Preliminary Evaluation of an Algorithm for Analysis of Stationary Random Data from a Multiple-Input Linear System.
Abstract
This paper deals with a preliminary evaluation of a mean-squared estimated algorithm for the extraction of a signal that is embedded in superposed stationary noise processes. The motivation for this study is the processing and interpretation of data obtained in recent experiments to measure gradients of magnetic fields induced by internal waves in a shallow ocean. The approach used is simulation of the signal and noise processes through construction of random time series for the signal and interfering noise processes whose spectra can readily be adjusted to study parametrically a wide range of situations. The principal conclusions are: (1) the algorithm, in the form commonly applied to the reconstruction of the output of the main data channel, is valid only if a very special form of linear relationship is assumed between each of the superposed time series in the main channel and the time series in each of the subsidiary channels; an assumption that can lead to grossly erroneous estimates of the signal spectrum and (2) severe cumulative bias errors can be introduced in the reconstructed signal, partial coherences and transfer functions through the algebraic operations necessary in the application of the algorithm. This procedure can lead to large uncertainties in the functional form of the reconstructed signal spectra and, even more important, to false indications of the state of partial coherence.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1979
- Accession Number
- ADA079329
Entities
People
- Henry Hidalgo
- Wasyl Wasylkiwskyj
Organizations
- Institute for Defense Analyses