edkelly wrote:
> Greetings,
>
> Can anyone tell me - or better still direct me to some code for converting
> PARtial CORrelation coefficients derived from the Schurr recursion into
> Line Spectrum Pairs?
You need to use what Hayes calls the "step-up" recursion to compute the
AR model coefficients a_k, 0<=k<=p. There is Matlab code for this on
Hayes' website:
http://users.ece.gatech.edu/~mhayes/stat_dsp/lev_m.html
Once you have the AR coefficients, you get a line spectrum by taking
their DFT and computing the point-wise complex inverse. You can control
the line spacing by zero-padding the coefficient vector.