AISHELL-2 中文语音数据库
收藏超神经2024-02-06 更新2024-05-15 收录
下载链接:
https://hyper.ai/cn/datasets/29347
下载链接
链接失效反馈官方服务:
资源简介:
希尔贝壳中文普通话语音数据库 AISHELL-2 的语音时长为 1000 小时,其中 718 小时来自 AISHELL-ASR0009-[ZH-CN],282 小时来自 AISHELL-ASR0010-[ZH-CN] 。录音文本涉及唤醒词、语音控制词、智能家居、无人驾驶、工业生产等 12 个领域。录制过程在安静室内环境中, 同时使用 3 种不同设备:高保真麦克风 (44.1kHz,16 bit);Android 系统手机 (16kHz,16bit);iOS 系统手机 (16kHz,16bit) 。 AISHELL-2 采用 iOS 系统手机录制的语音数据。 1991 名来自中国不同口音区域的发言人参与录制。经过专业语音校对人员转写标注,并通过严格质量检验,此数据库文本正确率在 96% 以上。(支持学术研究,未经允许禁止商用)
AISHELL-2, the Xier Beike Mandarin Chinese speech database, has a total audio duration of 1000 hours, with 718 hours sourced from AISHELL-ASR0009-[ZH-CN] and 282 hours from AISHELL-ASR0010-[ZH-CN]. The transcribed texts cover 12 domains including wake words, voice control commands, smart home, autonomous driving, industrial production, and others. Recordings were conducted in quiet indoor environments using three distinct devices: high-fidelity microphones (44.1kHz, 16 bit), Android smartphones (16kHz, 16bit), and iOS smartphones (16kHz, 16bit). Notably, AISHELL-2 exclusively utilizes speech data recorded with iOS smartphones. A total of 1991 speakers from various accent regions across China participated in the recording sessions. The transcriptions were proofread and annotated by professional speech revisers, and underwent strict quality inspections, achieving a text accuracy rate of over 96%. (For academic research purposes only; commercial use is prohibited without prior authorization.)
创建时间:
2024-02-06
搜集汇总
数据集介绍

背景与挑战
背景概述
AISHELL-2 是一个大规模中文普通话语音数据库,总时长达1000小时,覆盖唤醒词、智能家居等12个领域,由1991名不同口音区域的发言人使用iOS手机在安静环境中录制。该数据集经过专业转写和严格质量检验,文本正确率超过96%,适用于学术研究,但仅限于非商业用途。
以上内容由遇见数据集搜集并总结生成



