1,420小时普通话自然语音手机采集数据

中国普通话语音数据_自然（手机），由700位普通话发音人参与录制，其中男性占比35%，女性占比65%。无预制文本，录音人以自然方式进行手机通话，同时录制通话的内容。准确性高，为语音识别相关研究及应用提供了丰富的资源，经多家AI公司验证：有助于模型面对真实世界的多样性时能够表现出色。我们严格遵循数据保护法规和隐私规定，确保数据采集、存储和使用的过程中维护用户的隐私和合法权益，所有数据均遵循GDPR, CCPA, PIPL。

Natural (Mobile) Mandarin Chinese Speech Dataset. A total of 700 Mandarin speakers participated in the recording, with 35% being male and 65% female. No pre-written texts were used; participants made mobile phone calls naturally, and the call contents were recorded simultaneously. The dataset boasts high accuracy and provides rich resources for speech recognition-related research and applications. Verified by multiple AI companies, it helps models perform excellently when facing real-world diversities. We strictly comply with data protection regulations and privacy rules, ensuring that user privacy and legitimate rights and interests are fully safeguarded throughout the entire process of data collection, storage and usage. All collected data adheres to GDPR, CCPA and PIPL.

搜集汇总

背景与挑战

背景概述

该数据集包含1,420小时的普通话自然语音，通过安卓手机在相对安静环境中采集，由700名母语者（35%男性、65%女性）进行无脚本对话录制。数据格式为16kHz、16位单声道WAV，准确率达95%，并附有转录文本，适用于语音识别研究，且遵循隐私法规，可供商业使用。