Parametric characterization of random processes using Prony's method
Wentworth Smith, DARREL L. LAGER · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 1979
The compact parametric characterization of a random process is often valuable in problems involving signal classification and system identification. The method of Prony suggests two different approaches to obtaining such a characcterization from a finite record of sampled data. The first approach is to extract the parameters of the estimated autocorrelation waveform. The second approach is to estimate the parameters of a white-noise-driven linear system that generates an output with the same characteristics as the given random process. These two approaches are compared with other common signal processing methods. Also, the performance of each approach is evaluated theoretically and with computer-simulated and experimentally recorded data. Each approach is found to characterize rapidly both stationary and nonstationary random processes. The first approach is superior for the accurate parametric characterization of a stationary process. Alternately, the more precise second approach is best for monitoring changes with time in random process parameters. 20 figures, 6 tables.