Wearable sensors have become ubiquitous thanks to a variety of health tracking features. The resulting continuous and longitudinal measurements from everyday life generate large volumes of data. However, making sense of these observations for scientific and actionable insights is non-trivial. Inspired by the empirical success of generative modeling, where large neural networks learn powerful representations from vast amounts of text, image, video, or audio data, we investigate the scaling properties of wearable sensor foundation models across compute, data, and model size. Using a dataset of up to 40 million hours of in-situ heart rate, heart rate variability, accelerometer, electrodermal activity, skin temperature, and altimeter per-minute data from over 165,000 people, we create LSM, a multimodal foundation model built on the largest wearable-signals dataset with the most extensive range of sensor modalities to date. Our results establish the scaling laws of LSM for tasks such as imputation, interpolation and extrapolation across both time and sensor modalities. Moreover, we highlight how LSM enables sample-efficient downstream learning for tasks including exercise and activity recognition.

Proximity beacons provide simple, low-cost location data. However, beacon deployments remain rare. In this paper we introduce Bellrock, a framework that repurposes static personal devices (phones, laptops, etc.) as proximity beacons without revealing the location of the device owners, and provides conventional beacons with access control. This is done by using mutable pseudo-anonymous identifiers that can be unmasked by a cloud service.We develop Bellrock as a general framework, describing the repurposing scheme and the anonymisation techniques, before applying it to Bluetooth Low Energy beacons. We implement and demonstrate the scalability of the de-anonymisation server, which uses a series of heuristics. We implement a Bellrock client on Android and demonstrate negligible impact on battery lifetime. We evaluate Bellrock using extensive real-world office worker movements. We find that Bellrock was able to provide proximity locations for 8,542 of the 21,796 failed locations that would have occurred without it. We further find that office workers were in range of one or more of their co-workers over 90% of the time, indicating Bellrock can provide relative proximity information even in the absence of a conventional beacon deployment. Overall, we find that Bellrock is both feasible and practical, providing a beacon deployment where there was none, or supplementing existing deployments.