Estimation of the sinusoidal signal frequency based on the marginal median DFT

The marginal-median discrete Fourier transform (DFT) is used for estimation of complex sinusoidal signals embedded in an impulse noise environment. Expression for the marginal-median DFT of the sinusoidal signal in the neighborhood of the exact frequency is derived. Two specific displacement techniques are proposed in order to achieve an accurate estimation of frequency displaced from the frequency grid. They are based on specific ratio of the marginal-median DFT magnitudes for samples in the neighborhood of the marginal-median DFT maximum calculated over the frequency grid. Efficiency and accuracy of the proposed techniques is proved for mixed Gaussian and impulse noise environment.