Aiming at the problems of low localization prcision

fingerprint drifting caused by unstable signal during fingerprint localization

Kalman filtering is used to preprocess the collected data of Bluetooth received signal strength（ RSS）

and initially clustering on data is carried out by K-means algorithm. Distance between the data to be tested and each cluster centroid are calculated. Class cluster data

neighboring to data to be tested

are fused.For new data subset formed by fusion

corresponding support vector machine（ SVM） model is trained

and classification of data to be tested are achieved. By test

localization precision of algorithm is stable within 1. 5 m

the prospective goal is achieved.