Indoor Fingerprint Localization Algorithm Research Based On Support Vector Regression Mechine

With the explosive growth of the Internet of things and mobile Internet applications,people's access to location information is more and more urgent.However,traditional positioning sciences,such as the global positioning system(GPS)and the Beidou navigation system(BDS),cannot be accurately positioned indoors.At present,WIFI network has been widely deployed indoors,covering most areas of people's life,such as schools,shopping malls,airports and hospitals.Moreover,WIFI network has the advantages of high transmission rate,low cost and wide coverage,etc.WIFI access module has been widely installed to various wireless terminals,so there is a broad application prospect in indoor location fingerprint positioning which uses existing WIFI network resources.Based on a large number of WIFI access point(AP)can connect field acquisition RSSI sample scenario,adopting the method of combination of theory and experiment to analyze localization fingerprint characteristic.Find that if using all directly access AP signal to position,there will be high computational complexity and interference which may affect the precision of positioning.Therefore,after comparing and analyzing the dimensionality reduction e effects of various mainstream dimensionality reduction algorithms on fingerprint features,kernel principal component analysis is adopted to extract features from the original fingerprint database to eliminate the influence of noise.Based on the statistical characteristics of nonlinear,non-gaussian RSSI,using support vector regression machine based on kernel function to build indoor location model,but the performance of support vector regression machine is influenced by its own parameters and kernel function parameter,and the traditional parameters optimization algorithm at the expense of the time and the computational complexity to get better optimization effect.Therefore,after analyzing the reason why the traditional parameter optimization algorithm is not effective,the immune particle swarm optimization algorithm is put forward.It is proved that the immune particle swarm optimization algorithm saves a lot of computing time.Finally,the KPCA algorithm is combined with the support vector regression positioning model of optimal parameters to carry out experimental simulation on the data collected on the spot and compare with other positioning algorithms.The result shows that the proposed algorithm performs well in positioning accuracy and efficiency.