摘要
为了实现基于人体的连续血压检测,同时通过光电容积脉搏波信号获取血压参数,提出了一种新的VMD(Variational Modal Decomposition)-LSTM网络血压测量算法。变分模态分解定义为能够有效分解序列信号为不同频率段分信号的信号处理方法,适用于脉搏波信号的处理分析,提高了提取血压特征参数的准确率。在经过脉搏波信号的采集、预处理、特征点识别后得到的特征参数有利于进行血压值的预测。由于血压预测问题是一个非线性规划问题,难以实现血压的完美预测。利用LSTM(Long Short-Term Memory)网络模型使用1000条光电容积脉搏波提取出的血压特征参数得到合理的血压值。最后,实验结果表明,该算法具有较好的准确性,SBP(systolic blood pressure)与DBP(diastolic blood pressure)的精确度分别为2.89±5.38 mm Hg和3.95±7.09 mm Hg,血压预测指标误差较小。
In order to achieve the continuous blood pressure detection of human body and obtain the relevant blood pressure parameters through optical capacitance product pulse wave signal,a new vmd(Variational Modal Decomposition)-lstm network blood pressure measurement algorithm is proposed in this study.With variational modal decomposition as a signal processing method,the sequence signal is effectively decomposed into different frequency segments,which makes it suitable for the processing and analysis of pulse wave signal,thus improving the accuracy of extracting the parameters of blood pressure characteristics.After pulse wave signal acquisition,preprocessing and feature point recognition,the calculated characteristic parameters are used to predict blood pressure.Since blood pressure prediction is a nonlinear programming problem,it is difficult to make highly accurate prediction of blood pressure.By using LSTM(long short term memory)network model,the blood pressure characteristic parameters extracted by 1000 optical capacitance pulse waves are taken into account to determine a reasonable blood pressure value.Finally,the experimental results are obtained to show that the algorithm performs well in accuracy.The accuracy of SBP(systolic blood pressure)and DBP(diastolic blood pressure)reaches 2.89±5.38 mm Hg and 3.95±7.09 mm Hg,respectively,while the error of blood pressure prediction index is insignificant.
作者
庞宇
狄淳杰
PANG Yu;DI Chunjie(Chongqing University of Posts and telecommunications,Chongqing 400065)
出处
《生命科学仪器》
2022年第4期46-51,共6页
Life Science Instruments
基金
国家自然科学基金(U21A20447,61971079)
重庆市基础前沿项目(cstc2019jcyjmsxmX0666)
重庆市技术创新与应用发展专项(cstc2021jscx-gksbx0051)
重庆市创新群体项目(cstc2020jcyj-cxttX0002)
四川区域创新合作项目(2020YFQ0025)
重庆市教委科学技术研究项目(KJZD-k202000604)
关键词
连续血压检测
光电容积脉搏波
LSTM网络
变分模态分解
Continuous blood pressure detection
Optical capacitance pulse wave
LSTM network
Variational modal decomposition
作者简介
狄淳杰(1997-),男,重庆南岸人,硕士研究生。研究方向:智慧医疗与嵌入式系统;通信作者:庞宇(1978-),2010年博士毕业于加拿大McGill大学,现为重庆邮电大学教授,主要从事数字医疗器械研究很健康物联网体系结构研究。
