A Novel Infrastructure WLAN Locating Method Based on Neural Network

摘要

Comparing to the client based locating, infrastructure based methods do not require install special software or hardware at client side. So it not only fits for real deployment, but also supports some specific requirement (ex target tracking). Former researchers mainly adopt k-NN method in infrastructure based locating. However, its computing complexity is proportional to the size of sample set, which makes it unscalable when the system grows large. This paper proposes a novel infrastructure WLAN locating method by utilizing neural networks and a new training method to overcome the effect of different power levels of client devices. Through real deployment and testing, the result shows that the computing complexity is much lower than the k-NN method, while the accuracy is very close.