Improved Fourier analysis using parametric frequency-domain transfer-function estimators

Frequency components (or sinusoids) can be detected from noise-corrupted signals by means of FFT-based Fourier analysis techniques. The main drawbacks of FFT-based techniques are their poor frequency resolution and the presence of leakage errors. This paper shows that by means of frequency-domain transfer-function estimators it is possible to eliminate the leakage errors and to improve the resolution of the frequency estimates. The advantages of this approach are illustrated by means of both simulations and measurements performed on two applications on rotating machinery.