Hi-MIA
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/kaldi-asr/kaldi/tree/master/egs/hi_mia/v1
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为Hi-MIA，包含了来自340位发言人在普通话和英语环境下，在干净和智能家居条件下录制的1561小时音频。该数据集分为两个子集：AISHELL-wakeup和AISHELL-2019B-eval。此外，数据集还包括了近距离高保真麦克风和各种分布式麦克风阵列的录音，其录音条件与2020年远场说话人验证挑战赛的评估数据相似。规模上，该数据集涵盖了来自340位发言人的1561小时音频。其任务目标是说话人验证。
This dataset, named Hi-MIA, contains 1,561 hours of audio recorded by 340 speakers in both Mandarin and English environments, under both clean and smart home conditions. It is split into two subsets: AISHELL-wakeup and AISHELL-2019B-eval. Additionally, the dataset includes recordings collected using close-range high-fidelity microphones and various distributed microphone arrays, with recording conditions matching those of the evaluation data from the 2020 Far-Field Speaker Verification Challenge. In terms of scale, the dataset encompasses 1,561 hours of audio from 340 speakers. The task objective of this dataset is speaker verification.
提供机构：
Open-sourced
