It's a hard problem. A recent article in Science News shows some new approaches:
http://www.sciencenews.org/articles/20070421/bob8.asp Henry Rich > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of Raul Miller > Sent: Friday, May 18, 2007 7:14 PM > To: General forum > Subject: [Jgeneral] spectral analysis? > > Fourier Transforms seem rather poor for spectral analysis. > > [If the problem is "given this recording of middle C in five > different octaves, determine which octaves are being used > and their relative power", fourier transforms can be a perfect > fit. If the problem is "identify the sequence of notes being > played", fourier transforms are much less adequate. And if > the sound is more complex it gets worse.] > > Can anyone recommend a better basis or approach for > such analysis? > > Thanks, > > -- > Raul > ---------------------------------------------------------------------- > For information about J forums see > http://www.jsoftware.com/forums.htm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
