PUMA: An Improved Realization of MODE for DOA Estimation

The method of direction estimation (MODE) offers appealing advantages such as asymptotic efficiency with mild computational complexity and excellent performance in handling coherent signals, which are not shared by conventional subspace-based methods. However, the MODE employs additional assumption and constraints on the symmetry of the root polynomial coefficients, which might cause severe performance degradation in the scenario of low signal-to-noise ratio/small sample size, since any estimation error will be enlarged twice due to the symmetry. Moreover, the standard realization for MODE does not have a closed-form solution for updating its estimates. In this paper, the optimization problem of MODE is proved to be equivalent to that of the principal-eigenvector utiliztion for modal analysis (PUMA) algorithm. We show that PUMA which has closed-form solution, that does not rely on any additional assumption and constraint on the coefficients, is a better surrogate than MODE for minimizing the same cost function. Extensive simulation results are carried out to support our standpoint.