Scene Recognition for Device-Free Indoor Localization

Device-free indoor localization methods based on channel state information (CSI) have become an increasingly important topic. In complex indoor environments, both the line-of-sight (LOS) and non-LOS (NLOS) areas coexist, and the optimal parameters of the localization models for these two areas are different. To address this problem, in this article, the scene-recognition indoor localization (SRIL) method is proposed to identify LOS and NLOS areas. First, the scene recognition model is given by combining mutation particle swarm optimization (MPSO) with a backpropagation (BP) neural network. Then, the feature transformation method based on discriminant correlation analysis (DCA) is presented, which can effectively explore the correlation of amplitude data and phase data and form high-resolution location fingerprints, and further improve indoor localization accuracy. Experimental results show that compared with the parallel AdaBoost indoor localization (PAIL), LDA-based CSI amplitude fingerprinting (LCAF), long short-term memory (LSTM), broad learning system (BLS), passive indoor localization based on CSI and Naive Bayes (PCNB), AdaBoost positioning system (ABPS), and FapFi algorithms, the proposed algorithm has higher localization accuracy.