Description: In the article, the problem of increasing of efficiency of spectral analysis of the signals observed on the noise background by methods based on subspaces of eigenvectors of covariance (or correlation) matrix of input sequence is considered. The examples of such methods are Root-MUSIC, ESPRIT, Min-Norm, Fine. In order to solve this problem the preliminary data processing using the SSA (singular spectrum analysis) method which is the one of the noise reduction methods is performed. Furthermore, the persymmetric property of covariance matrix of the data (the maximum likelihood estimate of persymmetric covariance matrix) is used. The simulation results are presented, where the performances of Root-MUSIC, Root-MUSIC with using persymmetric property of covariance matrix, Root-MUSIC with using SSA and Root-MUSIC with using SSA and persymmetric property are compared. They confirm the increasing of efficiency of spectral analysis by methods based on subspaces of eigenvectors of the covariance matrix of input sequence in the case of joint using the SSA method and the persymmetric property of covariance matrix of the data. The results of paper can be used for performance improvement of radiotechnical systems. It is of interest to generalize the obtained result to real signals of communication systems. It is recommended to use the graphical processor units to realize the eigenvalue decomposition of data covariance matrix.
Keywords: signal subspace, spectral analysis methods, eigenvectors, eigenvalues