gftool.linearprediction.pcoeff_burg

gftool.linearprediction.pcoeff_burg(x, order: int)[source]

Burg’s method for linear prediction (LP) coefficients.

Warning

We found this method to be instable, consider using pcoeff_covar instead.

Burg’s method calculates the coefficients from an estimate of the reflection coefficients using Levinson’s method. Burg’s method guarantees that poles are inside the unit circle [Kay1988], thus it is stable.

Parameters:

x(…, N) complex np.ndarray: Data of the (time) series to be predicted.
orderint: Prediction order, has to be smaller then N.

Returns:

a(…, order) complex np.ndarray: Prediction coefficients.
rho(…) float np.ndarray: Variance estimate.