Subject: Re: Rhythm perception From: Dan Ellis <dpwe(at)EE.COLUMBIA.EDU> Date: Tue, 15 Nov 2005 19:25:54 -0500Dear List - Following our earlier discussion, I contacted Tristan Jehan to see if he had any examples of maracatu to listen to. He does, and they are interesting - see his message below. Also on his web site are some nice other examples including a James Brown excerpt where there is no instrument playing on the downbeat. (The example is to illustrate how Tristan's learning system can still correctly locate the downbeat, as indicated by the "triangle" sound mixed in with the example): http://web.media.mit.edu/~tristan/Blog/Results/one.JBgotFeeling2.mp3 His parent page with other examples: http://web.media.mit.edu/~tristan/Blog/Downbeat.html Many thanks to Tristan. Happy listening! DAn. Date: Tue, 15 Nov 2005 14:46:22 -0500 To: Dan Ellis <dpwe(at)ee.columbia.edu> From: Tristan Jehan <tristan(at)media.mit.edu> Subject: Re: downbeat [...] Here are some Maracatu examples. I also heard music from Africa, China, Japan and other parts of Latin America that are also very confusing to me, Western musician, although I play in a Brazilian band. You might want to listen to these in that order and try to find the downbeat without training. Here is the basic rhythmic pattern that makes Maracatu: http://web.media.mit.edu/~tristan/Tmp/maracatuBasicPattern.mp3 On the very last drum sound you should be saying ONE while it cuts. Here is a section of that example now actually starting on the downbeat: http://web.media.mit.edu/~tristan/Tmp/maracatu1StartingOnOne.mp3 Note that only the downbeat aligns with a loud bass drum sound. The 2 other bass drum sounds are syncopated late. Here is another variation around that same pattern, yet starting on the downbeat: http://web.media.mit.edu/~tristan/Tmp/maracatu2StartingOnOne.mp3 Might sound even more confusing. Another test example, this time with singing. Without experience, I believe it's easier to catch the downbeat by relying on the voice: This is the kind of music that was part of my training set: http://web.media.mit.edu/~tristan/Tmp/maracatuPhD1.mp3 I fade in the audio here in order to avoid any bias based on the start phase. Finally the result of that same excerpt with the "Maracatu model" after a few seconds: http://web.media.mit.edu/~tristan/Tmp/tmp.maracatuPhD1.mp3 Here is full Maracatu song by the band Maracatu Estrela Brilhante for reference: http://web.media.mit.edu/~tristan/Tmp/maracatuEstrelaBrilhante.mp3 For the experiment as presented in my paper, I used 6 of those songs (each having a different tempo). They share some common rhythmic patterns though, but you will notice that there's quite a few variations in that one already. Hope this helps. Tristan