FluSense: A Contactless Syndromic Surveillance Platform for Influenza-Like Illness in Hospital ...
summary
FluSense is a deployable edge system built from a Raspberry Pi, microphone array, thermal camera and Intel Neural Compute Stick that runs on-device models and logs 350,000 thermal images and 21 million non-speech audio samples over seven months in hospital waiting areas. A CNN cough detector reaches F1 0.88 after transfer learning, a Faster-RCNN thermal people detector gives 0.9 mAP and people counts correlate 0.93 with patient visits. Cough-based features raise daily influenza-test prediction from 0.48 to 0.59 and positive-case correlation from 0.29 to 0.40, with p < 0.05 in walk-in and pediatric rooms.
AI-written summary · about