Brunel University College,
Electronic mail: firstname.lastname@example.org
Telephone: +44 (0)181 891.0121 ext.2767 Fax: +44 (0)181 569.9198
Brunel University College, UK
Dynamic form movement-based dance is emerging as a new form of performance practice. Recent advances and developments in artificial intelligence, computing and multimedia technologies have made possible the interactive integration of kinetic art forms with computers. They provide means to determine certain effects generated by individuals' responses which are usually achieved by using heuristic rules gained from experience. They also allow the exploration of different choreographic compositions on the computer screen by multimedia animation techniques. These efforts however, are based on logic and rational representations of movement-based performance which are inappropriate in dynamic form performance where interpretations and representations are continuously changing. Dynamic dance performance can be seen as a creative search for patterns of harmony which are not fixed but continuously remapped as the performance dynamically organises itself. This paper discusses the possible use of artificial neural networks techniques to explore/generate patterns of harmony in the evolution of dynamic form in movement-based dance. A framework for a hybrid tool to search/evolve patterns of harmony in movement-based dance is also proposed.
Dynamic form is emerging as a new form of performance practice. It is particularly though not exclusively, associated with physical theatre performance classified as non-narrative, non-literal or non-text-based. This new form of performance is characterised by certain key elements : adaptability, emphasis on interaction between all agents in the creative process; and the notion of emergent rather than pre-planned performance structure. The performers operate as adaptive agents in the emerging process itself as they interact with the material, the people ( each other and the audience ) and the sound environments.
The OPTIK production  exemplifies the nature of this new form of performance practice. OPTIK is a verbally textless movement-based dance performance concerned with a series of loaded moments, each with a potentially limitless variety of spatial and somatic possibilities. What constitutes the performance emerges from the set of initial conditions rather than any pre-planned idea. This demonstrates the dynamic nature of the technique employed by the performers as they work, as well as the interactive nature of the whole process.
Recent advances in information technology, especially artificial intelligence and multimedia are enabling the interaactive integration of kinetic art with computers. This new class of robotics is being recognised as autonomous kinetic art sculptures in art galleries, dance concerts and performance arts . They provide means to determine certain effects generated by individuals' responses and explore different choreographic compositions on the computer screen via animation techniques. These systems, usually employing heuristic rules gained from experience, are based on a logic and rational representation art performance.
2 Nature of dynamic form performance
However in dynamic form performance (as indeed in all live performance arts), this a priori fixing of heuristic rules is inappropriate as alternatives, emotions, interpretations (audience and artists), representations, perceptions are all in continuous flux of change. OPTIK's movement-based dance performance for example, focus on the essential dynamics of time, space and human contact, recognising and releasing the energy between the watcher and the watched in a prescribed space at any given time. The performance makes use of a highly developed performative instinct. The dance patterns thus formed are the result of the somatic intelligence of the performers . Dynamic form performance can be seen as a creative search for patterns, wholeness, harmony or cognitive equilibra amongst all the contributory factors. These patterns which are both spatial and temporal cannot be fixed and permanent, but continuously reformulated just as the live performance is constantly organising itself. Dynamic form can be regarded as a continuous organising effort along the critical line between harmony and disharmony. The performers have no resolution to their action given to them in advance, and as a result the performance develops into a complex dynamic system of patterning.
3 Artificial neural networks in dynamic form performance
Advances in neuroscience, cognitive science and the associated psychological data support this view [3, 4]. Human thinking and actions involve search for stable patterns among all influencing factors. This is determined by the way human neural networks are structured as a whole - as a spontaneously wired and rewired self organising 'economy' of repeatedly propagated patterns of formulation and reformulation. Viewed in these terms, artificial neural networks may be employed to explore and search for patterns of harmony in the evolution of dynamic form in dance movements.
Artificial neural networks (ANNs) are computational models which attempt to simulate the behaviour of the human brain. An ANN has a parallel distributed architecture of large number of simple neuron-like processing elements (nodes) and a large number of weighted connections between the elements. The weights on the connections encode the knowledge of a network. ANNs solve problems by self-learning and self-organisation, deriving their intelligence from the collective behaviour of the nodes. ANNs can recognise, classify, convert and learn patterns. There are now a variety of ANNs  which are adept at pattern recognition.
In applying ANNs to dynamic form movement-base dance performance, a suitable ANN model(s) can be used to recognise and learn as well as recognise and classify underlying "factors" contributing to patterns of harmony. Once this is achieved, the ANN(s) can be employed to explore and generate new patterns interactively in real time, hence illuminating the performance as it is experienced.
4 Proposed hybrid tool for dynamic form performance
The framework of a tool employing a hybrid of artificial neural networks, artificial intelligence, computing and multimedia technologies for the analysis, exploration and generation of movement-base dance patterns is shown in Figure 1. The tool will also enable studies of the relationship between live performance and artificial performance and the patterning and search for harmony that runs through dynamic form performance
Figure 1 Framework of proposed hybrid tool
Initially it is required to observe and discover patterns of harmony and their underlying contributory factors. This is done via the motion capture and translation interface module. Reflective sensors attached to the performers will enable their motion in space and time to be recorded on video and a digitised image obtained. This will then be translated into suitable form for input to the ANN model. The knowledge base which may take the form of a simple rule base, will be encoded with information of the spatial constraints of the environment, time constraint, other physical law constraints and heuristic rules of harmonisation articulated by the performers. Using mappings of the digitised motion images and knowledge base information as inputs, the ANN will then recognise and classify underlying factors and patterns of harmony in the movement-based dance performance.
Having achieved this, the ANN can be used interactively to explore and generate new dynamically harmonious dance patterns in real-time. Initial conditions of space, time and physical constraints may be changed and the ensuing patterning effects may be studied and comparisons made with the actual patterns that performers evolve. An important factor in any live performance is the emotion(s) of the performers at the time of performance. As information is distributed throughout the nodes of the ANN, it may be possible that emotions be identified and their effects  on change(s) in dance patterns be studied.
The dance patterns thus generated by the ANN model will be presented to the outside world via the multimedia user interface using animation software. The multimedia user interface will also enable user-friendly input/edit facilities for the knowledge base module and the motion capture/translation module of the tool.
The proposed hybrid tool will enable the study of relationships between artificial and somatic intelligence as understood and practised in live performing arts, in particular movement-based dance performances. It will also explore the interactive potential of state-of-the-art artificial intelligence and multimedia technologies to simulate real live art performances. Using the OPTIK production as case study, work is currently undertaken to observe and establish underlying factors contributing to dynamic harmonious patterns in movement-based dance performance. Investigation into the suitable ANN model(s) is also being initiated. It is hoped that an initial prototype of the tool will be developed in the near future.
 Performing performance - the new authencity of OPTIK, Total Theatre, Vol 7, No 1, 1995, pp11.
 Brill L M : Robots - artists of the electronic palette, AI Expert, Vol 19, No 1, Jan 1994, pp28.
 Margolis H : Patterns, thinking and cognition, University of Chicago Press, 1987.
 Edelmen G M : Bright light, brilliant fire - on the matter of the mind, Allen Lane, London, 1992.
 Grossberg S (ed) : Pattern recognition by self-organising neural networks, MIT, 1991.
 Levine D S (ed) : Motivation, emotion and goal direction in neural networks, LEA, 1992.
For more SYNDICATE SPEAKERS
For other SYNDICATES