Most existing fingerprints-based indoor localization approaches are based onsome single fingerprints, such as received signal strength (RSS), channelimpulse response (CIR), and signal subspace. However, the localization accuracyobtained by the single fingerprint approach is rather susceptible to thechanging environment, multi-path, and non-line-of-sight (NLOS) propagation.Furthermore, building the fingerprints is a very time consuming process. Inthis paper, we propose a novel localization framework by Fusing A Group OffingerprinTs (FAGOT) via multiple antennas for the indoor environment. We firstbuild a GrOup Of Fingerprints (GOOF), which includes five differentfingerprints, namely, RSS, covariance matrix, signal subspace, fractional loworder moment, and fourth-order cumulant, which are obtained by differenttransformations of the received signals from multiple antennas in the offlinestage. Then, we design a parallel GOOF multiple classifiers based on AdaBoost(GOOF-AdaBoost) to train each of these fingerprints in parallel as five strongmultiple classifiers. In the online stage, we input the correspondingtransformations of the real measurements into these strong classifiers toobtain independent decisions. Finally, we propose an efficient combinationfusion algorithm, namely, MUltiple Classifiers mUltiple Samples (MUCUS) fusionalgorithm to improve the accuracy of localization by combining the predictionsof multiple classifiers with different samples. As compared with the singlefingerprint approaches, the prediction probability of our proposed approach isimproved significantly. The process for building fingerprints can also bereduced drastically. We demonstrate the feasibility and performance of theproposed algorithm through extensive simulations as well as via realexperimental data using a Universal Software Radio Peripheral (USRP) platformwith four antennas.
展开▼
