apparently, the continuous wavelet transform can be used to generate time-frequency plots that are in some ways superior to the short-time Fourier transform. See http://physics.ius.edu/~kyle/K/pdf/timefreq.pdf for an example. Specifically, the frequency resolution of low-frequency signals and time resolution of high-frequency signals can be achieved simultaneously.
It seems to me that this kind of function is not implemented in the Matlab Wavelet toolbox. The CWT plots I get from the toolbox require very different interpretation compared to the STFT. This "time-structure information" is undoubtedly often useful to complement Fourier methods, but I would imagine that many would be interested in old-fashioned time-frequency information too.
Why is such a desirable application of the CWT not in the toolbox?
yours,
Jussi
Hi,
you need to calculate scales into frequencies. Here is a code example:
scales = logspace(1, 3, scale);
S = cwt(x, scales, wName);
S = (abs(S) .^ 2);
F = scal2frq(scales, wName, 1 / fs);
If you have problems just ask ^^
Kai
Hi,
I know how to compute (pseudo)frequencies from scales, but however I choose to color the CWT magnitudes, I seem to get plots resembling this:
http://www.mathworks.com/access/helpdesk/help/toolbox/wavelet/plcwt2.gif
This is what I meant by a "time-structure" plot. Pure sinusoids appear as repetitive features in the plot. I'm asking if the toolbox also implements a "time-frequency" plot, using the CWT? In other words, it would look a lot like the output of spectrogram.m, but computed using the CWT (with the corresponding advantages).
thanks,
Jussi
"Jussi " <j...@iki.fi> wrote in message
news:hfll1f$2mp$1...@fred.mathworks.com...
If you use the code and do
imagesc(T, F, S)
you should get your TF-Plot (example:
http://dl.dropbox.com/u/1275974/TFD_BK_PR_GA.jpg)
So, I hope I understood your problem now!
Kai