摘要
基于光电容积脉搏波(photoplethysmography,PPG)的心率(heart rate,HR)估计已被广泛应用于可穿戴设备。然而由于用户的身体活动引起的运动伪影,难以从受污染的PPG中获得准确的心率估计。为应对这一难题,提出了一种称为LU_PPG的心率估计方法,该方法基于回归的思想,首先利用2015 IEEE spc数据集训练和测试LSTM_Unet神经网络模型,然后经网络输出类PPG信号(含心率信息),最后基于频谱分析来估计最终的心率。实验结果表明,LU_PPG方法在该数据集上得到的心率估计平均误差为2.27次/min,为心率检测提供了新思路和途径。
Heart rate(HR)estimation techniques based on photoplethysmography(PPG)signals are widely applied in wearable devices.However,it is difficult to obtain accurate HR estimations from contaminated PPG signals due to motion artifacts caused by the users’physical activities.In order to cope with this problem,a heart rate estimation method called LU_PPG was proposed based on regression ideas.Firstly,the LSTM_Unet neural network model was trained and tested with the 2015 IEEE spc dataset.Then the PPG-like signal(with heart rate information)was output via the above model.Finally,the final heart rate was estimated through spectral analysis.The experimental results show that the LU_PPG method obtained a mean error of 2.27 time/min for heart rate estimation on this dataset,which provides new approach to heart rate detection.
作者
黄伟安
黎峰
张雨
李济涵
高军峰
HUANG Wei-an;LI Feng;ZHANG Yu;LI Ji-han;GAO Jun-feng(College of Biomedical Engineering,South-Central University for Nationalities,Wuhan 430074,China)
出处
《科学技术与工程》
北大核心
2023年第5期1875-1881,共7页
Science Technology and Engineering
基金
国家自然科学基金(61773408)
中央高校基本科研业务费专项资金(CZZ19004,CYZ20039)。
