Home My Page Projects MPTK: The Matching Pursuit ToolKit
Summary Activity Forums Tracker Lists Docs News SCM Files

Forum: help

Monitor Forum | Start New Thread Start New Thread
RE: retrieving the coefficients [ Reply ]
By: RĂ©mi Gribonval on 2011-03-09 09:13
[forum:106565]
Dear Iswarya

As far as I remember, there is unfortunately no shortcut to directly retrieve the inner product between the atom and the residual that you are looking for. This topic has been discussed by Bob Sturm here:
http://media.aau.dk/CRISSP/mt/mt-search.cgi?IncludeBlogs=1&tag=MPTK%20unit%20norm%20atom%20energy%20amplitude&limit=20

I can see two ways to proceed

a) in practice, the 'amplitude' of the atoms might be enough, and you may just retrieve it with the book.atom(x).param.amp field as described here
https://gforge.inria.fr/forum/message.php?msg_id=106426&group_id=36

b) if you want to be more precise, you can retrieve the energy decay after mpdecomp as

[book,residual,decay] = mpdecomp(signal,sampleRate,dict,numIter)

decay is a vector containing the energy of the residual after each iteration, and sqrt(decay(1:(end-1))-decay(2:end)) gives you the magnitude of the inner products you are looking for.

I hope this helps,

All the best,
Remi.




retrieving the coefficients [ Reply ]
By: Iswarya Sudhakar on 2011-03-05 06:29
[forum:106545]

project_paper.pdf (73) downloads
Respected sir,
I need to retrieve the atom and coefficient values from the output and feed it to the audio,video atom fusion module for the blind audio visual source separation. How do I retrieve the values to correlate it with the video atoms. By coefficient i mean the cross product of the atom and the signal.