AISHELL-5
收藏资源简介:
AISHELL-5是一个开源的汽车多通道多说话人语音数据集,由西北工业大学计算机科学与技术学院、北京爱数智能科技有限公司和理想汽车有限公司联合创建。该数据集包括超过100小时的在真实驾驶场景中记录的多通道语音数据,以及40小时的现实世界环境噪声记录。数据集旨在支持自动语音识别和说话人分割研究,特别是针对复杂的驾驶环境。数据集共包含超过100小时的语音数据,分为94小时的训练数据、3.3小时的验证数据和两个测试集,每个测试集分别为3.3小时和3.58小时。数据集还包括一个40小时的大型噪声数据集,用于促进车内语音数据模拟研究。
AISHELL-5 is an open-source automotive multi-channel multi-speaker speech dataset jointly created by the School of Computer Science and Technology of Northwestern Polytechnical University, Beijing AISHU Intelligent Technology Co., Ltd., and Li Auto Inc. This dataset contains over 100 hours of multi-channel speech data recorded in real driving scenarios, as well as 40 hours of real-world environmental noise recordings. It aims to support research on automatic speech recognition (ASR) and speaker diarization, particularly for complex driving environments. The dataset consists of over 100 hours of speech data in total, divided into 94 hours of training data, 3.3 hours of validation data, and two test sets with durations of 3.3 hours and 3.58 hours respectively. The dataset also includes a 40-hour large-scale noise dataset to facilitate research on in-vehicle speech data simulation.
数据集概述
数据方案
- 语音识别
- 语音合成
- 声纹识别
- 自然语言处理
- 多语种发音词典
数据产品
- ChatGPT对话式数据
- 场景数据
- 基础数据
开源数据
- AISHELL-1
- AISHELL-2
- AISHELL-3
- WakeUp-1
- DMASH
- AISHELL-4
- AISHELL-5
- AISHELL-6A
- AISHELL-6B
- AISHELL-RealMAN
- AISHELL-7A
评测数据
- 2018-EVAL
- 2019-EVAL
- 2021-EVAL
- 2022-EVAL
- 2023-EVAL
数据竞赛
- 国际赛事
数据产品详情
- 智能家居语音数据-中文 (AISHELL-ASR0010)
- 智能车载语音数据-电动汽车 (AISHELL-ASR0031)
- 会议数据 (AISHELL-ASR0055)




