Signal Feature Extraction Algorithm In WLAN Indoor Localization

In recent years, due to the wide deployment of WLAN in indoor environment,people can access to wireless network conveniently through mobile phones、laptops orPDA and so on. Meanwhile, people pay more and more attention on location-basedservices. Indoor positioning system based on WLAN has attracted extensive attention,for the reason that GPS cannot be applied to the indoor environment. In WLAN indoorpositioning system, fingerprint localization algorithm has high location precision andstrong universality, and don’t need additional device. So it has been widely used.Fingerprint localization is divided into two stages: off-line stage and on-line stage.During the off-line stage, we build the database. During the on-line stage, we comparethe received signal with database and obtain the coordinate.At present, research pay attention on matching algorithm on on-line stagewhile ignore the process that extracting signal feature. But the positioning errorof WLAN location fingerprinting is partly due to the noise mixed in the collectedsignal in complex indoor environment. And because there are more and moreAPs in WLAN environment, the database built on off-line stage is getting largerand larger, which will influence the real-time property on on-line stage. In order toreduce the effect of noise and the amount of calculation, this paper propose asignal feature extraction algorithm based on reducing the workload andimproving positioning accuracy.Firstly, this paper analysed the process of fingerprint localization algorithm,simulated the matching algorithms and took KNN algorithm to test thepositioning accuracy. And then, this paper analyzed the characteristics of thesignal, preprocessed the samples of data, and removed the data has gross error.Secondly, the signal feature extraction algorithm is analyzed in two aspects:reducing the workload and improving positioning accuracy. In the aspect ofimproving positioning accuracy, this paper proposed to take the logarithmicaverage of the signal’s power spectrum to KL transform, as to reduce the effecton positioning accuracy brought by the noise. In the aspect of reducing theworkload, this paper used the improved K-means clustering algorithm to dividethe positioning area into sub-areas, then selected APs with strong positioningresolutions in each sub-areas in order to enhance the real-time property onon-line stage.Finally, by collecting data and performing in the real WLAN environment,experimental result proves that the proposed signal feature extraction algorithm effectively improves positioning accuracy and reduces the calculated amount onon-line stage.