Neural Network Based on Learning Method for WiFi Metric in Signal Space
In WiFi based indoor positioning,reasonable metric definitions are needed in the signal space to find the correlations in the signal space.The state-of-art metrics in the signal space are mostly apriori metric definitions and their combinations.However,for practical applications,the best metric definition is not among the apriori metric definitions and their combinations.In this paper,a neural network based learning method is proposed,which can predict the adjacency relations in the position space to indoor po-sitioning through learning from observations.At last,an open source dataset is adopted to experiment over the proposed method and the results have validated the proposed method.Compared with traditional metrics such as Euclidean metric,Jaccard metric,rank based metric and CMD compound metric,the fi-nal average positioning error has a decrease of about 0.6m,1.8m,1.3m and 0.4 respectively,which shows that the proposed method can indeed increase the positioning accuracy.
indoor positioningneural networkmetric in signal space
NETL
NSTL
万方数据
