Improved Fingerprint Localization Algorithm based on Channel State Information

At present, most indoor localization technologies based on Wi-Fi relay on received signal strengths. Due to the interferences of multipath and environment noise, the localization accuracies of these methods are not high enough. The channel state information (CSI) can describe channel state more accurate and has a stronger stability, which is dramatically suitable to localization. This paper proposes an improved fingerprint localization algorithm based on CSI, which combines the RSSI-distance and CSI model jointly. With the distances from APs to the point to be located, this algorithm narrows the matching range and reaches a higher localization accuracy. Then, to better evaluate the influence of the support vector machine's (SVM) key parameters on positioning error by the fingerprint database and online data, a sample capacity expansion method is further proposed. The experiment and simulation results show that the localization accuracy with CSI outperforms that with RSSI. And the proposed algorithm improves the performance of SVM algorithm based on CSI.