Effectiveness of simulation data on walking in Wi-Fi fingerprints using RNN

Abstract

This study focuses on Wi-Fi fingerprint localization using RNN. Walking survey is a method used for collecting fingerprint datasets. This method can collect continuous walking data while the measurer is walking. However, multiple walking patterns must be measured to accommodate users with different walking speeds and routes. We propose a method for generating simulation data with different walking patterns. The simulation data were created using the coordinate adjacency of the measurement data. Further, we evaluated the model trained with and without the simulation data. Results demonstrate that the accuracy of the proposed method shows its effectiveness.