Device-Free Indoor Localization Based on Supervised Dictionary Learning

As a promising intelligent localization technology, device-free localization (DFL) is an area to be developed urgently. We propose a supervised dictionary learning algorithm to model DFL. The supervised dictionary learning algorithm can accurately update the columns in the dictionary and train a linear transformation matrix for target localization. In the regularization item of dictionary learning, we use generalized minimax-concave (GMC) regularization to replace the l0-norm to obtain accurate and tractable solutions. We deploy a sensor network in the laboratory environment to perform localization experiments. In the current experimental environment, our proposed algorithm can achieve 100% localization accuracy. We add Gaussian-distributed noise to all experimental data to test the anti-noise performance of the proposed algorithm. When the signal-to-noise ratio (SNR) is 10dB, our proposed algorithm can still achieve 100% accuracy which outperforms the state-of-the-art algorithms. Moreover, we show the performance improvement of the supervised model to the unsupervised model.