Location estimation of multiple sources based on direction of arrival applying compressed sensing theory

摘要

Diction of arrival algorithms have been widely used in positioning systems. However, they have important restrictions to take into account in terms of spatial and temporal stationary requirements between the sources and correlation proprieties between them and the noise. Nevertheless, due to its nature the localization problem could be posed like a sparse reconstruction problem, and is possible to apply the compressive sensing and sparse reconstruction theory to estimate the position of several non-collaborative sources. Besides, considering a joint estimation system as we propose in this work, is possible to exploit both inter and intra-correlation signal aiming to improve the accuracy estimation. Methodology: In this work we simulate a localization system composed by several reference nodes (RN) which share information with a central entity named fusion center (FC) where the target estimation will take place. The process is divides in two stages: offline and online. In the first one we discretize the region of interest (ROI) in K candidates position where the sources could be located. Each RN builds its own dictionary that contains the covariance matrix of the steering vector for each cell into the grid. In the online stage, the target position estimation is performed. To do so, each RN receives the signal from the sources and calculates the compressed version of the covariance matrix, which is sent to the FC. In the FC the orthogonal matching pursuit (OMP) is performed to estimate the target coordinates inside the ROI. Results: The results show the system performance in terms of accuracy in the position estimation when parameters like number of sensors, system’s noise and compression rate in the measurement matrix are varied.Conclusions: The proposed method provides high accuracy in the estimation without restricting requirements on the spatial and temporal stationery and correlation properties of the sources and the noise, which are common in traditional direction of arrival algorithms. Financing: Miniciencias Colombia and Pontificia Bolivariana University.