


Smartwatches and other wearable devices can monitor our health and help detect irregular heart rhythms. But continuously analysing all that health data uses energy, and batteries only last so long.
Researchers from the University of Twente and Vrije Universiteit Amsterdam have found a way to make these calculations much more energy efficient. In their experiments, their approach improved power efficiency by 64.9%. The research received an Outstanding Paper award at the Euromicro Conference on Digital System Design (DSD) in September 2026.
A smartwatch that continuously monitors your heart must make many calculations. The more energy these calculations use, the sooner the battery runs out. So, the researchers asked a simple question: do all these calculations really need to be completely exact?
They developed a system that analyses the electrical activity of the heart to detect irregular heart rhythms, also known as arrhythmias. It looks at individual heartbeats and uses an AI system trained to recognise different types of arrhythmias.
The researchers then made some of the calculations less precise on purpose. This means the system works with less detailed numbers and needs less energy to perform the calculations. In their experiments, for example, they reduced the precision from 16 bits to 8 bits and replaced some exact calculations with approximate ones. They trained the AI system to deal with the small errors these approximations can cause.
“If you train well, you can compensate for imprecision (to some extent) and gain significant energy efficiency,” says Ghayoor Gillani, assistant professor in the Computer Architecture for Embedded Systems group at the University of Twente.
The experiments show that the system can use simpler calculations while still recognising irregular heart rhythms with acceptable accuracy. This suggests that health monitoring could require considerably less energy.
The technology is still at an early stage. If it can eventually be used in wearable devices, it could help them monitor people's health for longer before their batteries need to be recharged. This could be especially useful for devices that continuously monitor someone's health.
The research started in 2024 with Ghayoor Gillani and Arlene John from the University of Twente. Together with PhD candidate Yawar Rasheed, they investigated how wearable devices could use less energy to detect irregular heart rhythms.
Embedded Systems master's student Floriaan Bulten took on the challenge for his thesis in 2025. He developed the system that became the basis for the research. After Bulten completed his thesis, Vincenzo Stoico from Vrije Universiteit Amsterdam joined the team. By bringing together knowledge of health monitoring, computer hardware and software, the researchers developed the results into a research paper.
Their paper, Toward Energy-Efficient and Low-Power Arrhythmia Detection for Wearable Devices, was accepted for the Euromicro Conference on Digital System Design and received an Outstanding Paper award. The researchers have already made a preprint, an early version of the scientific paper, publicly available. The official conference proceedings are still to come.
For now, the technology is a proof of concept: an early version that shows the idea can work. The VU-UT alliance has provided seed funding to take the research further in a new project called DAISY-HP.
The researchers will use the project to build a working version that they can test in practice. In the longer term, they want to develop reliable ways to monitor people's health with wearable devices while protecting their privacy. They also aim to collaborate with industrial partners to develop the technology further.