Computationally efficient PUMA algorithm for two-dimensional frequency estimation of a single-tone via DFT beamspace transformation

The problem of two-dimensional frequency estimation of a single-tone in white Gaussian noise is addressed in this paper, through a novel computationally efficient approach, referred to as beamspace principal-singular-vector utilization modal analysis (B-PUMA). Due to the centro-Hermitian structure inherent in the data model, the complex data matrix can be converted to real-valued via a DFT beamspace transformation. Thus, all calculations involved in the B-PUMA scheme are real-valued operations. Besides, to further boost the computational simplicity and estimation accuracy, an iteratively weighted least squares technique is employed to estimate the frequencies. Note that one iteration is sufficient for the B-PUMA to obtain comparable frequency estimates. Numerical results are included to demonstrate the effectiveness of our method.