Abstract
The present paper gives high resolution spectral line estimation algorithms which involve optimized prefiltering of the observed signals consisting of sinusoids and white noise, before applying the linear phase prediction method. Two types of the prefilter are investigated for resolution enhancement: in the first one the coefficients are determined so as to minimize a specific criterion, and in the other one the center frequency and the bandwidth of the bandpass filter are decided to minimize an alternative criterion. It is clarified that the two kinds of prefiltering can improve significantly the resolution in the estimation of closely spaced frequencies and can reduce the biases in the estimates even if we utilize a low-order prediction model. The prefilters can be determined automatically in a data-adaptive way in accordance with the data number, signal-to-noise ratio and signal frequencies.
