Robust Feature Construction for Human Activity Recognition from Wi-Fi Signals Perturbation

点击翻译按钮获取中文摘要

The project concerns the development of a new technology for reliable human activity detection andrecognition in residential environments by using signals emitted by off-the-shelf WiFi devices. Through theoriginal Engage grant, a method has been developed for building an initial localization model that includesreceiving and analyzing wireless signals, while a user is present (e.g., standing or walking) in differentpositions within sensing environment; a new proximity index has been introduced which gives a probabilisticmeasurement of the location of the user as the user is approaching or walking away from the WiFi device. Thegoal of the Engage Plus project is to develop a method for real-time evaluation of indoor localization phase thatincludes receiving a live stream of wireless signals and estimating a distance indicator using the probabilisticmodel built in the offine training phase - thus bringing the technology closer to market release where it can beused for low-cost intrusion detection or health monitoring of the elderly.