Object Classification Based on Channel State Information Using Feature Spaces

Abstract

To date, a large amount of research has been carried out on the use of a Wi-Fi signal for positioning or classifying objects using Channel State Information (CSI). This paper explores the classification of metal objects placed between routers at a distance of 1 meter. In the future, this can be used as an additional means of control in narrow door openings. Based on the CSI amplitude values, feature spaces are calculated, which contain concentrated information about the classified objects. Feature spaces plots are shown. The result of the classification of physical Wi-Fi objects using the Random Forest algorithm was 83% using raw amplitude CSI values. Using feature spaces result was 99% with best combination of features.