SPRSound

该数据集名为SPRSound，是首个开放获取的儿童呼吸音标注集合，覆盖了从1个月大到18岁的儿童，其中包含了被标记为喘鸣、爆裂声、粗湿啰音、喉鸣或正常等类别的音频片段。所有音频片段均为2秒钟长，被处理成单声道波形，并使用谷歌的HeAR模型嵌入为512维度的表示。该数据集包含了3,319个已标记的音频剪辑，其任务是对儿童呼吸音进行早期哮喘检测。

This dataset, named SPRSound, is the first open-access annotated collection of pediatric respiratory sounds, covering children aged from 1 month to 18 years old, and contains audio clips labeled as wheezes, crackles, coarse rales, stridor, or normal. All audio clips are 2 seconds in length, processed into monaural waveforms, and embedded into 512-dimensional feature representations using Google's HeAR model. This dataset comprises 3,319 labeled audio clips, and the associated task is early asthma detection based on pediatric respiratory sounds.

搜集汇总

背景与挑战

背景概述

SPRSound是一个开源的儿科呼吸音数据库，包含来自1个月至18岁儿童的呼吸音记录，记录格式为.wav文件，并附带.json格式的注释文件。数据集按年份（2022-2025）组织，每年对应不同的挑战任务，包括分类、压缩和事件检测等。