resp-agent-dataset
收藏资源简介:
Resp-229K 是一个大规模呼吸音数据集,包含229,101个音频文件,总时长超过407小时。该数据集专为训练Resp-Agent系统(智能呼吸音分析与生成框架)而构建。数据集包含训练集(196,654个样本)、验证集(16,931个样本)和测试集(15,516个样本),平均音频时长为6.41秒。音频采样率主要为48000Hz(占85.67%)和44100Hz(占12.52%)。数据集整合了来自多个机构的呼吸音数据,包括UK COVID-19、COUGHVID、ICBHI等,各子数据集保留其原始许可协议。数据集包含AI生成的呼吸音描述文件(audio_descriptions.jsonl),记录音频特征、疾病标签等信息。该数据集适用于音频分类任务,特别是呼吸音识别、咳嗽检测等医疗音频分析场景,发布许可为CC BY-NC 4.0,仅限学术研究使用。
Resp-229K is a large-scale respiratory sound dataset consisting of 229,101 audio files with a total duration exceeding 407 hours. This dataset is specifically constructed for training the Resp-Agent system, an intelligent framework for respiratory sound analysis and generation. It is split into a training set (196,654 samples), a validation set (16,931 samples) and a test set (15,516 samples), with an average audio duration of 6.41 seconds. The main audio sampling rates are 48000 Hz (accounting for 85.67%) and 44100 Hz (accounting for 12.52%). The dataset integrates respiratory sound data from multiple institutions including UK COVID-19, COUGHVID, ICBHI and other sources, and each sub-dataset retains its original license agreement. The dataset also includes an AI-generated respiratory sound description file (audio_descriptions.jsonl), which records audio features, disease labels and other relevant information. This dataset is suitable for audio classification tasks, especially medical audio analysis scenarios such as respiratory sound recognition and cough detection. It is released under the CC BY-NC 4.0 license and is only permitted for academic research use.



