. Experimental results on the ASVspoof 2019 LA dataset show that our proposed method obtains an equal error rate (EER) of 0.47% and a minimum tandem detection cost function (min t-DCF) of 0.0159, achieving the state-of-the-art performance among all of the single systems.

Key findings

The proposed method achieved state-of-the-art performance on the ASVspoof 2019 LA dataset with an equal error rate (EER) of 0.47% and a minimum tandem detection cost function (min t-DCF) of 0.0159. The F0 subband proved highly effective, outperforming other frequency bands, and the SR-LA Res2Net architecture significantly improved generalization to unseen attacks.

Approach

The authors propose using the 0-400 Hz frequency band of the log power spectrogram as an F0 subband for fake speech detection. This subband is then fed into a modified Res2Net architecture enhanced with spatial reconstruction and local attention blocks to improve feature extraction and reduce redundant information.

Datasets

ASVspoof 2019 LA dataset, ASVspoof 2021 LA dataset

Model(s)

Spatial Reconstructed Local Attention Res2Net (SR-LA Res2Net), Res2Net, ResNet, SENet, ACNN, MCG-Res2Net, LCNN

Author countries

China

← Previous