Multiantenna Fusion Fingerprint Localization Based on Weighted Broad Learning System

For multiantenna indoor localization methods, data concatenation can be regarded as equal-scale fusion of multiantenna data and ignores the reliability difference between antenna samples. To deal with this problem, we first propose a multiantenna feature fusion network based on weighted broad learning system (WBLS), which uses the weighted penalty factor to constrain sample contribution on the feature fusion model and obtain more discriminative fingerprint features. Second, we present an amplitude-phase fusion (APF) scheme based on Spearman’s correlation coefficient, which uses the nonlinear correlation between antenna pairs to realize an effective fusion of amplitude and phase information at different locations. Experimental results show that compared with ensemble broad learning localization (EnsemLoc), parallel AdaBoost indoor localization (PAIL), broad learning system (BLS), AdaBoost positioning system (ABPS), long short-term memory (LSTM), and so on, the proposed algorithm has higher localization accuracy.