心跳声音数据集

该数据集最初是针对机器学习挑战来对心跳声音进行分类的。数据是从两个来源收集的：（A）通过iStethoscope Pro iPhone应用程序从公众那里获得的，以及（B）从使用数字听诊器DigiScope的医院的临床试验中获得的。这场比赛有两个挑战： 1.心音分割第一个挑战是产生一种方法，该方法可以在音频数据中定位S1（lub）和S2（dub）声音，从而对两个数据集中的Normal音频文件进行分割。 2.心音分类任务是产生一种方法，可以将真实的心音分类（也称为“心跳分类”），分为四个类别之一。 数据集分为两个来源：A和B： set_a.csv-通过iPhone应用程序从公众那里收集的心跳的标签和元数据 设置一个 timing.csv -含有金标准时序信息从集A的“正常”录音 set_b.csv-使用数字听诊器从医院的临床试验中收集的心跳的标签和元数据 音频文件 -长度不等，介于1秒和30秒之间。（已被裁剪以减少过多的噪音并提供声音的显着片段）。

This dataset was originally designed for a machine learning challenge focused on classifying heart sound signals. The data was collected from two sources: (A) obtained from the general public via the iStethoscope Pro iPhone application, and (B) obtained from hospital clinical trials using the DigiScope digital stethoscope. The challenge includes two tasks:
1. Heart Sound Segmentation: The first task is to develop a method that can locate the S1 (lub) and S2 (dub) sounds in the audio data, thereby segmenting the Normal audio files from both datasets.
2. Heart Sound Classification: The second task is to develop a method that can classify real heart sounds (also referred to as "heartbeat classification") into one of four categories.

The dataset is divided into two source groups: A and B:
- set_a.csv: Labels and metadata for heartbeats collected from the general public via the iStethoscope Pro iPhone application
- timing.csv: Contains gold-standard timing information for the "Normal" recordings from set A
- set_b.csv: Labels and metadata for heartbeats collected from hospital clinical trials using the DigiScope digital stethoscope

Audio files: Vary in length from 1 second to 30 seconds. They have been cropped to reduce excessive noise and retain salient segments of the heart sounds.

搜集汇总

背景与挑战

背景概述

心跳声音数据集是一个医学音频数据集，用于心跳声音的分类和分割任务，数据来源于公众和医院临床试验，包含音频文件和相关的标签及元数据。