zog@intacc.uucp (Scott Lepore) (02/01/90)
In article <11805@cs.yale.edu> beck-chris@CS.YALE.EDU (Chris Beck) writes: >I have a philosophical question for anyone outhere: > >I was wondering if anyone could recommend some books, or, preferably, had >some of their own thoughts, about possible representations for music that >can be understood by a computer (other than standard notation and MIDI). > >What constitutes a musical event? What would be a good strategyfor >representing not only the music itself but also all (if possible) aspects >of a given musical performance? > >- Chris This question is obviously very complicated. The first suggestion I can offer is to look at the quarterly, Computer Music Journal, published by MIT press. There is some fascinating work being done by its contributors on how to more completely define a sonic or musical event and how to place it in time. It's very speculative but it is based in practice. I am working on project which relates somewhat to this question: I am trying to design a powerful (ie. high level), yet flexible, music composition environment based on David Betz's XLISP for the IBM PC. In order to do this I have to do just what Chris is curious about - I have break *music* down into what I feel are its representative elements. The first major problem I have encountered in this process is how to structure time in a composition. I figure... Let's assume each musical event has a number of basic characteristics: fundamental pitch (for tonal events), timbre, volume (ie. loudness), duration and so on... Each event must also have an *absolute* starting time within the piece, that is, each *note* must start at an easily identifiable time relative to the beginning of a composition. This is where my problem begins. In traditional notation the composer and musician are required to count bars, consisting of x number of beats, in order to track their position in a piece. The counting of the bars is obviously relative to the beginning of the piece and the biggest thing differentiating one bar from another is its number. This system is fine and has worked for a very long time but one always has the feeling that this is not the most intuitive, and therefore easiest, way of working with music. I believe that humans perceive each sonic event as a distinct experience - each creating an individual feeling or sensation in the listener. In conjunction with their placement next to one another, a series of notes creates the experience of music. But each note/event is fundamentally perceived on its own. Traditional notations and composition techniques *anonimyse* individual notes and sonic events in a piece by placing too much emphasis on their start time and position. The musician makes mistakes while playing anonymous notes. After all when a *note* is not considered a unique event, but as something to be played at such and such a count, it is a simple matter to miscount and play it at the wrong time. Composition should be the process of intuitively identifying one sonic event from the next and arranging each accordingly. Each event should not lose its identity within the structure of a piece. In MIDI notation, a system very similar to traditional notation has been adopted. Bars and beats are preserved but each beat now consists of a quantum number of *ticks*. Now you can not only find where you are in a piece by counting the bar number but also by looking at the number of *ticks* elapsed. This is a good system to use from a computer's perspective because now there is a quantifiable time referent for each note. However the problem with this system remains the disproportional consideration of timing. In computer composition, because abstraction can be more creative (eg. unique visual note/event symbols - see Coda's Finale and some of the new software for the NeXT), a composition system can be developed to allow more easily recognisable and intuitive note/event *tags*. Imagine each note/event as single-hue Leggo block... If all music was monophonic, then the problem of absolute note placement in time (as the chief *anonimyzer* of sonic events) would be less serious. All events would follow one after another and would only require a duration, or note length, to place them in a piece. Silences between notes would be musical events similar to rests in traditional notation. Picture a stack of different coloured Leggo blocks... A musician playing such a piece would still have to count time, to play for the correct duration of a note (this seems unavoidable), and would therefore still suffer the *take it from the top!*, anonymous note/event timing. But composing and playing back such a *serial* or *string* composition on a computer would simply require an arbitrary choice of time unit (ie. CPU clock cycle, date clock second, MIDI pulses per quarter note, etc.) that relates to the computer and that could be assigned as the duration of a note/event. The composer could gauge the length of a note/event by listening and let the computer take care of figuring out the timing. A point and click, stretchable Leggo block... Each sonic event would be unique. Each note wouldn't need a *starting time* relative to the beginning of the piece. The apparent problem is that music is not this simple. It is not monophonic, it is polyphonic. Notes *occur* in *parallel* as well as *serial* streams. What to do? How can we deal with time in music and make it simpler and more intuitive? In a computer environment we have the freedom to (conceptually) overcome this problem. I invite your suggestions, comments, thoughts and snottiness (:-)). - Thanks for reading. | Let's have more ideas! - Be seeing you. | These are powerful tools for powerful times. - Scott. | New tools require new techniques for using them.
mgresham@artsnet.UUCP (Mark Gresham) (02/07/90)
In article <1990Feb1.074731.19127@intacc.uucp> zog@intacc.UUCP (Scott Lepore) writes: >In article <11805@cs.yale.edu> beck-chris@CS.YALE.EDU (Chris >Beck) writes: >>I have a philosophical question for anyone outhere: >> >>I was wondering if anyone could recommend some books, or, preferably, had >>some of their own thoughts, about possible representations for music that >>can be understood by a computer (other than standard notation and MIDI). >> >>What constitutes a musical event? What would be a good strategyfor >>representing not only the music itself but also all (if >possible) aspects >>of a given musical performance? >> >>- Chris > >This question is obviously very complicated. > >The first suggestion I can offer is to look at the quarterly, >Computer Music Journal, published by MIT press. There is some >fascinating work being done by its contributors on how to more >completely define a sonic or musical event and how to place it in >time. It's very speculative but it is based in practice. > >I am working on project which relates somewhat to this question: > >I am trying to design a powerful (ie. high level), yet flexible, >music composition environment based on David Betz's XLISP for the >IBM PC. [...] Each event must also >have an *absolute* starting time within the piece, that is, each >*note* must start at an easily identifiable time relative to the >beginning of a composition. This is where my problem begins. Yes, this is where your problem begins... Why assume 'note = event'? Ask yourself: What is a 'note'? Define it clearly; see what you exclude from that definition. Other possible events: phrase motive consonant (i.e. the voiced bilabial: 'b') timbral shift silence inhalation recaptulation variation etc.... What's an 'event'? >I believe that humans perceive each sonic event as a distinct >experience - each creating an individual feeling or sensation in >the listener. In conjunction with their placement next to one >another, a series of notes creates the experience of music. But >each note/event is fundamentally perceived on its own. One of the objectives in orchestration is to obscure the perception of certain individual events. Can you come up with some examples of what you think I might mean by that? >Traditional notations and composition techniques *anonimyse* >individual notes and sonic events in a piece by placing too much >emphasis on their start time and position. If one events on one level are obscured or anonymised, ask yourself what might be emphasised, individualised or clarified (perhaps on another level)? Can you likewise come up with some examples of what you think I might mean by this? >The musician makes mistakes while playing anonymous notes. After >all when a *note* is not considered a unique event, but as >something to be played at such and such a count, it is a simple >matter to miscount and play it at the wrong time. If an 'event' is exclusively perceived as an isolated, unique event, then there would be no such thing as 'wrong notes.' Is the sound of a cymbal struck and then partially dampened with, say, a finger one event or more? Is a single call of a nuthatch one event or a series of events? (Or pick your favorite songbird.) What is a sonic event? Is the sound of a piano string struck and then quickly loosened with a tuning hammer one event or more than one event? >Composition should be the process of intuitively identifying one >sonic event from the next and arranging each accordingly. Each >event should not lose its identity within the structure of a >piece. Well, this is a new one for the composing community to hear about. What constitutes 'identity' of an event? >If all music was monophonic, then the problem of absolute note >placement in time (as the chief *anonimyzer* of sonic events) >would be less serious. All events would follow one after another >and would only require a duration, or note length, to place them >in a piece. Silences between notes would be musical events >similar to rests in traditional notation. Picture a stack of >different coloured Leggo blocks... Picture instead a rope of intertwined multicolored threads, or a group of such ropes, each thread changing color along its length, each rope intertwining with others, sometimes thicker, sometimes thinner, intertwining indifferent ways, some dissappearing altogether at times and others appearing apparantly out of nowhere, waving up, down, and across in other directions. Follow this interplay of ropes, threads, colors, textures and thicknesses along its length with your eyes. This is far more like music (as a metaphor) than Leggo blocks. In fact, these images (continuous varied rope; stack of bricks) has been used for some time among singers as images for 'how to' and 'how not to' approach singing. Guess which is which. :-) >A musician playing such a piece would still have to count time, >to play for the correct duration of a note (this seems unavoidable), >and would therefore still suffer the *take it from the top!*, >anonymous note/event timing. When you 'make a mistake' in performing, you go on rather than stopping and starting all over because one 'note' was 1/3742 of a second too long or short. I think the issue of 'time' in actual performance practice is quite a bit more flexible than your giving it credit for. How long is long? What is the absolute duration of a 'fermata'? When a tempo is marked 'ca. 80 mm' how much faster or slower do you have to go to play the 'wrong' tempo? when the tempo is marked 'Andante'? What role does 'rubato' play in all of this? >But composing and playing back such a *serial* or *string* >composition on a computer would simply require an arbitrary >choice of time unit (ie. CPU clock cycle, date clock second, MIDI >pulses per quarter note, etc.) that relates to the computer and >that could be assigned as the duration of a note/event. The >composer could gauge the length of a note/event by listening and >let the computer take care of figuring out the timing. A point >and click, stretchable Leggo block... Each sonic event would be >unique. Each note wouldn't need a *starting time* relative to the >beginning of the piece. I really don't understand your problem, unless the only experience you've had with music and time is a sequencer that demands events be located by when they begin (in seconds or whatnot) after the beginning of the piece. This is, except for a few contemporary scores, completely unlike how duration is handled in either musical notation or performance. Historically, it's been a matter of proportion: the relative durations or notes to each other, and not to some fixed unit of time. Combine that with the notion of 'pace' or 'tempo'--the rate at which these propotional units occur (x beats per minute), the rate being flexible. There is no reason an event's duration cannot be given a quantity in relationship to event that preceeds it (hence, one pulse or beat followed by an event of two pulses or beats, followed by... which are only given an absolute duration (in seconds or whatever) when the tempo or pace of the piece at that moment is taken into account. >The apparent problem is that music is not this simple. It is not >monophonic, it is polyphonic. Notes *occur* in *parallel* as well >as *serial* streams. What to do? Polyphony is no problem. You simply have to give the 'impression' of simultenaity, not actual simultaneousness. (How quickly must the beginning of one event follow that of another in order to sound simultaneous?) (Again, though, what is an 'event'?) >How can we deal with time in music and make it simpler and more >intuitive? In a computer environment we have the freedom to >(conceptually) overcome this problem. I invite your suggestions, >comments, thoughts and snottiness (:-)). How to handle the information? Simply as a database with one event per record and as many fields as you need to describe it. You can manage the information in that form, and translate it if necessary into a form your simple sequencer can understand. >- Thanks for reading. | Let's have more ideas! >- Be seeing you. | These are powerful tools for powerful times. >- Scott. | New tools require new techniques for using them. Cheers, --Mark ======================================== Mark Gresham ARTSNET Norcross, GA, USA E-mail: ...gatech!artsnet!mgresham or: artsnet!mgresham@gatech.edu ========================================
dar@telesoft.com (David Reisner) (02/07/90)
In your article, you suggest representing monophonic/serial music as a
series of musical events of various durations - some sounds/notes, some
silence. You further observe that music is actually made up of many,
potentially simultaneous events, and ask how that should be represented.
I agree that a simple serial sequence is an appealing representation in
many cases, however I don't think a sequence should start at the beginning
of a piece and end at the end. The sequences are the fragments from which
music is contructed (from arpeggio, to phrase, to part/section).
Some sets of parallel events might be reasonably considered as several
simple sequences tied together in time; others float free of each other;
and some are tied together by some characteristic (e.g. tempo), but not by
specific times.
This could reasonably be represented as (attributed) connections between
simple sequences. A given connection would provide a tie between some
characteristics. A time attribute would specify that all the points in
all the sequences to which this connection exists must occur at the same
time. A tempo attribute would make the connected sequences have the same
tempo. Volume, inflection, rhythmic structure, motif could all be
possible attributes. (This implies that the sequences might also contain
only some of these attributes, so you could have (traditional) pitch
sequences, rhythm sequences, timbre sequences. These separations allow
some interesting musics to be described.)
(I'm not sure how to allow multiple connections with the same attribute
to a single sequence (a slow tempo connection near the beginning and a
fast one near the end, with the tempo characteristic of the sequence
"stretching" between the two points.)
-David
{uunet,ucsd}!telesoft!dar, dar@sdcsvax.ucsd.edumorgan@ms.uky.edu (Wes Morgan) (02/13/90)
>(I'm not sure how to allow multiple connections with the same attribute >to a single sequence (a slow tempo connection near the beginning and a >fast one near the end, with the tempo characteristic of the sequence >"stretching" between the two points.) This gets into the one realm of music which requires that human element: interpretation. Crescendos, decrescendos, cadenzas, ritardandos, and the like are what define the human touch. I've heard systems that use ran- dom number generators to determine rates of change, but it became rather disjointed. You may have to write some AI code to supply that, my friend. > >-David >{uunet,ucsd}!telesoft!dar, dar@sdcsvax.ucsd.edu Wes Morgan -- The opinions expressed above are not those of UKECC unless so noted. Wes Morgan \ {rutgers,rayssd,uunet}!ukma!ukecc!morgan University of Kentucky \ or morgan@engr.uky.edu Engineering Computing Center \ or morgan%engr.uky.edu@UKMA.BITNET