EACeleb

EACeleb是一个专注于东亚语言的名人语音数据集，由乔治亚理工学院创建，旨在为语音识别系统提供训练数据。该数据集包含1,641位名人的语音记录，总计76,522条语音片段，总时长约180小时。数据集通过自动化管道从YouTube视频中提取，使用面部追踪技术以提高数据采集效率。EACeleb特别适用于东亚语言的语音识别和验证任务，旨在解决现有数据集中东亚语音数据不足的问题。

EACeleb is a celebrity speech dataset dedicated to East Asian languages, developed by the Georgia Institute of Technology, with the objective of providing training data for speech recognition systems. This dataset encompasses speech recordings from 1,641 celebrities, totaling 76,522 speech segments with an approximate cumulative duration of 180 hours. The dataset is extracted from YouTube videos through an automated pipeline, and facial tracking technology is employed to enhance the efficiency of data acquisition. EACeleb is particularly well-suited for speech recognition and verification tasks targeting East Asian languages, and it aims to address the scarcity of East Asian speech data in existing datasets.