AS-70
收藏资源简介:
AS-70数据集是由西北工业大学等机构创建的首个公开的普通话口吃语音数据集,旨在支持自动语音识别(ASR)和口吃事件检测(SED)研究。该数据集包含70名成年口吃者的语音数据,总时长达48.8小时,涵盖对话和语音命令阅读两种任务,具有详尽的手动转录。创建过程中,参与者通过在线平台进行录音,并采用自愿口吃技巧以增加数据的真实性。AS-70数据集的应用领域主要集中在提升ASR模型对口吃语音的识别能力,以及开发更有效的口吃检测系统,以促进语音技术在特殊人群中的应用和包容性。
The AS-70 dataset is the first publicly available Mandarin stuttered speech dataset developed by Northwestern Polytechnical University and other institutions, designed to support research in Automatic Speech Recognition (ASR) and Stuttered Event Detection (SED). This dataset contains speech data from 70 adult people who stutter, with a total duration of 48.8 hours, covering two tasks: conversational speech and speech command reading, and is accompanied by comprehensive manual transcriptions. During the data collection process, participants recorded their speech via an online platform and adopted voluntary stuttering techniques to enhance the authenticity of the dataset. The primary application scenarios of the AS-70 dataset focus on improving the recognition performance of ASR models for stuttered speech, as well as developing more effective stuttered event detection systems, so as to promote the application and inclusivity of speech technologies for special populations.

- 1AS-70: A Mandarin stuttered speech dataset for automatic speech recognition and stuttering event detection西北工业大学 · 2024年



