A Robust Region-Aware Framework for Audio Forgery Localization

doi:10.1109/TASLPRO.2026.3661237

2026-01-01

PRE

DOI：10.1109/TASLPRO.2026.3661237

原文链接

原文求助

摘要

En 中文

含多个精细伪造段的音频定位对音频伪造定位对策构成了显著挑战。当此类音频受到压缩引入的噪声影响，并可能包含未知模型生成的内容时，该挑战进一步加剧。为应对此挑战，我们提出了一种新颖框架，通过捕获不同伪造音频区域间的差异并针对各区域应用不同策略来增强音频伪造定位。具体而言，我们引入了一种基于聚类的专家混合（CMoE）框架，用于动态分配专业专家至伪造音频的不同区域。此外，我们引入了一种可学习掩码模块，以有效从大规模预训练模型中提取鲁棒且泛化的特征。我们还设计了一种组对比损失函数，以增强区分不同音频区域差异的能力。实验结果表明，我们的模型在泛化和鲁棒性方面表现优异，同时能够有效定位不同压缩格式和未知生成模型中的伪造内容。

Keyword:

Feature extraction

Forgery

Location awareness

Robustness

Noise

Reviews

Propagation losses

Graph neural networks

Data mining

Contrastive learning

Audio forgery localization

audio forensics

deep clustering

contrastive learning

期刊

IEEE Transactions on Audio Speech and Language Processing

IF:

论文数:

151

被引数:

机构

sichuan university

学者数:

12.1W

论文数: 7.8W

被引数: 100

引用论文

Learning From Yourself: A Self-Distillation Method For Fake Speech Detection

ICASSP 2023 - 2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

IF0

2023-06-04

OAAI

Jun Xue; Cunhang Fan; Jiangyan Yi; Chenglong Wang; Zhengqi Wen; Dan Zhang; Zhao Lv

Coarse-to-Fine Proposal Refinement Framework for Audio Temporal Forgery Detection and Localization 音频时域伪造检测与定位的粗到精提议优化框架

Proceedings of the 32nd ACM International Conference on Multimedia

IF0

2024-10-28

PREAI

Junyan Wu; Wei Lu; Xiangyang Luo; Rui Yang; Qian Wang; Xiaochun Cao

ASVspoof 2021: Towards Spoofed and Deepfake Speech Detection in the Wild

IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING

IF5.1

2023-01-01

OAAI

Liu, Xuechen; Wang, Xin; Sahidullah, Md; Patino, Jose; Delgado, Hector; Kinnunen, Tomi; Todisco, Massimiliano; Yamagishi, Junichi; Evans, Nicholas; Nautsch, Andreas; Lee, Kong Aik

Integrating frame-level boundary detection and deepfake detection for locating manipulated regions in partially spoofed audio forgery attacks 集成帧级边界检测和deepfake检测，用于在部分欺骗的音频伪造攻击中定位操纵区域

COMPUTER SPEECH AND LANGUAGE

IF3.4

2024-04-01

PREAI

Cai, Zexin; Li, Ming

Toward Improving Synthetic Audio Spoofing Detection Robustness via Meta-Learning and Disentangled Training With Adversarial Examples

IEEE ACCESS

IF3.6

2024-01-01

OAAI

Wang, Zhenyu; Hansen, John H. L.