基于潜在空间特征的无袖带血压估计方法

A cuff-less blood pressure estimation method based on latent space features

针对目前基于光电容积脉搏波(photoplethysmography,PPG)的血压(blood pressure,BP)估计模型的不足,文章提出一种基于潜在空间特征的BP估计方法。该方法充分挖掘PPG中与BP相关的特征,利用梯度增强回归树(gradient boosting regression tree,GBRT)提取单周期PPG信号及其一阶和二阶导数的高阶交叉特征,并利用卷积神经网络(convolutional neural networks,CNN)提取PPG信号的时频图中的深层特征;随后将提取的特征输入支持向量回归器(support vector regression,SVR)构建BP估计模型。在UCI-BP数据库的12000个样本上进行了模型评估,实验结果表明所提出的方法优于最新的BP估计方法。

Aiming at the shortcomings of the current blood pressure(BP)estimation model based on photoplethysmography(PPG),a BP estimation method based on latent space features is proposed.This method fully mines the BP-related features in the PPG signal,uses the gradient boosting regression tree(GBRT)to extract the high-order cross features of the single-cycle PPG signal and its first and second derivatives,and convolutional neural networks(CNN)are used to extract deep features in the time-frequency diagram of PPG signal.Subsequently,the extracted features are input to a support vector regression(SVR)to build a BP estimation model.The model is evaluated on 12000 samples of UCI-BP database,and the experimental results show that the proposed method outperforms other state-of-the-art methods in BP estimation.