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
Edge effects from atoms [ Reply ]
By: Dillion Tan on 2008-10-14 02:57

I would like to understand more about edge effects or boundary conditions in MPTK. After observing the time frequency representation of various decompositions (using multiscale Gabor dictionaries), it seems that there is some broadband noise associated with the start and end of each atom; is this the so-called edge effect? What is the cause of it? Is there additional energy in the reconstructed signal due to these effects? Is it possible to reduce or even eliminate these effects?

If the dictionary used has sufficiently long blocks, the effect is negligible, but when the dictionary is limited to short length blocks e.g. only up to 256 window length, 512 fftSize, very audible artifacts are formed from the overlaying of atom edges. This is important for me as I would like to model an audio signal with short length blocks so that manipulation of atoms can be more flexible in time.