I couldn’t find anything to the problem of finding similar sequences of two (mono) tracks and using this overlap to align them, i.e. to concatenate/append them.
Say I have two recordings of the same song from recording over the soundcard output. The first track contains seconds from 0s to 10s and the second track from say 9s to 19s. I don’t know the exact period of overlap. So I’m looking for a tool that finds the overlap and possibly also aligns the second track to the first such that the second track starts at the right time of the first when the overlapping sequence begins. Is there such a tool? How is this process called? I would expect that some kind of cross-correlation has to be used to find the right position for best overlap of the two tracks. This would also allow to find a best overlap if the overlapping sequences are not completely identical.
Thanks for any help or hints
Thanks Trebor for Your quick reply.
As far as I understand that only shows how to align two tracks manually. However, I’m looking for a tool that tells me at what time shift a part of the end of the first track matches the beginning of the second track best. I don’t know the actual length of this overlapping period, i.e. the actual time shift.
In e.g. 2D image analysis You can find the best fitting position, i.e. the translation of one image to another to achieve best fitting overlap by using a 2D cross-correlation (see e.g.: http://imagejdocu.tudor.lu/doku.php?id=plugin:analysis:feature_finder:start or http://rsbweb.nih.gov/ij/plugins/template-matching.html).
I would expect this is also possible with audio using a 1D cross-correlation. Is there such a function in audacity?
Not at present, though a couple of the developers have been looking at a similar idea: http://wiki.audacityteam.org/wiki/Proposal_Audio_Diff
Audacity is in “feature freeze” so there will be no new features until after Audacity 2.0 has been released.