A weighted min-norm method with a recursive implementation

The min-norm method is related to the MUSIC bearing estimation method. This relation suggests a weighted version of min-norm method obtained from the eigenvalue method of Johnson and DeGraaf (1982). The method is obtained by weighting the eigenvectors with the inverse of the eigenvalues. Some experimental results illustrate the comparative behavior of the algorithm. Both the weighted and unweighted methods were implemented with a recursive algorithm that performs the spectral decomposition of the correlation matrix and is based on the properties of elemental matrices called dyads. The problem is reduced to an eigendecomposition of the sum of a diagonal matrix with a dyad, which is easily obtained. This computation can be executed in high degree of parallelism.<
>