SpoofCeleb

该数据集名为SpoofCeleb，专为语音深度伪造检测（SDD）和抗欺骗自动说话人验证（SASV）任务设计。它包含了超过250万条来自1251位独特说话人的语音样本，这些样本是在自然真实的条件下收集的。该数据集还精心划分了训练集、验证集和评估集，并配备了严格控制的实验协议。规模上，它拥有来自1251位说话人的超过250万条语音样本。这项工作的任务旨在处理语音深度伪造检测和抗欺骗自动说话人验证。

This dataset, named SpoofCeleb, is specifically designed for speech deepfake detection (SDD) and spoofing-resistant automatic speaker verification (SASV) tasks. It contains over 2.5 million speech samples from 1251 unique speakers collected under natural and realistic conditions. The dataset has been meticulously partitioned into training, validation and evaluation sets, and is equipped with strictly controlled experimental protocols. With a total scale of over 2.5 million speech samples sourced from 1251 distinct speakers, this work focuses on speech deepfake detection and spoofing-resistant automatic speaker verification.

搜集汇总

背景与挑战

背景概述

SpoofCeleb是一个专注于语音深度伪造检测和说话人验证的数据集，基于TITW数据集构建，包含真实人类语音和23个TTS系统生成的合成语音。数据集通过Hugging Face平台提供访问，需申请并遵守Creative Commons Attribution 4.0 International License许可，适用于语音安全研究。