scott@heim.uucp (Scotty) (08/04/89)
Well, it's time for "one of those questions" again... Is anyone reading this familiar with work done in the Markovian properties of music? In other words, has anyone done an analysis of the probability of any particular note (tone/duration) givin any N previous notes? This kind of analysis might be done for any one piece or composer for example, and it is *possible* that playing notes generated from such a table may produce a sequence which sounds like the piece or composer... I'm sure this is well trodden ground.. could somebody point me in the right direction? Thank you for your time - -- Scott Watson - "Inane little message goes here" uucp: {rutgers,ames}!elroy!grian!heim!scott Internet: scott@heim.UUCP
jakob@nada.kth.se (Jakob Cederlund) (08/04/89)
I've dunnit. Using a simple BASIC program I generated a new tune given an existing one. I only used monophonic tunes, and the result wasn't very interesting; it just sounded like a small variation of the original tune. I got some rather simple correlations: the more indices in the matrix (=number of preceding notes), the closer the result was to the original tune. Also, the longer the tune, the more alike were the result. Any others who've tried Markovian music-generation? /Jakob Cederlund jakob@nada.kth.se Computer Science Dept. Royal Institute of Technology, Stockholm, Sweden
eiverson@nmsu.edu (Eric Iverson) (08/05/89)
I have an article somewhere that wasgested sug to me by Charles Dodge in which some researchers used a 1/f1 fractal waveform to generate music. When they played it for people the general reaction was that it was music, albeit not exciting music. Tthen went on to analyze several hours of commercial radio transmission and found that it too had this 1/f1 relationship. Anybody know more about this and how it might be coupled with a Markovian approach?
scott@heim.uucp (Scotty) (08/07/89)
In article <1389@draken.nada.kth.se> jakob@nada.kth.se (Jakob Cederlund) writes: >I've dunnit. Using a simple BASIC program I generated a new tune given >an existing one. I only used monophonic tunes, and the result wasn't >very interesting; it just sounded like a small variation of the >original tune. > >I got some rather simple correlations: the more indices in the matrix >(=number of preceding notes), the closer the result was to the >original tune. Also, the longer the tune, the more alike were the >result. What happens if you feed it the multiple works of a composor? Hopefully, the end result would be something where small segments where more or less played verbatim from the source material, but, would make (more or less) reasonable transitions between them? In your BASIC attempt (no pun intended), I suppose you used the note and duration as the key features your analysis? Is this correct? Could you post this program to the newsgroup? -- Scott Watson - "Inane little message goes here" uucp: {rutgers,ames}!elroy!grian!heim!scott Internet: scott@heim.UUCP
chip@soi.UUCP (Chip Morris) (08/07/89)
scott@heim.uucp (Scotty) writes: >In other words, has anyone done an analysis of >the probability of any particular note (tone/duration) givin any N >previous notes? There is a commercial program for the Macintosh, I believe it is "Jam Session", that does just this. You can vary the depth of the analysis, but it does allow you to write melodies in someones "style". -- Chip Morris, Senior Engineer US Mail: Software Options, Inc., 22 Hilliard St., Cambridge MA 02138 Internet: chip%soi@harvard.harvard.edu UUCP: ...!harvard!soi!chip Phone: (617) 497-5054
swan@jolnet.ORPK.IL.US (Joel Swan) (08/08/89)
In article <497@heim.uucp> scott@heim.UUCP (Scotty) writes: >In other words, has anyone done an analysis of the probability of any >particular note (tone/duration) givin any N previous notes? >Scott Watson - "Inane little message goes here" > uucp: {rutgers,ames}!elroy!grian!heim!scott >Internet: scott@heim.UUCP There is a piece of software for the Amiga called "M." I've heard wonderful things about it. I don't know if it can also use the Markovian method, but I suspect that may be one of its many options. It allows one to do analysis and composition based on a number of variables. I'm not familiar enough to answer any specific questions about it, but someone in comp.sys.amiga may. So far M has had pretty good reviews. -- - Joel E. Swan [ swan@jolnet.UUCP <> PLINK ID: Amiga*joel ] [ "Amigas.... for the rest of us." <> CI$ : 74746,3240 ] [ "...peace with God through our Lord Jesus Christ." Romans 5:8 ]
mketch@pawl.rpi.edu (Michael D. Ketchen) (08/08/89)
In article <1273@jolnet.ORPK.IL.US> swan@jolnet.UUCP (Joel Swan) writes: >There is a piece of software for the Amiga called "M." I've heard >wonderful things about it. I don't know if it can also use the >Markovian method, but I suspect that may be one of its many options. >It allows one to do analysis and composition based on a number of variables. >I'm not familiar enough to answer any specific questions about it, but >someone in comp.sys.amiga may. So far M has had pretty good reviews. M is not really a music analysis program so much as it is a composition aid. It uses random processes to alter your musical data according to several variables (note density, rhythm, amount of randomness, etc.) that the user can set. You can use it to add a "human" feel to your music, to experiment with different musical ideas, or just to play around. It's a very powerful composition tool. However, I don't think it has any features for music analysis. (I used it last semester in a class with the guy who designed the program, and both from his descriptions of the program and my use of it, I don't see how one could use it for analysis.) - Mike -- |XXX| __/\__ |XXX|-----------------------+--------------------********========= |XXX| \ / |XXX| Michael D. Ketchen | This space for ********========= |XXX| /____\ |XXX| mketch@pawl.rpi.edu | rent... ================= |XXX| ][ |XXX|-----------------------+--------------------=================
jboggs@inco.UUCP (John Boggs) (08/11/89)
In article <1273@jolnet.ORPK.IL.US> swan@jolnet.UUCP (Joel Swan) writes: >There is a piece of software for the Amiga called "M." I've heard >wonderful things about it. I don't know if it can also use the >Markovian method, but I suspect that may be one of its many options. >It allows one to do analysis and composition based on a number of variables. >I'm not familiar enough to answer any specific questions about it, but >someone in comp.sys.amiga may. So far M has had pretty good reviews. > Voyetra Technologies, distributers of the Sequencer Plus series of IBM PC based software sequencers, also sells a version of "M" for the PC called, strangely enough, "M-PC". I don't know any more about it than that. For an address Voyetra advertises in Keyboard Magazine every month. -- John Boggs McDonnell Douglas Electronic Systems Company McLean, Virginia, USA
aj-mberg@dasys1.UUCP (Micha Berger) (08/13/89)
Yes, this has been discussed by Martin Gardner when he had his column in SciAm, and by Douglas Hostadter when he got his hands on the space. -- Micha Berger "Always should [the child of] Adam have awe of G-d in secret and in public, admit the truth, and speak truth in his heart."
rnm@well.UUCP (Robert Marsanyi) (08/23/89)
re: Markovian analysis There's a nice PD package by Nick Didkovsky written in JForth for the Amiga that'll take an smus file(s), generate an analysis/synthesis of new material using a 1-degree Markov table and spit it out as smus, rendering it usable with many other Amiga packages. mail me for more info, anyone who's interested. rnm@well sysop, HMSL Forum (415) 928 8246 ps: this has been the subject of many experiments and pieces using HMSL, the Hierarchical Music Specification Language, on the Mac and Amiga.