Skip to Main Content
A new efficient expectation-maximization (EM) algorithm for ML estimation is presented for multiple sources localization using a wireless sensor network (WSN). It uses acoustic signal energy measurements taken at individual sensors to estimate the locations of multiple acoustic sources. Instead of already existent multi-resolution (MR) search of projection solution and existing EM algorithms for ML estimation, the basic idea of our method is to decompose the observed sensor data (signal energy), which is a superimposition of multiple sources, into individual components and then estimate the corresponding location parameters separately. Our proposed EM algorithm involves two steps, namely expectation (E-step) and maximization (M-step). In the E-step, signal energy of sensors received from individual source is estimated. Then, in the M-step, the maximum likelihood estimates of the source location parameters are obtained through a global grid search. The two steps are iteratively repeated until the pre-defined convergence is reached. Simulation results show that our proposed EM algorithm have a good performance and is a better solution for ML estimation which approach a good trade-off between estimation error and computation complexity.