Authors:
Kevin Kristofer Kosasih
1
;
Carl Daniel Karlsson
1
;
Thilini Savindya Karunarathna
1
and
Zilu Liang
1
;
2
Affiliations:
1
Ubiquitous and Personal Computing Lab, Kyoto University of Advanced Science (KUAS), Kyoto, Japan
;
2
Institute of Industrial Science, The University of Tokyo, Tokyo, Japan
Keyword(s):
Eye-Tracking, Pupillometric, Heart Rate, Regression, Machine Learning.
Abstract:
Heart rate is a key indicator of health, typically measured through skin-contact methods such as electrocardiograms (ECG) or photoplethysmograms (PPG). However, these methods may not be comfortable for everyone, prompting interest in non-contact alternatives. Eye tracking presents a promising solution, as the autonomic nervous system links the eyes to heart rate. This research develops heart rate prediction models based on pupillometric features. We conducted data collection experiments to build a dataset of multi-modal measurements of pupillometric data and heart rate from 10 subjects at high sampling rates. Several regression models, including linear regression, ridge regression, random forest regression, and XGBoost regression, were trained on the dataset. The random forest model achieved the best performance with a R2 of 0.457 and a root mean square error (RMSE) of 9 beats per minute, representing a 52.3% improvement over the state-of-the-art. Future work should focus on expandin
g the dataset, refining feature extraction and selection, and incorporating 3D pupillometric data to enhance model accuracy and applicability.
(More)