数据堂—200人中文手机采集语音数据

200人中文手机采集语音数据是由200名中国本土发音人，覆盖各大方言区；录音涵盖安静和噪音环境，更贴合语音识别实际应用场景；录音内容均为常用口语化句子，朗读自然流利；人工转写，准确率高。200人中文手机采集语音数据可应用于语音识别、机器翻译

The 200-person Mandarin speech dataset collected via mobile phones is recorded by 200 native Chinese speakers covering major dialect regions of China. The recordings are conducted in both quiet and noisy environments, which better aligns with real-world application scenarios of speech recognition. All recorded content consists of commonly used colloquial sentences, read naturally and fluently. The dataset features high-accuracy manually transcribed transcripts. This 200-person mobile-collected Mandarin speech dataset can be applied to speech recognition and machine translation.

搜集汇总

背景与挑战

背景概述

该数据集包含200名中国母语者通过手机采集的语音数据，覆盖主要方言区，录音环境涵盖安静和嘈杂场景，内容为常用口语语句，适用于中文语音识别模型的测试任务。音频格式为16kHz、16bit的单通道WAV文件，由数据堂版权所有，为商业数据。