mSense: Mobile & Wearable Sensing at scale

The main objective of this effort is to provide a secure and lightweight framework for real-time analysis of sensory data from wearables. Our goal is to continuously monitor warfighter health and readiness at the edge in real-time while consuming less than 5% of the wearable battery over 10 days. Our system enables the provisioning and instantiation of different edge-deployed algorithms to address the symptom domains while it provides highly accurate sensing, context awareness, security, and privacy at the edge delivering seamless service and building a reliable monitoring system. Our system’s strengths are its ability to work seamlessly with different algorithms and its compatibility and operability on different edge devices. It can facilitate the design and implementation of algorithms related to real-time processing of diverse sensor data and their fusion towards specific goals such as human activity recognition or using physiological data as a wellness indicator. We envision our platform to provide an affordable and ubiquitous access to Digital Biomarker-based wellness detection and assessment, early treatments and therapies, enhanced telehealth clinical interventions, and better health outcomes. Additionally, this platform will not only enable innovation through secure and privacy-preserving edge deployment of biosensing algorithms fused across multiple sensors, but it will also present curated Digital Biomarker data sets available to researchers to support ongoing research in mental and physical health. We aim to develop a platform, called mSense, for collecting and curating digital biomarkers and sensor data. Seamless to the end user, the sensors will be measured passively and actively for predictive analytics deriving personalized levels of physical and cognitive wellness, human performance, and behavioral health across a wide range of activity contexts and operational environments. mSense will gather information from sensors of wearables and will pre-process them to remove any data discrepancies and to extract device and user context. Finally, we plan to leverage advances in low-power sensing at the chipset level present in modern mobile and wearable devices including but not limited to “always-on sensing” and we offer modular APIs to access sensor data and edge machine learning models/algorithms that can fit in the resource constrained environments provided by commercial wearables. By running the models at the edge, the platform can reduce latency, minimize the data transmission, and real-time feedback to the user. Also, by leveraging ML techniques at the edge and on-device training, our platform can provide personalized insights to the user based on their unique physiological and environmental data.