Minimum Variance Interpolation and Spectral Analysis

KAUSHIK MAHATA; DAMIÁN MARELLI

Santorini

Conferencia; International Conference on Digital Signal Processing (DSP); 2009

Institute of Electrical and Electronics Engineers (IEEE)

Interpolation and spectral analysis of signals from finite num-ber of samples is considered. When the observed data is offinite length, interpolation and spectral analysis of bandlim-ited signals using Shanon's framework leads to erroneous re-sults. In spectral analysis this phenomenon is known as thespectral leakage problem. In this paper we address this issuefrom a minimum variance estimation perspective, and treatthe generic case where the signal is not necessarily bandlim-ited. In contrast to traditional windowing based methods, theminimum variance framework leads to a convolutive transfor-mation of the data, which employs a linear predictor. Simula-tions indicate a significant improvement in the performance.