Advice on how to optimize a MISO Wiener Filter (predictor)

Dear all,

I have a basic knowledge of signal processing, but unfortunately I'm not a specialist of this field. I have to build an adaptive MISO (multiple-input single-output) FIR Wiener filter, which receives three (or more) signals as inputs and produces an output according to the Wiener filter theory (for example referring to the book of Haykin, Adaptive Filter Theory). I need to use the Wiener filter as a predictor. I am coding a MATLAB routine to achieve this task. I have already started with some codes I found online at this page

http://www.mathworks.it/matlabcentra.../authors/95913

but I need some advice to build a more efficient (as you know, using the so-called Levinson Durbin recursion) and numerically stable MATLAB code . Indeed the stability of the code strongly depends on the nature of the inputs: If for example I pre-filter the signals to reduce the noise, ths weights coefficient of the filter go "crazy" and assume incredibly high values. It would be very grateful if you can indicate some MATLAB fucntions, already build for this purpose, somewhere in the web. It is very hard to code the Levinson-Durbin recursion and I do not know where I can find an algorithm for the MISO case (in which block-Toeplitz matrix are involved). In addition, I would like to ask whether there are other alternative approaches, other than the Wiener filter, which might suit better the problem I am tackling. Please let me know. Best regards,

Em