Indoor wireless localization has attracted considerable attention with great improvements achieved in wireless technology in past decades year. In this paper, we address the problem of sparse beacon deployment due to an incomplete signals acquisition in the real-world scenarios. Considering the sparsity nature, it motivates us to exploit Compressive sensing (CS) algorithms for proximity service(PBS) using Bluetooth Low Energy beacons by referring to their success in indoor positioning system. For this purpose, a compressive sampling matching pursuit extended with generalized similarity filter is proposed and also concern about the effect of different similarity measures. In addition, another approach using a two-phase neural network including Deep Neural Network(DNN) and Stacked Denoising Autoencoder(SDA) to cope with predict the location of a mobile device since it has an excellent performance on extracting and reconstructing data among multitudinous deep learning algorithms. Simulation results show that the accuracy of the two-phase neural network can reach 0.875, and the accuracy of the generalized similarity filter for chord distance measurement can achieve 0.9, which can be considered as a better performance.
