onset detection (Andrew McGuiness )


Subject: onset detection
From:    Andrew McGuiness  <andymcguiness@xxxxxxxx>
Date:    Tue, 25 Dec 2007 22:57:37 +1100
List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>

--0-1673517764-1198583857=:4189 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: 8bit Hello - I have written a script in Praat for onset detection, and I'd like to get peoiple's opinions on my underlying assumptions. First, the script looks for negative slope in the energy curve of the sound, with energy calculate by averaging 100 millisecond blocks, to eliminate very local fluctuations. If the sound file is recorded clean, with no spill, this works well to identify the general area of the onset. Having found the general area of the onset, the script looks for the steepest point in the energy curve of the upper partials of the sound, to identify the precise point of onset. I don't take this to be the P-center but hope it might be in a consistent relation of absolute distance (in time) from the P-center - any comments? If looking for the steepest part of the energy curve is taken to be consistently related to the P-center, is it reasonable to expect the time between steepest-point and P-center to be the same for eg. a snare drum hit and, say, a violin attack? The steepest point is found in an energy curve obtained by average at the 0.5 millisecond level - does this sound right? I mean, is it feasible to expect accuracy at such a fine resolution? Hoping for some comments - Andy McGuiness Music Dept Open University, UK --------------------------------- Make the switch to the world's best email. Get the new Yahoo!7 Mail now. --0-1673517764-1198583857=:4189 Content-Type: text/html; charset=iso-8859-1 Content-Transfer-Encoding: 8bit <div>Hello - I have written a script in Praat for onset detection, and I'd like to get peoiple's opinions on my underlying assumptions.</div> <div>&nbsp;</div> <div>First, the script looks for negative slope in the energy curve of the sound, with energy calculate by averaging 100 millisecond blocks, to eliminate very local fluctuations. If the sound file is recorded clean, with no spill, this works well to identify the general area of the onset.</div> <div>&nbsp;</div> <div>Having found the general area of the onset, the script looks for the steepest point in the energy curve of the upper partials of the sound, to identify the precise point of onset. I don't take this to be the P-center but hope it might be in a consistent relation of absolute distance (in time) from the P-center - any comments?</div> <div>&nbsp;</div> <div>If looking for the steepest part of the energy curve is taken to be consistently related to the P-center, is it reasonable to expect the time between steepest-point and P-center to be the same for eg. a snare drum hit&nbsp;and, say, a violin attack?</div> <div>&nbsp;</div> <div>The steepest point is found in an energy curve obtained by average at the 0.5 millisecond level - does this sound right? I mean, is it feasible to expect accuracy at such a fine resolution?</div> <div>&nbsp;</div> <div>Hoping for some comments - </div> <div>&nbsp;</div> <div>Andy McGuiness</div> <div>Music Dept</div> <div>Open University, UK</div> <div>&nbsp;</div> <div>&nbsp;</div><p>&#32; <hr size=1> Make the switch to the world's best email. <a href="http://au.rd.yahoo.com/mail/taglines/default_all/mail/spankey/*http://au.yahoo.com/worldsbestmail/spankey/" target=_blank>Get the new Yahoo!7 Mail now</a>. --0-1673517764-1198583857=:4189--


This message came from the mail archive
http://www.auditory.org/postings/2007/
maintained by:
DAn Ellis <dpwe@ee.columbia.edu>
Electrical Engineering Dept., Columbia University