Nonlinear Detection of Weak Pseudoperiodic Chaotic Signal Frequencies from Noisy Environment

The extraction of weak pseudoperiodic chaotic signal frequencies from white Gaussian and colored additiwe noise is investigated by applying the nonlinear signal detection algorithms, based on phase-space embedding technique, principal component analysis and power spectral analysis. By analyzing Rossler chaotic time series, it is demonstrated, that the detection algorithm based on standard eigenvalue decomposition performing to the averaged covariance matrix of the reconstructed phase space matrix and the detection algorithm based on eigenvalue decomposition performing to the time-delayed covariance matrix are able to detecting of weak pseudoperiodic chaotic (or regular sinusoidal) signals hidden beneath the additive Gaussian or colored noise floor at SNR up to –20 dB. It is concluded, that these algorithms are preferable to the SPCA algorithm by detecting weak pseudoperiodic chaotic or sinusoidal signals buried in a additive Gaussian noisy background. Ill 4, bibl. 14 (in English; summaries in English, Russian and Lithuanian).