|
|
Équipe Analyse/Synthèse |
After some discussion on the experience gained in previous works and on philosophical reasoning on machine attributes in human-machine performance, I. Choi discusses the notion of Interactivity versus Control. She begins by specifying the notion of interactivity in terms of system connectivity and parametrization.
In the system described , movements effect system states defined as machine responses. Such responses are registered as measurements taken at an array of pressure sensors - in this case, sensors placed at specially designed shoes. Other input devices include magnetic tracking, computer vision and positional devices. These registered measurements undergo an interpretation by a fuzzy inference process, noting that only the measurements in the range of state changes will be reflected in the machine response.
The inference process then returns an output value which is passed to response dynamics encoded in the machine, representing a range of motion in the computational model. The motion then determines the change of states in effector functions that describe output signals (such as graphics or sound).
The author then discusses the notion of parametrization of interactive parameters. As an example, parametrization of a brass tone synthesis algorithm can be made in terms of pitch, loudness and spectral distribution of sinusoidal components, etc. or, in the case of physical models, breath pressure, lip pressure, etc. Choi considers the choice of parametrization as capable of trivializing or enhancing interactivity in a human-machine relationship.
Next the author discusses the relationship between the observers of a musical performance and the musician/performer's movements (as in musical experience) as a perceptual interface to the process of listening. She considers that in the proposed system, the performer also plays a role of an observer, and that his/her actions are seen by other observers both (traditionally) as part of the musical performance and (non-traditionally) as an expertise accessible to other observers, as opposed to virtuoso music performance conditions. She considers that this accessibility changes the mode of other observers' spectation.
The manifold interface provides an interface to parameter control spaces with more than three dimensions by means of graphical lines and surfaces. It allows the user to navigate in a high-dimensional parameter space from a visual display with continuous gesture input (with at least two degrees of freedom).
The manifold controller (MC) is a set of classes linking graphics, hardware input devices, and sound synthesis engines. It can be defined as an interactive graphical sound generation tool and composition interface. Computational models involve sound synthesis models (also physically-based systems) and composition algorithms.
The idea behind the manifold interface is that of organizing control parameters in order to provide efficient system access. They also seek to provide a representation of the system that has visual simplicity. This is done by proposing the concepts of control space, phase space and window space (please refer to [13] for details).
Control space refers to both phase space and window space as a whole. Phase space is an n-dimensional space where vector arrays correspond to states of a parametrized system. It represents all the permissible combinations of parameter values of an algorithm.
In order to make it possible to visualize the n-dimensional phase space, the authors [13] propose the concept of a window space, that represents the mapping of the phase space data to a three dimensional space. Said in another way, it defines how a three-dimensional visual representation is embedded in the high-dimensional space. Therefore, an observer may control the window space by panning and zooming in the phase space. Citing the author: "The window space provides a domain for generating and modifying classes of control point sets. These points represent combinations of parameter values as user-specified, and they are associated with particular sounds". This association is reported to enhance the ability to identify boundaries where character shifts occur in the states of the system.
Choi continues by considering two main gestural acquisition system types: externalized systems, where a sensing device makes an observational measurement according to world-centered coordinates and human-centered systems, where the measurements share the coordinate system with the observer. She claims that the information cues sent by human (body)-centered systems are complementary to world-centered positional cues (for example, from the performer's eyes) and to gravity-centered balance cues (from the inner ear).
She has implemented an example of human-centered system by designing a foot-mounted interface, sensitive to natural stance and bipedal locomotion. Four pressure sensors are placed on the base of the boot at the heel, the inner and outer foot ball and at the toe tip. The outputs of these 8 sensors (2 feet) are then sent to an inference system based on fuzzy rules. The result of the fuzzy process will then control both graphic engines and sound synthesis engines in real-time.