mr.t
Aug20-08, 07:45 AM
Hi there!
When designing fir filters there are three different scenarios, smoothing m<0,
filtering m=0, prediction m>0, by setting your m-variable in the right-side wiener-hopf equation.
When designing non-causal wiener filters using
H(z) = \frac{P_{dx}(z)}{P_{x}(z)},
how do you specify what scenario you want? as you dont use wiener-hopf.
My guess is that as the filter is non-causal, all restrictions are lifted and you set
your impulse-respons as you want. If we for example would like a 2-step-prediktor
we simply define h(n) = \delta(n-2) \Rightarrow H(z)=z^{-2} and use that
frequency-response when calculating the cross-spectrum that is needed to solve
the IIR-filter equation shown above.
Is this correct? Thanks for any input on this!
When designing fir filters there are three different scenarios, smoothing m<0,
filtering m=0, prediction m>0, by setting your m-variable in the right-side wiener-hopf equation.
When designing non-causal wiener filters using
H(z) = \frac{P_{dx}(z)}{P_{x}(z)},
how do you specify what scenario you want? as you dont use wiener-hopf.
My guess is that as the filter is non-causal, all restrictions are lifted and you set
your impulse-respons as you want. If we for example would like a 2-step-prediktor
we simply define h(n) = \delta(n-2) \Rightarrow H(z)=z^{-2} and use that
frequency-response when calculating the cross-spectrum that is needed to solve
the IIR-filter equation shown above.
Is this correct? Thanks for any input on this!