Pre- and Post-Processing Algorithms with Deep Learning Classifier for Wi-Fi Fingerprint-Based Indoor Positioning

摘要: To accommodate the rapidly increasing demand for connected infrastructure, automation industrial sites and building smart cities, development of Internet Things (IoT)-based solutions is considered one major trends in modern day revolution. In particular, providing high precision indoor positioning services such applications a key challenge. Wi-Fi fingerprint-based systems have been adapted as promising candidates applications. The performance degrade drastically due to several impairments like noisy datasets, variation signals over time, fading multipath propagation caused by hurdles, people walking area under consideration addition/removal access points (APs). this paper, we propose data pre- post-processing algorithms with deep learning classifiers positioning, order provide immunity against limitations database environment. addition, investigate proposed system through simulation well extensive experiments. results demonstrate that pre-processing algorithm can efficiently fill missing received signal strength fingerprints database, resulting success rate 88.96% 86.61% real-time experiment. improve from 9.05–10.94% conducted experiments, highest 95.94% 4 m positioning.