A Wifi Positioning Algorithm Based on Deep Learning

Based on the traditional Wifi positioning algorithm, the offline Wifi fingerprint database has a large number of redundant APs, which is easy to cause dimensionality disaster. Based on this situation, this paper proposes a Wifi positioning algorithm based on SVM-kmeans-ReliefF-SAE. First, the wifi fingerprint database is divided into building and floors by SVM algorithm, and the intra-floor region is divided by the K-means algorithm. Then the feature selection of the constructed Wifi fingerprint database is carried out. The redundant wifi offline fingerprint database of the divided area is removed by ReliefF algorithm, and the reduced offline fingerprint database is established. Finally constructing a deep neural network for the reduced wifi offline fingerprint database through the stack auto encoder. By pre-training the deep neural network, we can obtain a more accurate Wifi indoor positioning model. The experimental results by using the standard database UJIIndoor Loc show that the algorithm can effectively remove the redundant AP information in the offline Wifi fingerprint database and improve the positioning efficiency.