数据堂—245小时车载环境普通话手机采集语音数据

695名中国本土发音人参与录制，有效数据时长为245小时，覆盖全国多个地区。录制均在车内环境下进行，覆盖不同道路类型、不同车型、不同车窗开闭情况、是否开启音乐等多种情景。录音内容丰富，涵盖客户咨询、短信、新闻等多个领域

A total of 695 native Chinese speakers from multiple regions across China participated in the recording project. The dataset has a total valid audio duration of 245 hours. All recordings were conducted in in-vehicle environments, covering diverse scenarios including different road types, vehicle models, window opening/closing states, and scenarios with or without background music. The recorded content is rich and spans multiple domains such as customer inquiries, text messages, and news programs.

搜集汇总

背景与挑战

背景概述

该数据集包含在车载环境下通过手机采集的245小时普通话语音数据，涵盖中国多区域场景，由695名说话者录制，用于中文语音识别模型的测试任务。数据格式为16kHz单声道WAV文件，内容覆盖客服咨询、短信及新闻等30余个领域。