A method for an accurate estimation of the power spectral density - psd- of noisy signals in the frequency domain is presented. The estimation technique utilizes nonlinear averaging of weighted overlapped time slices of the data sample. The analytical expressions of bias and variance are derived and the optimal percentage of overlap is obtained in terms of the used window. Finally, the statistical properties of this method are compared with those of the modified version of the Welch’s method; it is shown that, even in this case, the nonlinear means produce a better behaviour in presence of additive noise.
|Titolo:||A new technique to perform highly accurate psd estimates through non-linear averagings|
|Data di pubblicazione:||1996|
|Nome del convegno:||IMEKO/XIV|
|Appare nelle tipologie:||4.1 Contributo in Atti di convegno|