Indoor localization using software defined radio: A non-invasive approach

Abstract

In the field of data-driven healthcare systems, wireless-based human activity monitoring in an indoor environment is getting popular. Localization of human activities can help patients to get remote healthcare. Wearable devices are employed in modern healthcare monitoring systems, despite the fact that they can be inconvenient and expensive. Several studies have shown that analyzing variations in Channel State Information (CSI) can be used to identify activities in an indoor environment and monitor health utilizing radio frequencies. The results of an experiment using Software Defined Radio (SDR) to identify and localize human activities while a human subject performed six different classes of activities (no-activity, leaning forward, sitting, standing, walking-Rx, walking-Tx). The CSI extracted from the communication channel between two Universal Software Radio Peripherals (USRP) devices communicating and operating as SDRs was used to collect a total of 2100 sample data. The Extra Tree (ET) classifier outperformed the prior benchmark approaches in localizing six activities across the 21 classes. The proposed ET classifier can identify if the room is occupied and a human subject’s walking directions in two separate zones, as well as localize six distinct activities inside the room with an accuracy of 91% and 100% respectively.