An Improved Indoor Fingerprint Localization Algorithm

Aiming at the problem that the fingerprint localization algorithm needs to deploy more reference points (RP) when the offline fingerprint databaseis being constructed, a partition fitting approximation method (PFAM) is proposed.To reduce the workload of the fingerprint localization algorithm in the offline phase, firstly, the whole target environment is partitioned and the virtual RP is redeployed in each partition. Then, the third order polynomial log-distance path loss model is adopted to fit the environmental coefficient of each partition and the error vector of each partition is established.Then, the received signal strength (RSS) of the virtual RP is obtained by using the fitting model and the error vector inversely.And the C-means clustering is used to cluster offline fingerprintdatabase to reduce the amount of computation in the online phase. Finally, WKNN is used to locate the target in the online phase. Test results show that the PFAM algorithm can still achieve high location accuracy with few RPs and the mean location error is only about 1.2 m. Cumulative distribution function (CDF) shows that 86% of the location errors are within 2 meters.