burg - AR-Model by the method of Burg
x specifies a stationary random process. With the function burg you calculate the characterisitc values of an AR filter. It is assumed, that a white noise signal, filtered with these characteristics, has nearly the same statistical properties as x.
Therefore, the transfer function specified by the AR filter represents nearly the spectrum of the random process x. With [H,W]=mtlb_freqz(gain,A,N); [N is for example 256] you get the transfer function in the Fourier area.
x=rand(1,4000,'normal'); // Generate white gaussian noise hz=iir(8,'bp','cheb2',[0.1,0.2],[1E-4,1E-4]); y=flts(x,hz); //bandpass filtering [A,Gain]=burg(y,11); //Burg estimation [H,W]=mtlb_freqz(Gain,A,512); plot2d(W,abs(H));