I-Corps: Translation Potential of Privacy-Preserving Internet of Things (IoT) for Clinical and Home Health Monitoring

Award ID: 2638897

This I-Corps project is based on the development of a privacy-preserving sensing platform that monitors movement, activity, and health without recording identifiable images or understandable speech. Falls are the most common preventable adverse event in American hospitals. Current technologies that detect falls use cameras and microphones, yet those devices are prohibited or unwelcome in exactly the places where risk is highest, including bathrooms, behavioral health units, pediatric areas, and private homes. This technology strips identifying detail inside the device at the moment of capture, preserving the information needed for automated detection while the recording that creates privacy and liability exposure is never made. The same capability supports rehabilitation, mobility and gait assessment, and continuous in-home health monitoring. Applications include hospitals and health systems, senior living and assisted living operators, home health providers, medical device and remote monitoring companies, and clinical research. Users may benefit from fewer injuries among older and hospitalized people, reduced avoidable health spending, and protection of personal privacy and dignity. This I-Corps project utilizes experiential learning coupled with first-hand investigation of the industry ecosystem to assess the translation potential of a sensing platform that enforces privacy in hardware at the physical sensing layer rather than through software policy or downstream data governance. This technology uses an imaging component that pairs a thermal sensor with a conventional camera and performs pose extraction on-device, replacing each person in real time with an anonymized skeletal or avatar representation that retains posture, movement, and activity while eliminating faces, clothing, and appearance. The output remains compatible with existing computer vision and machine learning models for fall detection, activity recognition, and physical therapy monitoring. In addition, the technology includes an acoustic component that attenuates speech-bearing frequency bands in hardware before digitization and extends sensing into the ultrasonic range, retaining non-speech signatures of motion, object interaction, and physiological events, with inference accuracy comparable to conventional full-band audio models. Both components run on low-power embedded hardware. Users gain a verifiable non-capture guarantee, which may permit accurate automated monitoring in regulated and privacy-sensitive spaces that prohibit conventional sensors. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

NSF Program Director: Ruth Shuman