Hi all,
We recently ran a benchmark test using the 60GHz mmWave Breathing and Heartbeat Detection Sensor Kit with XIAO ESP32C6 (MR60BHA2) (Getting started with MR60BHA2 | Seeed Studio Wiki), comparing its heartbeat detection to the Polar H10 chest strap monitor, which is known for its precision.
The test involved minute-by-minute comparisons between the two devices, and unfortunately, we found that the correlation was quite low—around 0.33, which isn’t reliable enough for hospital-grade remote patient monitoring.
We’re wondering:
- Is there any way to improve the precision of the MR60BHA2 for heartbeat detection?
- Are there any known tuning or calibration methods that could enhance accuracy?
Additionally, we noticed an issue with the mmWave.getPeopleCountingTargetInfo(PeopleCounting& target_info) function, it consistently returns 0 or 1, but never 2 or more, even when multiple people are present.
Has anyone encountered this issue or found a workaround?
Looking forward to insights from the community!
Thanks,
Sergio
Thank you for conducting the comparative test between the MR60BHA2 and the Polar H10 chest strap monitor! Your detailed feedback is extremely valuable, and we will carefully document this information for continuous product improvement.
Regarding the heartbeat detection precision:
We understand your expectations for hospital-grade remote patient monitoring. However, we need to clarify that with current technological limitations, mmWave radar sensors (including the MR60BHA2) cannot be used for actual medical applications and cannot obtain relevant medical device certifications. The current breathing and heart rate detection accuracy is roughly comparable to consumer-grade wearable devices like smartwatches, primarily suitable for non-medical health reference scenarios.
The radar algorithms are currently not open-source, and our engineering team is continuously working to optimize algorithm performance. While we cannot provide direct tuning methods, ensuring that the subject remains still and is positioned in the optimal detection zone (typically 0.5-1.5 meters directly in front of the sensor) can improve measurement accuracy.
Regarding the people counting issue:
The multiple people detection feature is indeed still in the algorithm optimization stage, and the measurement results should be considered as reference only. When multiple people are present with front-to-back occlusion or when some individuals remain stationary, detection accuracy can be affected. We’ve found that results tend to be more reliable when all people in the measurement area are clearly moving.
We will directly forward your feedback to our product development team as important reference for future firmware updates and algorithm optimizations. Thank you for your support and valuable feedback on our product!