Bispectral Inversion: The Construction of a Time Series from Its Bispectrum
Abstract
This report presents a method of producing a stochastic time series that has any desired bispectral characteristic. This was done using a quadratically nonlinear, infinite order, moving average model whose bicovariance was obtainable in terms of a finite linear function of the kernels. This expression was inverted to produce an equation for the kernels in terms of the bicovariance. Knowing the kernels, the time series can easily be computed. Alternately the DFT of the time series can also be computed directly. This actually easier than the time domain computation, as it only involves one summation as opposed to two. Keywords: Computer programs. (KR)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 13, 1988
- Accession Number
- ADA202508
Entities
People
- Peter S. Allison
Organizations
- University of Texas at Austin