Sonification of data and data-clusters using 2-dimensional waveguide mesh

Jonathan Berger
Stanford University
CCRMA

Jonathan Berger, Kyogu Lee, Gregory Sell - CCRMA, Stanford University


We describe possible applications of a 2-dimensional waveguide mesh for
sonification of complex data, object identification and classification. Our
overall goal is to distinguish highly dimensional data sets from one another
in such a way that reveals meaningful differences in a particular context.
Using a 2-dimensional mesh of impedance matrices we explore various methods
in which data controls or effects wave dispersion. We first consider
situations in which the data effects mesh size, and proceed to describe
mapping data to points or regions of impedance within the mesh. In preliminary experiments data parameters of a high-dimensional data set
were mapped as mesh control parameters. Since there is no physical
limitation to the size of the mesh N-dimensional data, no matter how large N
may be, can, for example, be mapped to the initial excitation condition of
each junction in the mesh.
In these initial experiments three mapping strategies were tested. These
include:


1. a mesh is created with the number of points equals the number of
dimensions in a particular data set. The initial condition may be any type
of wave variable - displacement, velocity, or force. Since the 2-D mesh is
rectilinear, we can have different mesh dimensions with any given (even)
number of points.


2. a mesh of size NxM is created, where N corresponds to the dimension of the data to be sonified, and M can be arbitrary. A planewave whose initial
conditions are determined by the data is used as an excitation along the
axis of N points.


3. a mesh of size NxM is created, where N corresponds to the dimension of
the data to be sonified, and M can be arbitrary. Instead of mapping the data
to an initial excitation condition as before, they are mapped to the
boundary condition of the mesh. Since one pole filters are used at the
boundaries the data can be mapped to control the gain or to change the pole
location of the filters. In this case, the initial excitation can be
anything - impulse, planewave, or a set of impulses.
Since the resulting timbre of the sound produced by the mesh is largely
determined by its size, success in producing sounds with salient
differentiation in timbre was limited. We thus proceeded to describe a new
technique using impedance is applied with the aim of establishing a sense of
perceptual distance and character.


In our current work the reflective behavior of a wave as it encounters
greater impedance is harnessed to direct the wave to certain regions within
the mesh. If a shell of high impedance is wrapped around a region of low
impedance, the interior becomes a zone of resonance producing a unique and
identifiable timbre and frequency when responding to an impulse. A promising aspect of altering the impedance within the shell to reflect
resonant properties is the audible percept of proximity to the region from
the impulse source. Because of the nature of the dispersion of a wave, a
point further from the impulse will receive a smaller proportion of the
wave's original energy. Thus, an impulse placed closer, but not within, a
resonance zone will filter more of its energy through the shell, creating a
significantly lower degree of resonance resulting in a noticeable alteration in timbre.


Back to MGA Workshop III: Multiscale structures in the analysis of High-Dimensional Data