Matt Gough Blog [closed]: Immeasurable
Tuesday, May 24 2005, 01:54
i received the details for Immeasurable? The Dance in Dance Science over the weekend. there is much work to do before then, a recent sharing of my work revealed complexities i had not observed. whilst the 'dance' audience understood the somatic context of the work they found it difficult to understand the technology. more than anything they showed a desire for a more extended explanation if the terminology, rather than failing to understand the links i was making. what's interesting (for me) is that my technical paper was under discussion, rather than a text targeted for the dance audience. so now, i look to see what changes i can make when presenting my work in the dance setting.
the problem of transferring my work for easy consumption by dance scientists, academics and artists is compounded by some rather dubious computer science (CS) movement theories. There is a common belief in CS that a global movement language exists which can be identified to describe al human motion. They hope to identify this language through pattern recognition and the statistical decomposition of human motion to discover the basic motor primitives or 'movemes'.
Yet human movement has no embedded meaning other than that imposed by the spectator or performer. Because the range of potential movement is so large (limited only by the constrains of the human body), movement techniques reduce the number of possible solutions to provide a usable classification. these classifications are not so much languages but dialects whilst their construction is systematic but subjective and essentially arbitrary.
a gesture recognition system trained to recognise features that match a particular dialect is not performing global recognition but probabilistic assessment of feature minima. each of these minima, when considered within another dialect or sub dialect could represent different, incomplete or invalid movemes. with many movement domains (eg. ballet, t'ai chi ch'uan, figure skating) being pluricentric the problem of identifying the feature minima of a gesture becomes more difficult. diglossia in a movement domain (variations between the technical and applied form) adds additional layers of complexity.
further more gestures are not always mutually intelligible, whilst a ballet dancer may recognise aspects of the salchow, lutz and axel (figure skating) their method of execution may not be technically understood (effect / affect divergence).
if CS is going to borrow from linguistics to classify movement it needs to pay more attention to the theories and practice of each domain rather than looking inwardly to speech and language processing or natural language processing. there is a clear difference between physical phenomena and their structuralist abstractions, it is this difference that most CS papers on gesture recognition fail to acknowledge alongside confusing effect and affect.
so whilst these basic concepts are nothing new to dancers the language and parallels drawn are unusual. what i hope to show is that we, as dancers hold valuable phenomenological knowledge about our practice that can be applied to a wide range of contexts.
science can learn from the arts, but only if it takes the time to listen
Filed by Article19 at 1:54 PM | 
Stumble It!