Antenna agreement helps recover breathing under motion interference
In plain words
This paper introduces a way to recover breathing from signals received by many wireless antennas. A learned system combines the antenna signals, then adjusts those combination weights for each recording without being given the true breathing pattern.
It checks whether estimates made from different groups of antennas agree, since breathing can show up across many antennas while a nearby moving object may affect only a few.
In simulated scenes with strong motion at breathing-like frequencies, the method matched the known breathing waveform better than a hand-crafted signal-combining method. The real recordings tested had no second moving body.
The research
Wireless respiration sensing can confuse breathing with a second moving body when both produce motion in the same frequency band. Can a system still recover the breathing waveform when frequency and signal power cannot distinguish the two?
SR-TTA adapts learned antenna-combination weights without labels, using agreement across antenna subsets to separate breathing from localized motion.
The authors ask whether breathing can be recovered when a moving body produces motion in the same frequency band. Their method adapts a learned antenna combiner for each recording without labels, seeking agreement among estimates from different antenna subsets.
In randomized simulated scenes with strong in-band interference, the adapted system had higher waveform correlation and lower breathing-rate error than a hand-crafted combiner.
On real single-person recordings, adaptation also reduced failures after the same one-time calibration, supporting an added benefit on those recordings.
The real recordings contained no second moving body; the interference added to them was simulated. The results therefore do not show how the method performs against a real moving interferer.
Analysis
Deep
- License
- CC BY 4.0