University of Twente Student Theses


Using activity recognition to improve heart rate monitoring accuracy

Thoonen, Maarten (2022) Using activity recognition to improve heart rate monitoring accuracy.

[img] PDF
Abstract:Heart rate monitoring in continuous automatic vital signs monitoring often produces false alarms, because alarm thresholds are static and do not take context into account. Alarm management could be improved by taking into account what an expected heart rate would be. This work attemps to create a data-driven model of the heart rate reponse to various activities. An experiment was conducted which resulted in a dataset with 9 participants. In this experiment, participants wore IMUs (Inertial Measurement Unit, a type of movement sensor) and a dry electrode ECG recorder and were asked to perform Activities of Daily Living (ADL). The dataset was used to create a data-driven model that predicts the heart rate from the current activity. To do so, a Support Vector Machine (SVM) classifier with Radial Basis Function (RBF) kernel is used to perform Human Activity Recognition (HAR). A k-NN regressor is used to perform heart rate prediction based on the predicted activity, activity intensity and activity duration. Five-fold cross validation was used to evaluate system performance. The HAR classifier had a median accuracy of 87%, with a minimum of 82% and a maximum of 92%. The heart rate prediction algorithm had a median absolute error of median 3.82 BPM, minimum 2.94 BPM and maximum 4.79 BPM. This is deemed an acceptable result for a preliminary system. Future work includes building a more practical system to wear, creating a more general model that does not need to be individually trained and clinical validation.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:50 technical science in general, 53 electrotechnology
Programme:Embedded Systems MSc (60331)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page