Deep Learning based Pulse Rate Prediction in Noisy PPG Signals with Practical Application to Wearables
作者
Husnu Suleyman Hanayli,Alisher Kholmatov
出处
期刊: [Institute of Electrical and Electronics Engineers] 日期:2022-05-15卷期号:: 1-4被引量:2
标识
DOI:10.1109/siu55565.2022.9864877
摘要
When the quality of photoplethysmography (PPG) signals degrades, the measurement of heart rate (HR) becomes highly challenging, if not impossible. To alleviate the problem, we utilize subject’s kinematic information and devise a neural network-based deep learning model to predict heart rate. A labeled data set of varying signal quality totaling to 1750 hours and comprising of comprehensive set of physical activities such as walking, running, biking, rowing, elliptical machine was used to train and validate the model. We demonstrate that the proposed method enhances overall accuracy of the state-of-the-art heart rate measurement (HRM) algorithm by 1% and reducing both mean absolute (MAE) and root mean square (RMSE) errors, respectively. Besides, proposed method is integrated onto a memory and computation constraint micro controller (MAX32664) that demonstrates its usability for real life applications. To the best of our knowledge, our HR prediction approach is unique both in a way it approaches the problem and has the practical implementation.