Indoor localization based on CSI in dynamic environments through domain adaptation

Abstract

As the demand for indoor localization applications continues to grow, device-free localization based on Wi-Fi Channel State Information (CSI) has become a popular research topic. Wi-Fi signals are, however, easily affected by environmental factors such as furniture changes. These factors disable the original localization system, and rebuilding it will cost a lot of time and workforce. This is a major challenge of device-free Wi-Fi localization. To address this issue, we use a transfer learning method, “Integration of Global and Local Metrics for Domain Adaptation (IGLDA),” and improve it, aiming to adapt the original localization model to the changing environment. Consequently, the localization accuracy is improved from 26.3 % to 82.2 % by only recollecting 37.5 % of data.