Wavelet Packets to Dominant Frequency
in [Matlab]
|
Prev: engPutVariable doesn't put any variable in workspace but program works?!?!
Next: Errors while compiling MATLAB GUI m-file
From: Jonathan on 8 Jul 2010 13:43 Greetings all, I have been learning about wavelets for the past couple of weeks. I have the wavelet toolbox and have some trouble extracting certain information. Right now, my primary concern is how to extract a dominant frequency from a given transform. For example, if I have the coefficients of an N-level discrete wavelet transform (DWT), then I can easily see which level carries the most energy. I can even use the scal2freq function to get a pseudo-frequency that corresponds to that scale. However, if I want to think of a scale as a band of frequencies, how do I determine how far on each side of the pseudo-frequency to extend the bounds of this band? This leads to thinking about the wavelet packet transform (WPT). I think this gives information about different scales at different times (just as the DWT) and gives this information for different frequency bands (determined by the position of a node in the packet tree). So, for this case, how can I find the dominant frequency of a given 1D signal at a certain time using the WPT coefficients? That is, given a particular node in the packet tree, what is the corresponding frequency band? I should note that I already understand the difference between the natural ordering of nodes and the frequency ordering of such. Most of all, how can I do these calculations from the command line? Thank you for your help. ~Bubba
From: Wayne King on 8 Jul 2010 13:52
"Jonathan " <bubbathekid(a)gmail.com> wrote in message <i152nq$9sh$1(a)fred.mathworks.com>... > Greetings all, > > I have been learning about wavelets for the past couple of weeks. I have the wavelet toolbox and have some trouble extracting certain information. Right now, my primary concern is how to extract a dominant frequency from a given transform. For example, if I have the coefficients of an N-level discrete wavelet transform (DWT), then I can easily see which level carries the most energy. I can even use the scal2freq function to get a pseudo-frequency that corresponds to that scale. However, if I want to think of a scale as a band of frequencies, how do I determine how far on each side of the pseudo-frequency to extend the bounds of this band? > > This leads to thinking about the wavelet packet transform (WPT). I think this gives information about different scales at different times (just as the DWT) and gives this information for different frequency bands (determined by the position of a node in the packet tree). So, for this case, how can I find the dominant frequency of a given 1D signal at a certain time using the WPT coefficients? That is, given a particular node in the packet tree, what is the corresponding frequency band? I should note that I already understand the difference between the natural ordering of nodes and the frequency ordering of such. > > Most of all, how can I do these calculations from the command line? Thank you for your help. > > ~Bubba Hi, do you have the R2010b prerelease? If so, my answer is going to be a lot shorter :) Wayne |