Good morning experties,

I'm using the generalized Levinson-Durbin algrithm to solve the system
R*hinv=q. Where q(l)=h(N-l), N is the length of the filter to be inverted
and hinv is its inverse. R is a symmetric Toeplitz matrix.
I used the MATLAB code from Bob Cain in 
http://www.musicdsp.org/showone.php?id=188  which work very well.
But I can't understand why it works!
Could you please help me?

Thank you in advance
Stef