A comparative performance analysis of some discrete Fourier transform based frequency estimators

This thesis is about frequency estimation. FFT interpolation method is used for this estimation. Fast Fourier transform (FFT) is the general name given to algorithms that compute the discrete Fourier transform of a sequence. Interpolation is an application that is used to find unknown values from known values. In the study, complex sine data model is used as the data model. With FFT interpolation, it is aimed to compare some frequency estimators in the noisy and noiseless conditions experimentally in MATLAB environment. These estimators; It is an improved version of Quinn's estimator, as well as the estimators of Quinn, Jacobsen, and Macleod. 0, 0.5 and 1 values were chosen for the noise variance. The algorithms presented by Ligges and Quinn are the basis of the estimators in this thesis. Jacobsen's estimator showed the best bias performance. The Jacobsen and Quinn algorithms offer simpler computation and very good variance performance, while the Macleod and Improved Quinn methods offer excellent variance performance but are slightly more complicated to compute. The choice of the estimator can be made by striking an acceptable balance between the estimator's performance and computational complexity depending on application and design constraints.