Dinner Party Corpus (DiPCo)
收藏arXiv2019-09-30 更新2024-06-21 收录
下载链接：
https://s3.amazonaws.com/dipco/DiPCo.tgz
下载链接
链接失效反馈官方服务：
资源简介：
Dinner Party Corpus (DiPCo) 是由亚马逊创建的一个语音数据集，模拟了家庭环境中晚餐聚会的场景。该数据集包含10个会话，每个会话记录了4名亚马逊员工志愿者围绕餐桌进行的自然英语对话。数据集通过一个单通道近讲麦克风和五个远场7麦克风阵列设备进行录音，总时长介于15至45分钟。数据集旨在促进噪音鲁棒性、远场语音识别、说话人识别和分离等领域的研究，为公共研究和基准测试提供数据支持。
Dinner Party Corpus (DiPCo) is a speech dataset created by Amazon that simulates dinner party scenarios in home environments. The dataset contains 10 sessions, each recording natural English conversations among 4 volunteer Amazon employees around a dining table. The data is recorded using a single close-talking microphone and five far-field 7-microphone array devices, with a total duration ranging from 15 to 45 minutes. This dataset aims to facilitate research in fields including noise robustness, far-field speech recognition, speaker recognition and speaker separation, and provides data support for public research and benchmarking.
提供机构：
亚马逊
创建时间：
2019-09-30
