LDA-Based CSI Amplitude Fingerprinting for Device-free Localization

Abstract

With the advantage of not requiring additional hardware equipment and low application development costs, the CSI amplitude fingerprinting has attracted more and more attention and has become a research hotspot. However, the current position fingerprint information based on the CSI amplitude can only roughly characterize the differences between different positions, which is easy to make the subsequent localization process ambiguous, resulting in limited positioning accuracy. In order to solve the above problem, we propose a LDA-based CSI amplitude fingerprinting (LCAF) algorithm. Firstly, the algorithm uses pre-processing method to eliminate the interference of environment and noise. And then uses channel state information (CSI) and linear discriminant analysis (LDA) to generate a fingerprint, a more stable and distinguishable than the original fingerprint, which better characterizes the impact of the target on the wireless signal. Finally, the localization of the estimated target is determined by the data fusion algorithm based on Naive Bayesian. The experimental results show that the accuracy of the LCAF algorithm is 13% higher than the FapFi localization algorithm, and 42% higher than FIFS localization algorithm.