INNER ROOM EXTENSION OF A GENERAL MODEL FOR SPATIAL
PROCESSING OF SOUNDS
Shahrokh Yadegari
Center for Research in Computing and the Arts
California Institute of Telecommunications and Information Technology
Department of Theatre and Dance, University of California, San Diego
ABSTRACT
F. Richard Moore proposed a general model for spatial
processing of sounds in 1982. [5] This model separates
space in two nested areas. The outer room is an imagimary
acoustic space within which the inner room (or the real
performance space) is located. The inner room is denoted
by the location of the speakers which simulate the sound
heard in the inner room as if the speakers were "openings"
connecting the inner and the outer room. The spatial impression is produced by diffusing simulated direct sound
rays, early echos, and global reverbration of the sound
sources as heard at each speaker location. The model does
not allow sound to travel through the walls of the inner
room, and thus, when sound sources are near the walls of
the inner room, or travel through these walls, unexpected
results may be heard in opposing speakers. Furthermore,
the simulation of sound sources inside the inner room are
not as convincing as simulation of sound sources outside
the inner room. This paper discusses an inner room extension of the general model. This extension defines the inner
room as multiple nested imaginary rooms, and it provides
an improved ray intersection algorithm. It also slightly alters the algorithm by which the delay time and attenuation
factors of the direct and reflected rays are calculated. This
extension ameliorates a number of undesirable effects and,
according to our subjective tests, provides a more convincing spatialization impression when the sound sources are
inside the inner room.
1. INTRODUCTION
This paper describes an extension to F. Richard Moore's
"general model for spatial processing of sound". Moore's
general model simulates the most perceptibly recognized
effects of room acoustic to produce the desired spatial
impressions and it draws on the works of Gardner [2],
Blauert [1], and Stevens [8], on psychophysics of spatial
perception, and those of Schroeder [7] and Moorer [6], on
simulation and the use of artificial reverberation.
The general model does not simulate spatial impressions for a specific listener position; however, it produces
spatial impressions for a concert setting where audiences
are located in various locations of the performance space.
The algorithm separates the space in two nested rooms.
The outer room is an imaginary acoustic space, inside
which the inner room is located. The speakers are located
on or near the perimeter of the inner room, and are considered to be "openings" connecting the inner and the outer
room. No sound is to be propagated through the walls of
the inner room. The sound of the imaginary outer room
is heard through the sound propagated through the "openings" at the location of the speakers by which the simulated sound of the outer room is diffused inside the inner
room. Thus Moore's model achieves spatial impressions
which are minimally dependent on the location of the audience in the performance space.
The strength of this model is to realistically localize
sound sources in the area inside of the outer room and
outside of the inner room. If we were to follow the same
algorithm for calculation and diffusion of direct rays and
reflected rays for sources inside the room, as we did for
sources outside the room, the simulated sound rays no
longer would mimic a physically realistic scheme; meaning that when a source is inside the inner room, the general
model cannot be applied to it. This is due to the fact that
in a realistic situation, even if the speakers were openings
to the outer room, the sound is no longer heard through
these "openings." This situation causes the produced spatial effects simulated for sound sources inside the inner
room not be as effective as those simulated for sources
outside the inner room. Furthermore, since no sound travels through the walls of inner room, any time a sound
source comes close to an inner wall or travels through it,
undesired effects could result in the simulated sounds rays
for speakers located opposite of this inner wall.
This paper discusses an extension of the general model
which improves the impression of simulated spatial effects for sound sources located inside the inner room, and
provides an algorithm to produce smooth curves for turning speakers on and off as the sound sources pass through
the walls of the inner performance room. This extension makes two modifications to the original Moore's algorithm and provides an improved ray intersection algorithm. The modifications are as follows: 1) the inner room
is defined as multiple nested imaginary rooms with imaginary speakers on their perimeter; when a source is inside the outmost inner room (the primary inner room), the
largest imaginary room is chosen so that the source is outside of that room, and delay and attenuation factors are
adjusted to diffuse the sound as if the sound was being
