IMRP-Loc: An Indoor Localization Method Based on CSI Images of Multiple Reflection Paths

Fingerprint-based indoor localization methods have become an important technology for indoor localization due to their wide availability, low hardware cost, and fast growing localization needs. In this paper, we propose a method called IMRP-Loc based on channel state information, which uses images of multiple reflection paths information generated by the 2D-MUSIC algorithm as the input for machine learning. We train the machine learning model in the offline phase, and then test it in the online phase. The scheme was successfully implemented in two different classical experimental scenarios and validated with multiple network models, the test results were maintained at high accuracy and the feasibility of using the approach was confirmed by the experimental results.