Automotive In-Cabin Vital Sign Monitoring : A Review of State-of-the-Art Methods and an Experimental Study on Dual-Sensor Fusion
Anttola, Ella (2026)
2026
Bioteknologian ja biolääketieteen tekniikan maisteriohjelma - Master's Programme in Biotechnology and Biomedical Engineering
Lääketieteen ja terveysteknologian tiedekunta - Faculty of Medicine and Health Technology
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2026-04-21
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202604204086
https://urn.fi/URN:NBN:fi:tuni-202604204086
Tiivistelmä
Driver and passenger monitoring systems have become increasingly common in modern vehicles. Monitoring driver’s drowsiness and attention is already mandatory in new vehicles, and safety rating protocols drive the implementation of occupancy detection systems. Remote vital sign monitoring has been proposed as a means to detect of driver drowsiness, sudden illness of the driver, or the presence of a child left in the vehicle.
This study aims to provide an overview of camera- and radar-based vital sign monitoring systems and the challenges present in the in-cabin environment. The recent research focuses mainly on improving robustness against motion and illumination interference. Illumination-related interference affects camera-based remote photoplethysmography (rPPG) measurements. Use of near-infrared imaging can reduce the interference, but it comes at the cost of lower signal-to-noise ratio of the rPPG signal compared to the signal captured using visible wavelengths. Motion interference is a challenge in both rPPG- and radar-based measurements, and especially large head turns and body rotations are difficult to compensate.
The proposed solutions are mainly algorithmic, as hardware solutions alone are insufficient. In addition to conventional signal processing, deep learning methods have been shown to improve vital sign detection accuracy. A third challenge arises from the large inter-subject diversity and the diversity of measurement environments, which complicates the design and evaluation of such systems.
In addition, benefits of sensor fusion in heart rate measurement are studied. An RGB camera is used to measure rPPG, and the position data of the target captured by a radar is used to compensate motion-related artefacts. The results showed large inter-participant variation and challenges in detecting higher heart rate. Sensor-fusion method did not improve outperform rPPG alone. This was likely due to unsuitable processing approach of the position data and limitations in the measurement setup, which may have affected the performance.
This study aims to provide an overview of camera- and radar-based vital sign monitoring systems and the challenges present in the in-cabin environment. The recent research focuses mainly on improving robustness against motion and illumination interference. Illumination-related interference affects camera-based remote photoplethysmography (rPPG) measurements. Use of near-infrared imaging can reduce the interference, but it comes at the cost of lower signal-to-noise ratio of the rPPG signal compared to the signal captured using visible wavelengths. Motion interference is a challenge in both rPPG- and radar-based measurements, and especially large head turns and body rotations are difficult to compensate.
The proposed solutions are mainly algorithmic, as hardware solutions alone are insufficient. In addition to conventional signal processing, deep learning methods have been shown to improve vital sign detection accuracy. A third challenge arises from the large inter-subject diversity and the diversity of measurement environments, which complicates the design and evaluation of such systems.
In addition, benefits of sensor fusion in heart rate measurement are studied. An RGB camera is used to measure rPPG, and the position data of the target captured by a radar is used to compensate motion-related artefacts. The results showed large inter-participant variation and challenges in detecting higher heart rate. Sensor-fusion method did not improve outperform rPPG alone. This was likely due to unsuitable processing approach of the position data and limitations in the measurement setup, which may have affected the performance.