Neural network-based indoor localization system with enhanced virtual access points

Wi-Fi indoor positioning systems are based on received signal strength indicator (RSSI) measurements and fingerprinting, by matching measured data with the database. Hence, generation of RSSI fingerprint database is essential for a Wi-Fi-based indoor positioning system; this requires significant time and effort. The study utilizes virtual access points to increase the number of access points in an indoor environment without necessitating additional hardware. Increases in the total access points is advantageous because it makes the database more granular, and the Kriging algorithm is introduced to solve the issue with less effort. The study also aims to apply deep learning neural network (DNN) in Wi-Fi fingerprinting using RSSI. The proposed system utilizes the neural network (NN) and Kriging algorithm to perform standard Wi-Fi fingerprinting, without difficulties in generating a fingerprint map. The result of a simple location estimation test led to an accuracy of 97.14%, thereby indicating that the application of NN and Kriging improves indoor localization. Further experiments should be conducted to test the complete effectiveness of the proposed system as compared to other systems employing DNN.