Method for prediction of a channel coefficient

1. A method for prediction of a channel coefficient in a wireless communication device adapted to communicate over a propagation path in a mobile radio channel with multipath propagation, comprising:

using a canonical random variable analysis on a signal having a known sequence to predict the channel coefficient in the wireless communication device, wherein using the canonical random variable analysis comprises:

producing a past matrix P and a future matrix F from a known time series x k of estimated values of the channel coefficient, wherein x k is associated with the signal having the known sequence;

solving the equation system:

F=B·M+E

M=A·P

 in order to obtain estimates  and {circumflex over (B)}, for matrices A and B, wherein F is the future matrix, P is the past matrix, M is a memory matrix representing an intermediate set of variables and a finite number of past states of the respective channel coefficient, E is an error matrix that contains prediction error variables and represents an error in a linear prediction of the future matrix;

producing a past vector p from a new time series of known estimated values of the channel coefficient; and

calculating a prediction vector {circumflex over (f)} for estimated values of the channel coefficient from the equations:

{circumflex over (f)}={circumflex over (B)}·m

m=·p,

wherein m is an optimum dimension portion of the memory matrix identified by ascertaining canonical correlations.

2. The method of claim 1 , wherein the method is used in an estimator for channel coefficients in a base station or in a mobile station in a mobile radio system.

3. The method of claim 1 , further comprising estimating the channel coefficient using the prediction vector {circumflex over (f)}.

4. A method for prediction of a channel coefficient in a wireless communication device adapted to communicate over a propagation path in a mobile radio channel with multipath propagation, comprising:

using a canonical random variable analysis on a signal having a known sequence to predict the channel coefficient for the wireless communication device, wherein using the canonical random variable analysis to predict the channel coefficient comprises:

evaluating the signal having the known sequence to identify a sequence of values x k associated therewith, wherein k comprises a discrete integer time variable;

producing a past matrix P and a future matrix F from the identified sequence of values x k , and

using the past matrix P and the future matrix F to generate a prediction vector {circumflex over (f)}, wherein a portion of the prediction vector {circumflex over (f)} represents the predicted channel coefficient.

5. The method of claim 4 , wherein the sequence of values x k comprise estimated values of the channel coefficient, and wherein the sequence of values x k extends over a discrete time sequence of k=1 to N+n, wherein the sequence k=1 to N represents values of the channel coefficient in the past, and wherein the sequence k=N to n represents values of the channel coefficient in the future based on the signal having a known sequence.

6. The method of claim 4 , wherein using the past matrix P and the future matrix F to generate the prediction vector {circumflex over (f)} comprises:

calculating a cross-covariance matrix Σ PF from the past matrix P and the future matrix F;

performing a singular value decomposition of the cross-covariance matrix Σ PF to determine square matrices J and L;

using the square matrices J and L to calculate matrices  and {circumflex over (B)} representing estimations of matrices A and B, respectively;

using the past matrix P and the estimated matrix  to calculate a memory matrix M; and

using the matrices  and {circumflex over (B)} along with a memory vector m associated with the memory matrix M and a past vector associated with the past matrix P to calculate the prediction vector {circumflex over (f)}.

7. A method of predicting a channel coefficient for estimating a channel impulse response in a wireless communication device, comprising:

receiving a signal having a known sequence at the wireless communication device; and

evaluating the signal using canonical random variable analysis to determine the channel coefficient, wherein using the canonical random variable analysis to evaluate the signal and determine the channel coefficient comprises:

evaluating the signal having the known sequence to identify a sequence of values x k associated therewith, wherein k comprises a discrete integer time variable;

producing a past matrix P and a future matrix F from the identified sequence of values x k ; and

using the past matrix P and the future matrix F to generate a prediction vector {circumflex over (f)}, wherein a portion of the prediction vector {circumflex over (f)} represents the predicted channel coefficient.

8. The method of claim 7 , wherein using the past matrix P and the future matrix F to generate the prediction vector {circumflex over (f)} comprises:

calculating a cross-covariance matrix Σ PF from the past matrix P and the future matrix F;

performing a singular value decomposition of the cross-covariance matrix Σ PF to determine square matrices J and L;

using the square matrices J and L to calculate matrices  and {circumflex over (B)} representing estimations of matrices A and B, respectively;

using the past matrix P and the estimated matrix  to calculate a memory matrix M; and

using the matrices  and {circumflex over (B)} along with a memory vector m associated with the memory matrix M and a past vector associated with the past matrix P to calculate the prediction vector {circumflex over (f)}.