Spectral estimation through cubic-spline approximation of a discrete time series (Q801656)

A method for obtaining accurately the Fourier transform of continuous data is developed. A discrete time series is generated by uniformly sampling the data during a finite period. The method is based on cubic- spline approximation of the data. An analytical expression for the Fourier coefficients is derived in explicit form. The fast Fourier transform algorithm is used in the numerical computations. A new procedure to estimate spectrum and correlation functions for continuous random data is developed. The intrinsic properties of the procedure are clarified through a series of test computations.