Spectral Analysis of Short Record Time Series Data.
Abstract
Spectral estimation of data from some radar applications and seismilogical events is not accurate when short records are evalauted using traditional techniques. A record of data is short if the number of samples from the process is more than an order of magnitude smaller than the reciprical of the lowest frequency of interest. This analysis considers records of fewer than 128 samples. Techniques that produce improved frequency and amplitude resolution over smoothed periodograms and Fast Fourier Transforms (FFT) are considered. Specifically, the Burg Maximum Entropy Method (MEM) and Papoulis Bandlimited Extrapolation are derived. These techniques are shown to produce estimated that become unbiased and consistant. Additionally, the effects of windowing, a problem inherent with periodograms, are not observed in these techniques. Papoulis bandlimited extrapoloation techniques provided accurate results when short records are evaluated.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1979
- Accession Number
- ADA080156
Entities
People
- Paul B. Terry
Organizations
- Air Force Institute of Technology