数据堂—88小时噪音环境口音普通话手机采集语音数据

中国普通话噪音环境口音语音数据_朗读（手机），基于给定的脚本朗读并模拟录制，由中国各地共203名发音人参与录制，覆盖各大方言区；涵盖多种噪音场景，如地铁、超市、餐厅等，与实际应用场景更契合，文本经过人工校对，准确率高，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, PIPLMandarin Chinese Accented Speech Dataset with Noise Environments (Read Speech, Recorded via Mobile Phones): This dataset is collected from 203 speakers across various regions of China covering all major dialect areas, who read from given scripts and complete recordings under simulated noise scenarios. It includes multiple common noise environments such as subways, supermarkets, restaurants and others, which better fits real-world application scenarios. The scripts have been manually proofread with high accuracy, providing abundant resources for speech recognition-related research and applications. Verified by multiple AI companies, this dataset helps models perform excellently when facing the diversity of real-world scenarios. We strictly abide by data protection laws and privacy regulations to safeguard users' privacy and legitimate rights and interests throughout the process of data collection, storage and utilization. All data complies with GDPR, CCPA and PIPL.

搜集汇总

背景与挑战

背景概述

该数据集包含88小时的口音普通话语音数据，通过手机在噪音环境下采集，涉及203名来自中国不同地区的说话者，覆盖地铁、超市等多样噪声场景。数据格式为16kHz、16bit的单声道WAV文件，适用于中文语音识别模型的测试任务，以及机器翻译和说话人验证等应用。