Forums

Project on Power Spectral density estimation.

Started by Tiken January 23, 2006
HI, I am currently doing a project on estimation of power spectral
density using DSP processort(Blackfin).I am bit confused abt how to
approach initially for the same.I am thinking to implement  algorithms
based on parametric estimation. What could be the best and convinient
type of the signal to be used for the same(Test vectors are to be used
actually) and how to generate such test vectors?

Tiken wrote:
> HI, I am currently doing a project on estimation of power spectral > density using DSP processort(Blackfin).I am bit confused abt how to > approach initially for the same.I am thinking to implement algorithms > based on parametric estimation. What could be the best and convinient > type of the signal to be used for the same(Test vectors are to be used > actually) and how to generate such test vectors? >
You could start with test vectors that match your parametric estimation scheme. For example, a simple spectral estimator (I forget the name) can be obtained as follows. For a given order of predictor, find the optimum linear predictor, h, for the signal of interest. An estimate of the spectrum is given by S(w) = 1/(H(w)H*(w)), where is H is the DFT of h. In this case, generate test vectors which are the output of analysis filter 1/h driven by white noise. This will provide the best case scenario in which your estimator will work. Sorry, but I don't have any references in front of me to provide more details of other estimators. Greg