VoxAging
收藏资源简介:
VoxAging 是一个大规模纵向数据集,包含来自293位讲者(226位英语讲者和67位普通话讲者)的录音,跨越了17年(约900周),每周进行记录。该数据集包含了2,629,100个片段,总计7,522小时的视听数据。数据来源于YouTube和Bilibili,通过手动筛选确保视频质量和适当的时间跨度。VoxAging 数据集的独特之处在于其连续的每周间隔,以及长跨度时间,这在以往的讲者老化数据集中是罕见的。数据集的收集过程采用了动态模板,以适应讲者面部和声音特征的衰老变化。VoxAging 数据集旨在解决讲者验证系统中讲者老化带来的挑战,并研究讲者老化现象及其对高级讲者验证系统的影响。
VoxAging is a large-scale longitudinal dataset containing audio recordings from 293 speakers (226 English speakers and 67 Mandarin speakers), spanning 17 years (approximately 900 weeks) with weekly recordings. This dataset comprises 2,629,100 segments, totaling 7,522 hours of audio-visual data. The data was sourced from YouTube and Bilibili, with manual filtering conducted to ensure video quality and appropriate temporal spans. What sets the VoxAging dataset apart is its consistent weekly intervals and long temporal duration, which are rare among existing speaker aging datasets. The dataset collection process employs dynamic templates to accommodate the aging variations of speakers' facial and vocal features. The VoxAging dataset aims to address the challenges posed by speaker aging in speaker verification systems, and investigate the phenomenon of speaker aging and its impacts on advanced speaker verification systems.
VoxAging 数据集概述
基本信息
- 数据集名称:VoxAging
- 关联会议:Interspeech 2025
- 状态:即将发布
说明
- 该数据集为Interspeech 2025会议相关论文及配套数据,目前尚未正式发布。
- 页面提示"coming soon",具体发布时间和细节需等待进一步更新。




