Research And System Design Of Support Vector Regression Learning Algorithm For Indoor WLAN Positioning

With the rapid development of wireless communication technology,and the increasing demand for location services,positioning technology is attracting more and more attention.In wide outdoor environments,the global positioning system and cellular network positioning system has been developed,which can meet the demand of pinpoint accuracy,however,in the indoor environment where the building shading is serious,the positioning system can not receive the effective signal and does not meet the requirement of positioning accuracy.Due to the WLAN location technology can use the existing public network,and it only needs intelligent terminal that has spread,which can achieve indoor positioning,and pinpoint accuracy is relatively high,so it can become a hot research direction of indoor location technology.In WLAN indoor positioning,the received signal strength(RSS)is affected by many factors,showing serious uncertainty and non-linearity,resulting in RSS and physical location is not one by one mapping relationship between seriously affect the positioning accuracy,prompting In this paper,the SVR learning and localization algorithm is adopted,which has better generalization and higher positioning accuracy.In view of the fact that SVR location calculation directly introduces a lot of noise information,the original RSS signal is processed by feature extraction to reduce the uncertainty of RSS.However,the traditional feature extraction method can not make use of the nonlinear characteristic of RSS effectively.In this paper,the kernel-based direct discriminant analysis method can make full use of the non-linear characteristic of RSS signal,increase the credibility of RSS information,and improve the localization accuracy of SVR learning localization algorithm.The SVR learning localization algorithm performs sample learning for all regions will increase the amount of algorithm,prone to learning problems,so this paper uses whitened RSS signal k-means clustering algorithm,SVR learning localization algorithm can be limited to a relatively small area,And the whitening processing can solve the problem that the clustering accuracy is not high due to the correlation between the signals.Compared with the traditional algorithm,the localization error is increased by 37% within 1 meter,and the workload of positioning fingerprints at each reference point is reduced by 80%.Improve the positioning accuracy and positioning efficiency.Based on the study of the above algorithm,this paper introduces the software development environment of positioning system,and designs the WLAN indoor positioning system on Android platform.