Skip to Main Content
Measurements of signals recorded via digital recording equipment often contain blocks of missing data due to equipment errors, the nature of the process under observation, or physical limitations of the experiment or hardware. Therefore, it is of great interest to reconstruct these signals in terms of their time or frequency domain characteristics, or in this case their power spectra. Of particular interest here are those signals for which a parametric model is appropriate for describing their spectral content, notably auto regressive. The "Kim method" is an efficient algorithm for reconstruction of segmented autoregressive signals using an extrapolative method based on estimates of the autoregressive model parameters of the existing data. It has proven effectiveness for segmented autoregressive narrow-band data with pronounced peaks. This paper examines the effectiveness of the method on those signals that have less pronounced spectral peaks and are more broadband in nature. The performance of the estimator is quantified for these signals and compared to the performance attained with the narrow band signals.