WLAN Fingerprint Localization with Stable Access Point Selection and Deep LSTM

With the development of communication technologies, the demand for location-based services is growing rapidly. The presence of a large number of Wi-Fi network infrastructures in buildings makes Wi-Fi-based indoor positioning systems the most popular and practical means of providing location-based services in indoor environments. This paper proposes a machine learning indoor positioning method based on received signal strength. This algorithm considers the access point (AP) selection strategy to reduce the computational load and enhance noise robustness whereby improving the positioning accuracy. The local feature extraction method is used to extract powerful local features to further reduce the noise impact. We then employ the Long Short-Term Memory (LSTM) network to learn high-level representations for the extracted local features. The proposed method has been tested both in the simulation environment and the real environment. The experimental results show that the algorithm can greatly improve the accuracy and computational complexity of position prediction.