Photoplethysmography (PPG) Signal Heart Rate Monitoring During Exercise and Reduces Motion Artifacts

Abstract

Photoplethysmography (PPG) is a non-invasive technique for monitoring cardiovascular parameters such as heart rate during various activities, including exercise. However, the accuracy of PPG-based heart rate monitoring can be compromised by motion artifacts caused by body movements. This study explores the effectiveness of three distinct algorithms – Random Forest, Decision Tree, and a novel Lion Optimization Algorithm-enhanced Long Short-Term Memory (LOA-LSTM) – in improving PPG-based heart rate monitoring accuracy during exercise while mitigating motion artifacts.The Random Forest algorithm harnesses ensemble learning to aggregate Decision Trees, providing robustness against noise and improving heart rate predictions. Decision Trees offer transparent decision-making based on PPG features, aiding in rapid classification of heart rate trends. The LOA-LSTM algorithm uniquely combines the Lion Optimization Algorithm's ability to adaptively explore and exploit with the temporal sequence learning capacity of LSTM. This integration aims to achieve high accuracy by dynamically optimizing LSTM parameters, effectively reducing motion artifacts and improving exercise-related heart rate predictions.In this comparative study, these algorithms were evaluated using a diverse dataset collected during exercise sessions. Experimental results demonstrate that while all three algorithms enhance heart rate monitoring accuracy and reduce motion artifacts, the LOA-LSTM algorithm outperforms the others, consistently achieving the highest accuracy rates about 99%. The proposed approach holds significant promise for improving real-time heart rate monitoring accuracy during exercise, contributing to more reliable fitness tracking and healthcare applications.