A Convolutional Neural Network Based Device-Free Indoor Localization Using CSI

Device-free indoor localization has gained significant attention in recent years in the fields of wireless sensing and machine learning. This paper proposes an improved CSI-based device-free indoor localization (DFIL) system using a machine learning-based classification approach. The proposed system exploits the variations in wireless signals reflected from objects in the environment to locate and track targets without requiring any device to be carried by them. The system uses machine learning techniques such as random forest and support vector machines to classify the received signal strength indicator (RSSI) data into different target zones. The classification accuracy is improved by using channel state information (CSI) data alongside RSSI data. The experimental results show that the proposed system achieves higher accuracy than existing DFIL systems, with an average accuracy of 94% depending on the number of training samples. This enhanced localization method can be used in real-world settings like smart house automation systems, security, and the healthcare industry.