基于改进小波变换与BiLSTM-PPG信号中的应用  

Application of Improved Wavelet Transform and BiLSTM in PPG Signal Processing

作　　者：蔡俊[1] 李阿会 CAI Jun;LI Ahui(School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan Anhui 232001,China)

机构地区：[1]安徽理工大学电气与信息工程学院,安徽淮南232001

出　　处：《佳木斯大学学报(自然科学版)》2024年第12期17-20,共4页Journal of Jiamusi University:Natural Science Edition

基　　金：安徽省自然科学基金(2108085MF200);安徽高校自然科学重点项目(KJ2020A307)。

摘　　要：探讨了心电图(ECG)和光电容积(PPG)信号在现代医疗和健康监测中的关键作用,特别是在实时监控心脏健康和血流动态方面的应用。ECG和PPG信号在日常活动和运动条件下容易受到噪声干扰,因此开发有效的噪声去除技术显得尤为重要。提出了一种改进的小波变换去噪算法,并结合BiLSTM(双向长短期记忆网络)模型进行信号处理。实验结果表明,改进的小波变换方法在去噪效果和信号细节保留方面优于传统方法,而BiLSTM模型通过结合前向和后向信息,提高了心率估计的准确性。基于ISPC数据库的实验数据,BiLSTM-PPG模型的平均绝对误差(MAE)最低为2.09 BPM,优于TROIKA和Deep PPG模型。This paper discusses the critical role of Electrocardiogram(ECG)and Photoplethysmography(PPG)signals in modern medical and health monitoring,particularly for real-time cardiac health and blood flow dynamics.ECG and PPG signals are susceptible to noise interference during daily activities and exercise,highlighting the need for effective noise removal techniques.The paper proposes an improved wavelet transform denoising algorithm combined with a BiLSTM(Bidirectional Long Short-Term Memory)model for signal processing.Experimental results show that the improved wavelet transform method outperforms traditional methods in noise reduction and signal detail preservation.The BiLSTM model,leveraging both forward and backward information,enhances heart rate estimation accuracy.Based on ISPC database experimental data,the BiLSTM-PPG model achieves the lowest mean absolute error(MAE)of 2.09 BPM,surpassing TROIKA and Deep PPG models.

关 键 词：光电容积(PPG) 小波变换 双向长短期记忆网络 心率测量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]