MSDFL: a robust minimal hardware low-cost device-free WLAN localization system

The research on fingerprint-based device-free (DF) indoor localization has attracted great interest due to the ubiquitous of radio frequency signals and its accuracy. Most such approaches typically require a large number of transmitters and receivers to achieve acceptable accuracy, and there is another problem that the accuracy is degraded due to the expiration of the fingerprint database over time. In this paper, we present an accurate fingerprint-based device-free localization system using only a single stream over time, termed MSDFL. By analyzing the CSI fingerprint patterns and comparing the matrix similarity between CSI fingerprints and testing CSI samples matrix, the proposed MS algorithm is able to provide accurate DF localization with only a single stream. To cope with the noisy CSI samples and remove the subcarriers greatly affected by multipath, a novel CSI pre-processing scheme is applied on CSI data to reduce the noise and extract the most contributing subcarriers. In addition, to overcome the decrease in positioning accuracy caused by the expiration of the fingerprint database, a rigorously designed update scheme using artificial neural network is utilized for the renewal of fingerprinting database, which further enhance the performance of our system over time. Experimental results are presented to confirm that MSDFL can effectively improve location accuracy, compared with the other existing methods in representative indoor environment.