Exam zone
Overview of FIR Wiener Filtering For the optimal FIR Wiener filter, the estimate of the takes the form: LX¡ 1 dˆ[n ]= w[k ]x [n ¡ k ]:
signal of inter est (SOI)
k =0
The corresponding estimation error is given via: LX¡ 1
e[n ]= d[n ] ¡
w[k ]x [n ¡ k ]: k =0
The orthogonality principle associated with this MMSE optimization problem is given by: E f e[n ]x ¤ [n ¡ l ]g =0 ; 0 · l · L ¡ 1: The Wiener–Hopf equationsfortheoptimalFIRWiener filterderivedfromthe orthogonality principle are given by: LX¡ 1
wopt [k ]R xx [l ¡ k ]= R dx [l ]; 0 · l · L ¡ 1 k =0
Reformulated into a matrix system the normal equations can be written as: R xw
opt
= r dx ;
where R x istheautocorrelationmatrixoftheobservationsand r dx isthecrosscorrelation vector between the SOI and the observations. If a unique solution for the optimal FIR filtering problem is to exist, r dx 2 Range(R x ), i.e., the matrix R x must be invertible and from the properties of covariance and correlation matrices, also has to be positive–definite: aTR ax > 0; a 2 R L : The corresponding minimum me an-squar e d err (MMSE) or for the optimal FIR filter is given via: T T ² 2min = R dd [0] ¡ w opt r dx = R dd [0] ¡ w opt R xw
opt
= R dd [0] ¡ r Tdx R ¡x 1 r dx :
In the context of linear transversal filtering seen in digital communication applications, the linear FIR–MMSE filter is sometime referred to as the finite tap e qualizer . Note that eventhough the FIR Wiener filter is suboptimal compared to the non-causal and causal IIR filters, it is more practical for real-time applications such as equalization, echo-cancellation, and interference mitigation. Furthermore there are no stability problems associated with the FIR–MMSE filtering that one encounters with IIR filtering.
1