MobiVital is an 24-hour large-scale dataset collected from 12 subjects on vital signal sensing using IR-UWB radar. We also provide ground truth waveforms collected from wearable sensors, which are time-synchronized with the radar time series. This study is approved by UCLA IRB under IRB#23-000754. The full anonymized dataset will be released on Zenodo following the protocol specified in the IRB documents.
We select Ultra-wideband (UWB) radar as the RF media for its low energy consumption, fine-grained resolution, and availability in mobile devices. Volunteers were asked to sit in a private section in our lab, relax, and breathe normally. A laptop may be provided if the volunteers prefer to watch their recreational videos. We used a tripod-mounted or handheld UWB radar facing the volunteer to sense the environment and collect a UWB radar response signal containing information about his/her chest movement, which can be used to extract vital signals like respiration and heartbeat. We also collect accelerometer and gyroscope traces to estimate the handheld platform'smovement.
As ground truth, we collected respiration and heartbeat waveforms from wearable chest bands and finger clips. All the data is resampled and time-synchronized to form a dataset for RF-based human vital signal measurements using mobile devices.
We do not seek to make medical or healthcare inferences from the data. The final dataset will not contain any personally identifiable information and will be made publicly available to encourage reproducible research.
To be added after paper publication.