Subject: Re: Robust method of fundamental frequency estimation. From: Geoffroy Peeters <Geoffroy.Peeters@xxxxxxxx> Date: Thu, 1 Feb 2007 16:19:10 +0100 List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>This is a multi-part message in MIME format. --------------010405070508000503080301 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: quoted-printable X-MIME-Autoconverted: from 8bit to quoted-printable by drizzle.cc.mcgill.ca id l11FIwwY027807 Dear Roisin Loughran, We did some works on that topic recently, the goal was to estimate the pitch note rather than the exact pitch=20 frequency but across a wide range of instruments and octaves. The best method was a simple combination (multiplication) of 1) an Auto-correlation of a DFT with 2) a Frequency-mapped cepstrum (cepstrum lags expressed as inverse=20 frequencies). You can find details about our algorithm in the last ICASSP proceedings Peeters Music Pitch Representation by Periodicity measures based on=20 Combined Temporal and Spectral Representations IEEE ICASSP 2006 Toulouse (France) Best regards Geoffroy Peeters Ircam - R&D tel: +33/1/44.78.14.22 email: peeters@xxxxxxxx Roisin Loughran a =E9crit : > Dear list, > > I was wondering if any of you know the most robust way to calculate=20 > the fundamental frequency of a note across the range of a variety of=20 > instruments? > > I'm currently working on a matlab program and have tried using the=20 > auto-correlation method and the cepstrum method but have found that=20 > these both have difficulty in calculating f0 of timbre-rich tones such=20 > as those from a piano - particularly in the lower pitch ranges. Does=20 > anyone know of a method that is more reliable in these regions or is=20 > it necessary that I investigate such complex tones by a different=20 > means? From examining a number of the FFTs from these signals it is=20 > tempting to just pick the first strongest partial - the complex=20 > overtones just seem to confuse the more complicated algorithms, but I=20 > realise that this is hardly a reliable approach. > > Any suggestion would be greatly appreciated, > Thanks in advance, > > Roisin Loughran > > -----------------------------------------------------------------------= - > New Yahoo! Mail is the ultimate force in competitive emailing. Find=20 > out more at the Yahoo! Mail Championships=20 > <http://uk.rd.yahoo.com/mail/uk/taglines/default/championships/games/*h= ttp://uk.rd.yahoo.com/evt=3D44106/*http://mail.yahoo.net/uk/>.=20 > Plus: play games and win prizes. <http://www.ircam.fr> --------------010405070508000503080301 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit <!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"> <html> <head> <meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type"> </head> <body bgcolor="#ffffff" text="#000000"> <font face="Verdana">Dear Roisin Loughran,<br> <br> We did some works on that topic recently, <br> the goal was to estimate the pitch note </font><font face="Verdana">rather than the exact pitch </font><font face="Verdana">frequency but</font><font face="Verdana"> </font><font face="Verdana">across a wide range of instruments and octaves.<br> The best method was a simple combination (multiplication) of <br> 1) an Auto-correlation of a DFT with <br> 2) a Frequency-mapped cepstrum (cepstrum lags expressed as inverse frequencies).<br> You can find details about our algorithm in the last ICASSP proceedings<br> </font><span class="author">Peeters</span> <span class="title">Music Pitch Representation by Periodicity measures based on Combined Temporal and Spectral Representations</span> <br> <span class="publication">IEEE ICASSP 2006</span> <span class="location">Toulouse (France)</span><span class="date"></span><br> <font face="Verdana"><br> Best regards<br> Geoffroy Peeters<br> <font color="#555555" size="-1"> Ircam - R&D<br> tel: +33/1/44.78.14.22<br> email: <a class="moz-txt-link-abbreviated" href="mailto:peeters@xxxxxxxx">peeters@xxxxxxxx</a><br> <br> </font></font>Roisin Loughran a écrit : <blockquote cite="mid20070131171152.74585.qmail@xxxxxxxx" type="cite"> <style type="text/css"><!-- DIV {margin:0px;} --></style> <div style="font-family: times new roman,new york,times,serif; font-size: 12pt;"> <div>Dear list,<br> <br> I was wondering if any of you know the most robust way to calculate the fundamental frequency of a note across the range of a variety of instruments?<br> <br> I'm currently working on a matlab program and have tried using the auto-correlation method and the cepstrum method but have found that these both have difficulty in calculating f0 of timbre-rich tones such as those from a piano - particularly in the lower pitch ranges. Does anyone know of a method that is more reliable in these regions or is it necessary that I investigate such complex tones by a different means? >From examining a number of the FFTs from these signals it is tempting to just pick the first strongest partial - the complex overtones just seem to confuse the more complicated algorithms, but I realise that this is hardly a reliable approach.<br> <br> Any suggestion would be greatly appreciated,<br> Thanks in advance,<br> <br> Roisin Loughran<br> </div> </div> <br> <hr size="1"> New Yahoo! Mail is the ultimate force in competitive emailing. Find out more at the <a href="http://uk.rd.yahoo.com/mail/uk/taglines/default/championships/games/*http://uk.rd.yahoo.com/evt=44106/*http://mail.yahoo.net/uk/">Yahoo! Mail Championships</a>. Plus: play games and win prizes.</blockquote> <br> <div class="moz-signature"><a href="http://www.ircam.fr" hreflang="fr"> </a><br> </div> </body> </html> --------------010405070508000503080301--