BCG dataset

BCG dataset: The bed based ballistocardiography dataset collected and shared by Carlson et al. [3]. Both ABP and PPG signals were recorded from 40 subjects, with one record per subject. The original sampling rate of the signals is 1000 Hz. We resampled them at 125 Hz to be consistent with the previous data sets. Then, we multiplied the ABP signals by a factor of 100 since they are scaled at 100 mmHg/volt in the downloaded dataset. After preprocessing, there are 2559 segments left, which is around 4 hours of recording. To sum up, the BCG dataset is a small to medium-sized set with less data variation given its low number of subjects; its remarkably high ratio of segments per subject (63.97); and a narrower BP distribution. We applied a 5-fold CV to partition the data. [3] Carlson, C. et al. Bed-based ballistocardiography dataset (2020). IEEE Dataport https://dx.doi.org/10.21227/77hc-py84.

BCG数据集：由Carlson等人[3]采集并公开的心冲击图（ballistocardiography, BCG）数据集。该数据集包含40名受试者的动脉血压（ABP）与光电容积脉搏波（PPG）信号，每名受试者对应一条原始记录。信号的原始采样率为1000 Hz，我们将其重采样至125 Hz，以与既往数据集保持一致。由于下载的原始数据集中动脉血压信号的缩放比例为100 mmHg/伏特，我们将其数值乘以100进行校准。经预处理后，最终剩余2559个数据片段，总记录时长约为4小时。

综上，该BCG数据集属于中小规模数据集：因受试者数量有限，数据变异度较低；单受试者对应数据片段数极高（63.97）；且血压（BP）分布区间相对狭窄。我们采用5折交叉验证对数据集进行划分。

[3] Carlson, C. 等. 基于床铺的心冲击图数据集（2020）. IEEE Dataport https://dx.doi.org/10.21227/77hc-py84.

搜集汇总

背景与挑战

背景概述

BCG数据集是一个基于床的球状心动图数据集，包含40名受试者的动脉血压（ABP）和光电容积脉搏波（PPG）信号，原始采样率为1000 Hz，经重采样至125 Hz和信号缩放预处理后，形成2559个数据段（约4小时记录）。该数据集具有中小规模、数据变异度低、每个受试者数据段比例高（63.97）以及血压分布较窄的特点，主要用于机器学习非侵入性血压估计的基准研究，并采用5折交叉验证进行数据划分。