MATLAB fileexchange should be an appropriate place to post this question or find the code

I have already asked the question and I have not had any response in matlab file exchange

Better chance to have a constructive answer if you describe a little bit more about your request. For example, post related papers, applications, goals ?

thank you for your reply, here are the methods detailed in the articles of base of Boashash.

I want to extract the Doppler acoustic signal and I work on Matlab these methods are just a passage for information retrieval and use data

thank you.

I believe that methods for projecting the line of maxima of time-frequency representations on the time-frequency plane work better for instantaneous frequency estimation. There is a Matlab Toolbox for time-frequency representations on the web page of Patrick Flandrin

thank you for your reply, I need a technique for estimating the instantaneous frequency real time. like Phase Differencing .....

  the time frequency distibution is very costly in computation time.

Hello. You could determine the spectrogram of the signal, then apply a ridge searching algorithm for determining the temporal evolution of the frequency affected by Doppler shift. The resulting sequence of data could be associated with the theoretical Doppler-shifted frequency [ fD(t) = f0 * c / ( c+v*cos(beta) ) ] , where v(t)=dx(t)/dt is the instantaneous velocity over a rectilinear path.

If computational load is an important concern, you may apply bandbass signal and apply the spectrogram just around the frequency of interest. We also worked in Doppler compensation for acoustic signals. For more detais, see these papers:

http://www.cimec.org.ar/ojs/index.php/mc/article/viewFile/2711/2653

http://www.cimec.org.ar/ojs/index.php/mc/article/viewFile/3978/3895

I'm not shure that the techniques based on adaptive filtering (LMS, RLS) are faster than the techniques based on time-frequency representations. See in attachement a Matlab code written by Philippe Regalia as demo for a instantaneous frequency estimation method based on adaptive notch filter.

If you neen adaptive instantaneous frequency estimation, you should check Hilbert Huang Transform http://en.wikipedia.org/wiki/Hilbert%E2%80%93Huang_transform. It is generally used for ocean wave analysis by the original author Norden Huang; however, its application is not limited in 1-D signal analysis.

I agree with Anh. Hilbert Huan transform (HHT) is quite useful for nonlinear and nonstarionary signals. However, its computational load is high and depends on the signal length. Maybe a good idea is applying the HHT to the signal filtered around the frequencies of interest fmin=f0*c/(c+v0) and fmax=f0*c/(c-v0).

The algorithms you need are reported in the excellent report of P.J. Kootsookos that you can find here:

http://espace.library.uq.edu.au/eserv.php?pid=UQ:10626&dsID=comparison-t.pdf

To learn filtering techniques, you should use the book written by Quinn and Hannan:

http://www.amazon.com/Estimation-Frequency-Statistical-Probabilistic-Mathematics/dp/0521804469

If you are looking for codes to compare with yours, you will find very good codes on the Peter K's website:

https://sites.google.com/site/kootsoop/frequency-algorithms

Are you looking for a mono or multi-component signals?

If your signal is monocomponent, I doubt that the computation of the spectrogram will give you insight.

Regards,

Nicolas

@ Anh Cat Le Ngo and Fernando Marengo Rodriguez thank you very much for your answers, the idea of ​​estimating the instantaneous frequency with adaptive filter is very helpful because my signal is very noisy.

@Fernando Marengo Rodriguez: I'm using a band pass to filter around the frequencies of interest fmin=f0*c/(c+v0) and fmax=f0*c/(c-v0).

I think that the MLS and RLS are faster than time-frequency representations. I didn't test it but I Think that I read somewhere that.

Regards,

Hakima

Ok Timlelt Hakima, I'm glad to have been helpful. I haven't used the RLS and MLS tools, but for sure time frequency representations are precise, most of all the HHT, which is also useful for noisy signals as yours. If you happen to need more help, just ask. And if you get some results, I will be very glad to know about them.

Regards,

Fernando

PS: The post of Nicolas Cusseau is indeed very interesting.

@Nicolas Cusseau, thak you for your answer,

Sorry I forgot du menetion that my signal is monocomponent.

the book is very interesting.

as soon as I have the results I will share with you.

Thank you

good luck

Hakima

Dear Hakima,

what kind of noise is present in your signals. If you deal with impulse noise or mixed impulse and Gaussian noise you should maybe try robust time-frequency distributions. Also, there are some hardware solution for very fast calculation of TF distributions. Related papers can be found on www.tfsa.ac.me.

If you need further help, please do not hesitate to contact me.

Regards,

Nikola

Dear Timlelt,

We need some figures to help you. Otherwise this question will become a troll.

What is the sampling rate of your signal?

Is your system an embedded system (DSP) or a computer? That changes the definition of real-time...

What frequency resolution do you need to achieve?

Regards,

Nicolas

Dear Nikola Zaric

Thank you for your help, you asked very objective questions, I think that my signal present a Guassian Noise. I want to know if there is a technique to detect the nature of the noise in a signal??? I use an acoustic signal so the frequency band is between (20Hz - 20Khz).

in this article, a technique for estimation of intantaneous frequency with any Time frequency distibution and Viterbi Algorithm in low SNR

http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=1201633&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D1201633

so what's your opinion about this méthode !!!

I'm sorry, I ask a lot of questions, I am a beginner in this field.

Regards,

Hakima