ASVspoof2019_LA
收藏资源简介:
ASVspoof 2019 LA数据集是ASVspoof 2019挑战赛中逻辑访问(LA)评估分区的基准就绪版本,专门用于语音反欺骗和合成语音检测研究。该数据集的核心任务是二元分类:区分真实的人类语音(bonafide)与合成或转换的伪造语音(spoof)。数据集包含总计71,237个语音样本,其中真实语音样本7,355个,伪造语音样本63,882个。每个样本由四个字段构成:`path`(稳定的归档相对路径)、`audio`(16 kHz采样率的单声道音频波形)、`label`(分类标签,0对应‘bonafide’,1对应‘spoof’)以及`notes`(一个包含`utterance_id`、`speaker_id`和`subset`信息的JSON字符串,用于标识话语、说话人及所属子集)。该数据集以Parquet格式提供,适用于训练和评估语音反欺骗模型,是音频深度伪造检测领域的一个重要基准数据集。
The ASVspoof 2019 LA dataset is a benchmark-ready version of the logical access (LA) evaluation partition from the ASVspoof 2019 challenge, specifically designed for research in speech anti-spoofing and synthetic speech detection. The core task is binary classification: distinguishing genuine human speech (bonafide) from synthetic or converted spoofed speech. It contains a total of 71,237 speech samples, with 7,355 bonafide samples and 63,882 spoof samples. Each sample consists of four fields: `path` (stable archive relative path), `audio` (mono audio waveform sampled at 16 kHz), `label` (classification label, 0 for bonafide, 1 for spoof), and `notes` (a JSON string containing `utterance_id`, `speaker_id`, and `subset` information to identify the utterance, speaker, and subset). The dataset is provided in Parquet format, suitable for training and evaluating speech anti-spoofing models, and serves as an important benchmark in the field of audio deepfake detection.
ASVspoof 2019 LA 数据集详情
基本信息
- 数据集名称: ASVspoof 2019 LA
- 许可证: Open Data Commons Attribution License (ODC-By)
- 语言: 英语
- 任务类别: 音频分类(audio-classification)
- 规模: 10K < n < 100K
- 标签: 反欺骗、音频深度伪造检测、语音、基准测试、Arena-ready
- PapersWithCode ID: asvspoof-2019
- 相关论文: arXiv: 1911.01601
数据集概述
该数据集是ASVspoof 2019挑战赛中逻辑访问(Logical Access, LA)评估分区的基准测试封装版本,用于语音反欺骗和合成语音检测。任务为二分类:区分真实(bonafide) 人类语音与伪造(spoof) 合成或转换语音。
数据模式
| 字段 | 类型 | 描述 |
|---|---|---|
path |
string |
稳定的归档相对路径(如 LA_E_2834763.flac),数据集中唯一 |
audio |
Audio(16000) |
音频波形,16 kHz 单声道 |
label |
ClassLabel |
"bonafide"(索引0)或 "spoof"(索引1) |
notes |
string |
JSON格式,包含 utterance_id、speaker_id、subset |
notes 示例:
json
{"utterance_id": "LA_E_2834763", "speaker_id": "LA_0039", "subset": "eval"}
数据统计
| 统计项 | 数值 |
|---|---|
| 总样本数 | 71,237 |
| 真实样本(Bonafide) | 7,355 |
| 伪造样本(Spoof) | 63,882 |
配置与文件
- 配置名称:
default - 数据文件:
data/test-*.parquet(测试集)
快速使用
python from datasets import load_dataset
ds = load_dataset("SpeechAntiSpoofingBenchmarks/ASVspoof2019_LA", split="test") print(ds[0])
{path: LA_E_2834763.flac, audio: {array: ..., sampling_rate: 16000},
label: 1, notes: {"utterance_id": "LA_E_2834763", ...}}
来源信息
- 原始数据集: https://www.asvspoof.org/index2019.html
- 评估协议:
protocols/ASVspoof2019.LA.cm.eval.trl.txt
引用
原始论文: https://arxiv.org/abs/1911.01601
arXiv版本引用: bibtex @misc{wang2020asvspoof2019largescalepublic, title={ASVspoof 2019: A large-scale public database of synthesized, converted and replayed speech}, author={Xin Wang and Junichi Yamagishi and Massimiliano Todisco and Hector Delgado and Andreas Nautsch and Nicholas Evans and Md Sahidullah and Ville Vestman and Tomi Kinnunen and Kong Aik Lee and Lauri Juvela and Paavo Alku and Yu-Huai Peng and Hsin-Te Hwang and Yu Tsao and Hsin-Min Wang and Sebastien Le Maguer and Markus Becker and Fergus Henderson and Rob Clark and Yu Zhang and Quan Wang and Ye Jia and Kai Onuma and Koji Mushika and Takashi Kaneda and Yuan Jiang and Li-Juan Liu and Yi-Chiao Wu and Wen-Chin Huang and Tomoki Toda and Kou Tanaka and Hirokazu Kameoka and Ingmar Steiner and Driss Matrouf and Jean-Francois Bonastre and Avashna Govender and Srikanth Ronanki and Jing-Xuan Zhang and Zhen-Hua Ling}, year={2020}, eprint={1911.01601}, archivePrefix={arXiv}, primaryClass={eess.AS}, url={https://arxiv.org/abs/1911.01601}, }
同行评审出版物引用(Computer Speech & Language, 2020): bibtex @article{wang2020asvspoof, title={ASVspoof 2019: A large-scale public database of synthesized, converted and replayed speech}, author={Wang, Xin and Yamagishi, Junichi and Todisco, Massimiliano and Delgado, Hector and Nautsch, Andreas and Evans, Nicholas and Sahidullah, Md and Vestman, Ville and Kinnunen, Tomi and Lee, Kong Aik and others}, journal={Computer Speech & Language}, volume={64}, pages={101114}, year={2020}, publisher={Elsevier} }
维护者
联系: k.n.borodin@mtici.ru




