Konstantin,

Within Durbin's (for that matter, also Levinson's) recursion, you have
to calculate the residual energy at each order step. So, all you need is
to test it and break out of the loop when underrun occurs.

Regards,
Miguel

geistgesicht wrote:
> Dear group,
>
> The impulse response of a pure recursive filter with linear prediction
> coefficients as filter coefficients is a good spectral description of
> a signals spectrum but the Levinsons recursion always calculates a
> fixed number of coefficients.
>
> Does anyone know an algorithm that calculates linear prediction
> coefficients and stops calculation after a certain amount of
> error-energy is under-run?
>
> Regards,
> Konstantin
> Start a New Thread

Dear group,

The impulse response of a pure recursive filter with linear prediction
coefficients as filter coefficients is a good spectral description of
a signals spectrum but the Levinsons recursion always calculates a
fixed number of coefficients.

Does anyone know an algorithm that calculates linear prediction
coefficients and stops calculation after a certain amount of
error-energy is under-run?

Regards,
Konstantin Start a New Thread