Non-parametric power spectrum estimation techniques make use of FFT algorithm and are easy to compute. However, there are some limitations which makes them inappropriate to use in certain applications: due to finite length windowing of the data, these methods produce spectral leakage. Because of this, weak signals are masked and are difficult to detect in the estimated PSD. Other limitation is that in order to achieve a desirable frequency resolution non-parametric methods require long data records. Moreover, non-parametric power spectrum estimation techniques don’t make use of the information about the data-generating process in estimating the power spectrum density. Using parametric power spectrum estimation methods, however, it is possible to incorporate a model of the process into the power spectrum estimation techniques to get accurate and potentially better frequency resolution. Since parametric methods make use of the information on the generation of the data, these methods extrapolate the values the autocorrelation for lags m≥N (as opposed to assuming that the values are zero). This can also avoid the use of window functions and in turn the spectral leakage problem. A signal generation model and its parameters can be estimated from the observed data. This model and the estimated parameters can be then used to compute the power spectrum density. These methods are mostly useful to process short data records.