Radio Tomography Imaging based on Gaussian Mixture Model

Radio tomographic imaging (RTI) system needs to have enough observed samples to ensure the performance of reconstruction. The method based on compressed sensing (CS) can effectively solve the problem of insufficient observed samples. However, this method cannot solve the problem that wireless signals in RTI network are affected by environmental noise, which results in the decrease of positioning accuracy. To solve this problem, we analyzed the characteristics of the received signal strength (RSS), used the Gaussian mixture model (GMM) to determine the acquired RSS observation samples to obtain the effective link in the wireless network, and then reconstructed it by the alternating direction multiplier method (ADMM).The experimental results show that when the voxel size is set as δ= 0.11m , the average positioning accuracy is improved by 36% and 9%, respectively, compared with the classical orthogonal matching pursuits (OMP) algorithm and Tikhonov algorithm. In the case of the same data, the reconstruction time based on GMM-ADMM algorithm is only 0.06 seconds, which is conducive to the real-time performance of the system.