Device-Free Indoor Localization of a Person Based on Channel State Information

The work is devoted to the development of an algorithm for localizing a person in space based on CSI values received from a Wi-Fi router. During the work, a laboratory installation was developed and implemented. The proposed method is based on a fully connected neural network, which uses as input the phase and amplitude values calculated from the CSI data. Localization occurs based on information measured over 56 sub carriers and two pairs of antennas. For the practical part of the work, data from two different people who participated in the experiment were collected. Ultimately, more than 90 million phase and amplitude values were collected for each participant in the experiment, from which the neural network was trained. The experiments performed showed a localization accuracy of over 90%.