Analysis of parametric model signal processing techniques for signature analysis

Abstract

Five parametric modeling techniques have been identified to be possible alternatives to the Fast Fourier Transform (FFT) for signature analyses involving short data records. The developments in signal processing that have lead to these techniques are reviewed. Mathematical definitions for parametric models are provided in terms of time-domain stochastic difference equations as well as in terms of frequency-domain rational transfer functions. Computer programs are developed for implementation of each of the five parametric modeling techniques. Results are presented from studies conducted on simulated stochastic signals to characterize the performance of the parametric modeling techniques and to compare the performance of these techniques to the FFT. One of the parametric modeling techniques, Pisarenko Harmonic Decomposition shows outstanding performance in comparison to the FFT and to the other parametric modeling techniques.