A New Approach to Detecting Deterministic Periodic Components in Noise

摘要

A new approach to detecting Deterministic Periodic Components of non-stationary time series corrupted by random component is considered. The approach is based on usage of the new time series, which is the union of the principal singular vectors of the data matrix of the original time series. The approach has universal character as it allows significant improvement of resolving capacity of any existed method of pseudospectra estimation. The comparative examples, illustrated practical efficiency of the approach relatively to existed methods of pseudospetra estimation, are considered.