Improved spectral estimation based on extrapolated and smoothed data records

The basic problem explored is the use of extrapolated and smoothed information to estimate discrete spectral components in additive white noise from a short finite length continuous-time record. Using approximations to optimal causal Wiener filters, smoothed and extrapolated data is obtained and used as an input to the Maximum Entropy and Classical Periodogram methods for spectral estimation. Results show a significant reduction of bias for the ME and periodogram methods at the expense of a small increase in variance for relatively low S/N-ratios.