Forums

implementing MUSIC/Pisarenko algorithm in Matlab

Started by vorad June 7, 2010
Hi guys,

I am trying to implement the MUSIC algorithm and the more specialized case
of Pisarenko and I am not sure if my approach is a good one. I am following
the algorithm from the book "Spectral analysis of signals" by Moses and
Stoica and the way I figured it out that it would work is by:

- having a method that gets as input the signal and return the frequency
estimates
- compute the covariance matrix ( I am wondering whether corrmtx from
Matlab is enough )
- compute the eigendecomposition for the given covariance matrix
- determine the frequency estimates based on the pseudospectrum relation 

Is there any real implementation in Matlab or in any other language that
shows a good example of MUSIC ?

PS: Sorry if I didn't post on the right section, I am new around here.
Thanks!


On 7 Jun, 22:40, "vorad" <vorad.1100@n_o_s_p_a_m.gmail.com> wrote:
> Hi guys, > > I am trying to implement the MUSIC algorithm and the more specialized case > of Pisarenko and I am not sure if my approach is a good one. I am following > the algorithm from the book "Spectral analysis of signals" by Moses and > Stoica and the way I figured it out that it would work is by: > > - having a method that gets as input the signal and return the frequency > estimates
Wrong. That's what the whole MUSIC algorithm does.
> - compute the covariance matrix ( I am wondering whether corrmtx from > Matlab is enough )
Roll your own. It's trivial.
> - compute the eigendecomposition for the given covariance matrix
Sure.
> - determine the frequency estimates based on the pseudospectrum relation
Almost. You only need to minimize the denominator of the 'pseudo spectrum' relations.
> Is there any real implementation in Matlab or in any other language that > shows a good example of MUSIC ?
What do you mean by 'real implementation'? It's a trivial matter to implemnt MUSIC in matlab, as all the required linear algebra is easily available. Rune