Abstract :
Prony´s method of spectrum analysis models a time series as a linear combination of complex exponentials plus a white noise. The performance of the method is very dependent on the peculiarity of the signal to be analyzed. Four algorithms for autoregressive estimation are experimentally compared to provide valid indications for the choice of the most suitable one for estimating the Prony´s parameters. L. Marple´s algorithm (1980) seems to be the best, while well-known facts about bias in frequency estimation produced by Burg´s algorithm are confirmed; nevertheless, it performs better than the covariance and the singular-value-decomposition-based algorithms
Keywords :
estimation theory; spectral analysis; white noise; autoregressive estimation; bias; complex exponentials; extended Prony´s method; frequency estimation; spectrum analysis; time series; white noise; Covariance matrix; Damping; Frequency; Frequency estimation; Least squares approximation; Least squares methods; Parameter estimation; Performance analysis; Polynomials; Signal analysis; Singular value decomposition; Time series analysis; White noise;
