SPECTRAL ESTIMATION VIA STATIONARITY-CONSTRAINED BANDLIMITED EXTRAPOLATION
MCCASLIN, SHAWN ROBERT
Master of Science
Bandlimited extrapolation is beset with two problems that make it difficult to use as a spectral-estimation technique in many practical applications. First, the frequency support of the signal being estimated must be accurately known, information which is normally not available. Second, the signal must be of finite energy; for minimum error, the signal should ideally contain minimum energy, with most of its energy concentrated in time near the signal samples. Unfortunately, the energy in most real-world signals is not bounded in any known way. In this thesis, a deterministic energy-density stationarity constraint is applied to bandlimited extrapolation to reduce the dependence on these data. To demonstrate the attributes of this approach, this modified bandlimited-extrapolation algorithm is applied to two example applications--the short-time Fourier transform and finite-record filtering.
Electronics; Electrical engineering