REAL-TIME SIMULATION OF SOUND SOURCE OCCLUSION

Chris Share

Sonic Arts Research Centre

Belfast, U.K.

Graham McAllister

Sonic Arts Research Centre

Belfast, U.K.

Keywords:

Audio, multimedia, occlusion, virtual auditory environments.

Abstract:

Sound source occlusion occurs when the direct path from a sound source to a listener is blocked by an inter-

vening object. Currently, a variety of methods exist for modeling sound source occlusion. These include ﬁnite

element and boundary element methods, as well as methods based on time-domain models of edge diffraction.

At present, the high computational requirements of these methods precludes their use in real-time environ-

ments. In the case of real-time geometric room acoustic methods (e.g. the image method, ray tracing), the

model of sound propagation employed makes it difﬁcult to incorporate wave-related effects such as occlusion.

As a result, these methods generally do not incorporate sound source occlusion. The lack of a suitable sound

source occlusion method means that developers of real-time virtual environments (such as computer games)

have generally either ignored this phenomenon or used rudimentary and perceptually implausible approxima-

tions. A potential solution to this problem is the use of shadow algorithms from computer graphics. These

algorithms can provide a way to efﬁciently simulate sound source occlusion in real-time and in a physically

plausible manner. Two simulation prototypes are presented, one for ﬁxed-position sound sources and another

for moving sound sources.

1 INTRODUCTION

1.1 Sound Source Occlusion

Sound source occlusion occurs when the direct path

from a sound source to a listener is blocked by an in-

tervening object. An example of an occluded sound

source is shown in Fig. 1. In situations such as that of

Fig. 1., sound emanating from the source often still

reaches the listener, however its characteristics are al-

tered due to the size and orientation of the objects it

encounters during its propagation.

This is a common occurrence. In everyday life the

effects of sound source occlusion surround us, shap-

ing our perception of our physical environment and

providing important information about the size and

orientation of objects within this environment. If we

are to create immersive and believable acoustic simu-

lations of real environments it is important to include

this physical phenomenon.

Despite the existence of several different methods

for simulating sound source occlusion, in general this

phenomenon has not been included in computer sim-

ulations of acoustic environments. This is especially

true in the case of real-time simulations such as com-

puter games. In this case, developers have generally

either ignored the phenomenon of sound source oc-

clusion or used rudimentary and perceptually implau-

sible approximations. This is detrimental to the simu-

lation and detracts from its realism. Thus, as Martens

(Martens, 2000) points out, the simulation of sound

source occlusion remains one of the unsolved prob-

lems in the rendering of real-time virtual acoustic en-

vironments.

In this paper we present a novel approach to simu-

lating sound source occlusion in real-time virtual en-

vironments. This approach is based on the use of

graphical shadow algorithms. It is shown that this

approach has several advantages over existing meth-

ods and is capable of producing plausible results in

a timely and efﬁcient manner. Two simulation pro-

totypes are presented, one for ﬁxed-position sound

sources and another for moving sound sources.

In Section 2 of this paper we review current ap-

proaches to simulating sound source occlusion in vir-

tual environments. In Section 3 of this paper we

present a novel approach to the simulation of sound

193

Share C. and McAllister G. (2006).

REAL-TIME SIMULATION OF SOUND SOURCE OCCLUSION.

In Proceedings of the International Conference on Signal Processing and Multimedia Applications, pages 193-199

DOI: 10.5220/0001571501930199

 SciTePress

Figure 1: Occluded sound source.

source occlusion based on the use of shadow algo-

rithms from computer graphics. Two prototype im-

plementations using this approach are described. Fi-

nally, we discuss several ways in which the existing

two prototypes can be extended.

1.2 Real-time Rendering

This paper focuses on real-time applications so it is

important to be clear about the meaning of real-time

in this context. Akenine-Moeller et al. (Akenine-

Moeller and Haines, 2002) describe real-time graph-

ics rendering as a process of interaction between a

user and a screen image in which the user reacts to

the image, thereby producing appropriate changes in

the image. It is generally accepted that a real-time ap-

plication should be responsive to user input and that

the response time should be relatively short, prefer-

ably in the millisecond range. For graphics applica-

tions Akenine-Moeller et al. recommend a framer-

ate of between 15 and 72 frames per second in order

for the images to be perceived as continuous. This

gives a minimum rendering time of approximately 66

milliseconds for each frame. It is important to note

however, that in real-time graphics libraries such as

OpenGL, the framerate is determined by the complex-

ity of the scene that the application renders. The more

computationally intensive the scene, the slower the

framerate will become. This is because the applica-

tion must wait for the current call to the render loop

to complete before the next call to the render loop be-

gins. Although framerates below 15 frames per sec-

ond produce stilted, choppy animations and are con-

sidered less than optimal, they may still be acceptable

to users for short time periods.

The case of real-time audio rendering is somewhat

different to that of real-time graphics rendering. Au-

dio applications typically use a buffered, interrupt-

driven approach to audio generation. In this case,

a time-based operating system callback is periodi-

cally called to ﬁll an audio buffer. This callback

runs regardless of whether the audio rendering has

been completed or not. This means that developers

must ensure that the time taken to generate each au-

dio buffer is equal to, or less than, the time taken to

play the buffer (see (Bencina, 2003) for further dis-

cussion of this issue). Otherwise, the output audio

stream will be compromised. The time period of the

audio callback is normally user-deﬁned and can be

as little as 16 samples (approximately 0.3628 ms at

a sample rate of 44100 Hz) in low-latency applica-

tions. For this reason, the audio component of virtual

environments must be as efﬁcient as possible. The

constraints of real-time operation mean that computa-

tional complexity must be limited. Note also that in

the case of environments such as computer games, a

CPU allowance is usually allocated to the audio com-

ponent of the application. This is usually in the vicin-

ity of 5% to 10% of the application’s total CPU usage.

This further limits the allowable computational com-

plexity of the audio algorithms that are used.

1.3 Simulation vs. Modeling

The goal of the approach presented in this paper is

to produce a plausible output rather than one which

is necessarily physically accurate. This approach

mirrors that commonly taken in real-time computer

graphics simulations where the goal is to produce a

perceptually plausible output.

2 BACKGROUND

In comparison with other areas of acoustics research,

the effects of an occluding object on the wavefront

produced by a sound source has received relatively

little attention. In one of the few systematic studies

into this subject, Farag (Farag, 2004) has shown that

the general effect consists of two parts: a lowpass ﬁl-

tering effect due to the diffraction of the wavefront

around the occluding edge, and a comb ﬁltering ef-

fect due to the combination of the various different

paths that the wavefront takes on its way to the user.

Experiments by Martens et al. (Martens et al., 1999)

conﬁrm the presence of these two effects.

These changes in the wavefront produce signiﬁ-

cant acoustic effects that can be perceived by the lis-

tener. For example, Farag (Farag, 2004) has shown

that when a sound source is occluded by an object,

SIGMAP 2006 - INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING AND MULTIMEDIA

APPLICATIONS

194

the listener perceives a phantom source located at the

edge of the object from which the signal arrives ﬁrst.

Another ﬁnding in this study was that sound source

occlusion produces a perception of increased distance

to the sound source. Studies by Russell (1997) (Rus-

sell, 1997) suggest that listeners may also be able to

perceive information such as passability or the size of

openings from the acoustic signal. That sound source

occlusion produces signiﬁcant physical effects, and

that these effects are perceived by listeners, highlights

the need for the inclusion of this phenomenon in vir-

tual environments. By so doing, the believability and

immersiveness of the acoustic simulation will be en-

hanced.

At a basic level, the problem of simulating the ef-

fects of sound source occlusion involves solving the

wave equation at the position of interest (usually the

listener position) in terms of the wave equation at sur-

rounding positions. The two most commonly used ap-

proaches to this problem are the ﬁnite and boundary

element methods (see for example, Kopuz and Lalor

(Kopuz and Lalor, 1995)). These approaches solve

the wave equation by expressing it as a set of linear

equations derived from the discretization of the envi-

ronment in space and time. To do this accurately is a

difﬁcult computational challenge and is generally too

computationally intensive for use in real-time appli-

cations.

A related approach using waveguide meshes has

been proposed by Murphy and Beeson (Murphy

and Beeson, 2003) and Savioja and Lokki (Savioja

and Lokki, 2001). This approach extends the one-

dimensional digital waveguide technique pioneered

by Smith (Smith III, 1987) to meshes of two and

three dimensions. In terms of occlusion, an advan-

tage of this technique is that the wave propagation ef-

fects are an inherent part of the implementation. How-

ever, there are several difﬁculties with this approach.

Firstly, despite the simpliﬁcations, the method is still

computationally intensive. Also, there are unsolved

problems concerning dispersion and boundary condi-

tions that impact upon the accuracy of the simulation.

An alternative approach is the use of time-domain

models of edge diffraction (see for example, Lokki et

al. (Lokki et al., 2002)). In this approach, the im-

pulse response of the occluding edge is directly cal-

culated. The difﬁculty with this approach is that the

calculated impulse response is often quite long and

therefore, too computationally intensive for real-time

use. The authors suggest that the calculated response

can be simulated by using a warped IIR ﬁlter with

a matching frequency response. Unfortunately, dif-

ﬁculties involved in this matching process precludes

the use of this approach in real-time environments.

A variety of other approaches based on geometric

methods have been used to generate room acoustic

models (e.g. (Allen and Berkley, 1979), (Krockstadt

et al., 1968)) however these rarely incorporate diffrac-

tion or occlusion effects. In these cases, the model

employed assumes that sound propagates in the same

manner as light, that is, that the wavelength of the

propagating wave is much shorter than the size of the

objects in the environment. This means that low fre-

quency effects such as diffraction and occlusion are

not included.

Simpler approaches to simulating sound source

occlusion include the use of Fresnel zones (Tsin-

gos and Gascuel, 1997) which generate a frequency-

dependent visibility factor that can then be used to

set the parameters of a ﬁlter that is then applied to

the source signal. More simple again is the approach

taken by Takala and Hahn (Takala and Hahn, 1992)

who chose to reduce the effects of sound source oc-

clusion to that of changes in source gain using a value

proportional to the degree of occlusion. A similar ap-

proach is generally taken in computer games where

sound source occlusion is either completely ignored

or simulated using a simpliﬁed model that attenuates

the source signal, and in some cases, also ﬁlters it us-

ing a lowpass ﬁlter. A typical example of this latter

approach is that of Creative’s EAX software (Envi-

ronmental Audio Extensions 5.0, 2006).

From the above discussion it can be seen that at

present there is no method capable of producing plau-

sible sound source occlusion effects in a real-time en-

vironment such as a computer game. Lokki et al.

(Lokki et al., 2002) suggest that in order to simulate

this phenomenon, a mapping directly from a geome-

try (positions of source, receiver and edge) to diffrac-

tion ﬁlter coefﬁcients is needed. A potential solution

to this problem is the use of shadow algorithms from

computer graphics. This approach can be used to ef-

ﬁciently and plausibly simulate sound source occlu-

sion. An implementation of this approach is described

in the following section.

3 IMPLEMENTATION

3.1 Soundshadows

One possible way of mapping virtual scene geome-

try to ﬁlter coefﬁcients is through the use of real-time

graphical shadow algorithms. The effect of this ap-

proach is to produce what we have termed soundshad-

ows. These soundshadows can be used to store the

parameters of a ﬁlter system that simulates the aural

effect that a listener perceives when an object inter-

venes between their position in the environment and

that of the sound source.

The generation of shadows has been a focus of

computer graphics research since the 1970s. Since

then, a wide variety of algorithms for generating

REAL-TIME SIMULATION OF SOUND SOURCE OCCLUSION

195

graphical shadows have been developed. In general,

these methods can be grouped into ﬁve distinct cate-

gories:

1. Lightmaps (Fake shadows)

2. Projected Planar Shadows (Blinn, 1988)

3. Shadow Volumes (Crow, 1977)

4. Shadow Maps (Williams, 1978)

5. Soft Shadows (e.g. (Assarsson, 2003))

In this paper we describe implementations of

soundshadows based on the ﬁrst two methods listed

above: lightmaps and projected planar shadows.

These algorithms were implemented in C using the

OpenGL graphics library (OpenGL, 2006) and GLUT

windowing toolkit (Robbins and Kilgard, 2006). A

visual scene comprising a sound source, an occlud-

ing object and a listener was developed. The user ob-

serves the scene from a ﬁrst-person perspective and

can move about within the scene using keyboard com-

mands. In order to simulate changes in scene geom-

etry, and hence changes in occlusion effects, a door

which can be opened or closed by the user is included

in the scene.

As this project has a focus on real-time applica-

tions, and in particular, computer games, it was de-

cided to use the OpenAL audio library (OpenAL,

2006) for audio output. OpenAL is a free, open

source, cross-platform 3D audio library. The library

models source-listener distance and azimuthal source

position. The library has both hardware rendering and

software rendering options available. In the case of

the latter, source distance and position effects are sim-

ulated using changes in output signal amplitude. In

the case of the prototype software, the listener posi-

tion was simulated using a mono output signal, and

the effects of the occluding object were simulated in

two dimensions only. OpenAL does not incorporate

any audio ﬁltering capability so in order to simulate

the effects of an occluding object it was necessary to

add this to the library. Hence, a second-order IIR ﬁl-

ter was added to the library. The update rate of the oc-

clusion calculation is determined using a combination

of the OpenGL frame-rate and an operating system-

based timer. It was found that an update period of

0.05 seconds provided suitable results for changes in

the occlusion value.

The general algorithm used for implementing

graphical shadows as soundshadows is shown in Fig-

ure (2).

The algorithm is implemented in the following

manner. First, a test is performed to determine

whether the listener is in an area designated as the oc-

cluded sound ﬁeld. If the result of this test is negative,

then no further processing of the soundshadow algo-

rithm takes place. If the result of the test is true, then

the algorithm returns an occlusion value between 0.0

Figure 2: General soundshadow algorithm.

and 1.0. This value is then used to set the cutoff fre-

quency of a lowpass ﬁlter, thus simulating the effect

of the occluding object. A key component of the al-

gorithm is the manner in which the occlusion value is

calculated. This is described further in the following

two sections.

3.2 Lightmaps and Soundmaps

Lightmaps (also know as shadow maps) are a com-

mon part of virtual environments, especially com-

puter games. They provide a fast, efﬁcient way

to simulate lighting effects for ﬁxed-position light

sources. Lightmaps are essentially graphical textures,

usually greyscale bitmaps, that are layered over an

existing texture and surround a light source. The

lightmap is blended with the underlying texture so

that a light-to-dark fade effect, corresponding to dis-

tance from the light source, is produced. This simu-

lates the effect of the light source on the surrounding

object. It is important to note that lightmaps suffer

from several inherent limitations, in particular, that

they cannot interact with other objects in the environ-

ment. For example, an object moving past a lightmap

will not receive any light from it. Another limitation

is that lightmaps can only be applied to ﬁxed position

light sources, as it is too computationally expensive to

recalculate the lightmap as the light source moves.

A ﬁrst soundshadow prototype has been developed

using the lightmap algorithm as a model. As noted

above, this algorithm can only be applied to station-

ary sound sources (although it does support chang-

ing environment geometry). A screen capture from

this prototype is shown in Figure 3. Note that in the

screencapture the soundmap is visible but that in gen-

eral use, it will not be seen by the user.

In the case of the soundmap approach, the occlu-

sion value is calculated by ﬁnding the value stored in

the soundmap at the current user position. The man-

ner in which the soundmap is initially created is ﬂex-

ible and can be as simple or as complex as the devel-

oper wishes. For the prototype described above the

SIGMAP 2006 - INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING AND MULTIMEDIA

APPLICATIONS

196

Figure 3: Using a soundmap to simulate sound source oc-

clusion.

soundmap was created by drawing it using a graph-

ics program. However, because the creation of the

soundmap is an ofﬂine process, more accurate ap-

proaches could certainly be used and are currently un-

der development.

3.3 Projected Planar Shadows

Blinn (Blinn, 1988) ﬁrst introduced the concept of

projected planar shadows. In this approach a graphi-

cal shadow is generated by projecting the light source

on to a two-dimensional plane. The algorithm is sim-

ple and fast however it does not produce intensity val-

ues within the shadow ﬁeld. The algorithm for gener-

ating a projected planar shadow is shown in (1).

P = V +

d − N.V

N.V − N.L

+ (V − L) (1)

where P is a point on the plane, V is the vertex po-

sition of the object casting the shadow, d is the dot

product of P and N, N is a normal to the plane and L is

the light position. Unlike the lightmap algorithm de-

scribed above, the projected planar shadow algorithm

can handle moving light sources.

A second soundshadow prototype based on the pro-

jected planar shadow algorithm has also been im-

plemented. This prototype supports moving sound

sources. A screencapture from this prototype is

shown in Figure 4. Again note that although the

soundshadow is displayed, it will not usually be seen

by the user.

In the case of the projected planar soundshadow ap-

proach, the occlusion value O is calculated using the

two distance measurements d and c shown in Figure

5. The ﬁrst measurement used is the distance from

the listener position to the occluding object, while the

second measurement used is the distance from the lis-

tener position to the centreline of the occluded sound-

ﬁeld. These two values are combined to produce an

overall occlusion value according to (2).

Figure 4: Using a projected planar shadows to simulate

sound source occlusion.

Figure 5: Derivation of the occlusion value in the case of

projected planar shadows.

O = (1 −

) × (1 −

) (2)

Note that by using this method the value at the edge

of the occluded soundﬁeld will always be equal to

0.0 which is interpreted as meaning that no occlusion

takes place.

4 DISCUSSION

The initial implementations of the soundmap and pro-

jected planar shadow prototypes incorporated only a

lowpass ﬁlter effect. While their simulation of sound

source occlusion was reasonably effective, it immedi-

ately became apparent that the phantom sound source

REAL-TIME SIMULATION OF SOUND SOURCE OCCLUSION

197

arising at the nearest occluding edge needed to be in-

cluded in the simulation to make it plausible. In cases

where the prototype simulated multiple paths from the

sound source to the listener around several occluding

edges, it was found that the comb ﬁlter effect was a

crucial part of the simulation. Its absence reduced

the plausibility to a large degree. This was particu-

larly true in cases where sound reached the listener

via paths of approximately similar length.

The comb ﬁlter effect was found to be particularly

important in the case of the moving sound source pro-

totype. In this case it was found that the introduction

of a type of amplitude panning between the various

occluding edges was necessary in order to create a

convincing effect as the sound source moved within

the environment. By incorporating a pathlength-

dependent delay into the sound source’s signal it was

found that the comb ﬁlter effect produced by the mul-

tiple phantom sources could be approximated.

Implementing sound source occlusion using graph-

ical shadow algorithms has several key advantages

over existing real-time methods. Of primary impor-

tance is that the algorithms are designed to run in real-

time. A secondary advantage is that the algorithms

used may already be part of the virtual environment

software. Most computer games now include some

form of real-time shadow simulation. This means that

implementing soundshadows requires less additional

coding effort on the part of the environments develop-

ers. Another potential advantage is that as these algo-

rithms become more widely used it is likely that they

will be implemented in graphics hardware, thereby al-

lowing further acceleration.

Simulating sound source occlusion as a sound-

shadow can be justiﬁed on the grounds that sound,

like light, is a wave phenomena and so can be mod-

eled in a similar manner. However, it is important

to note that there are signiﬁcant differences between

sound and light which must be taken into consid-

eration and which limit the accuracy of the simula-

tion. These differences, described by Funkhouser et

al. (Funkhouser et al., 2003) include wavelength,

speed, coherence, dynamic range and latency. These

differences impose natural limits on the accuracy of

the simulation.

In order to test the accuracy of the soundshadow

algorithms an application is currently being devel-

oped which compares the simulated frequency re-

sponse at particular listener positions with the pre-

dicted analytical frequency response. Farag (Farag,

2004) showed that Svenssons edge diffraction model

(Svensson et al., 1999) agrees well with experimental

evidence. As a result, it was decided that Svensons

model would be used as the standard by which the ac-

curacy of the various sound shadow algorithms output

would be evaluated. Due to the computational com-

plexity of the Svensson model this application cannot

run in real-time.

5 CONCLUSIONS AND FUTURE

WORK

From the preliminary results produced by the sound-

shadow approach to simulating sound source occlu-

sion it can be concluded that graphical shadow al-

gorithms can be used to simulate the effects of this

physical phenomenon in a plausible manner in real-

time. The approach produces better results than those

currently achieved in commercial software and has a

negligible impact on the graphical framerate.

Further research remains to be done in order to in-

crease the quality of the simulations. In the case of

the soundmap prototype, the ﬁrst step is to develop a

method of accurately generating the soundmap. This

could be performed using an ofﬂine process such as

that outlined in Svensson (Svensson et al., 1999).

In the case of the projected planar shadow proto-

type it is possible that a soundmap could also be

used, however the constantly changing positions of

the source/occluder/listener may produce occlusion

effects that are too complex to simulate accurately

with this approach. Instead, it may be more useful

to investigate more sophisticated ﬁlter algorithms that

better approximate the predicted frequency response,

and then use the approach described above to control

the ﬁlter parameters.

The virtual environment created in the two proto-

type applications is very simple, comprising only a

single sound source, occluding object and listener. In

practice, most virtual environments are far more com-

plex than this, consisting of possibly hundreds or even

thousands of sound sources and objects. Hence, an

important question is the degree to which the sound-

shadow method can be scaled to larger virtual envi-

ronments. Closely related to this question is the issue

of performance. In the case of more complex environ-

ments it will be necessary to evaluate the efﬁciency of

the soundshadow algorithms using benchmarking and

asymptotic analysis so that the efﬁciency of the algo-

rithms is ensured.

Another important question is the degree of accu-

racy necessary in the occlusion algorithm in order to

produce an acceptable occlusion effect. In general,

occlusion has not been considered an important effect

to simulate in real-time environments and so has been

coarsely simulated. However, as games and virtual

environments become more realistic, it is likely that

more accurate acoustic simulations will be required.

For this reason it is necessary to determine the level

of simulation error that is acceptable to the user in the

occlusion model. Perceptual testing using the proto-

types described above may be able to provide an an-

SIGMAP 2006 - INTERNATIONAL CONFERENCE ON SIGNAL PROCESSING AND MULTIMEDIA

APPLICATIONS

198

swer to this question. Initial feedback from informal

listening tests that were conducted as the prototype

software was being developed suggest that the accu-

racy of the soundshadow approach provides a satis-

factory level of plausibility for most users.

The development of a simple, efﬁcient yet plausi-

ble method of simulating sound source occlusion in

virtual environments will contribute to making these

environments more immersive and believable. It is

hoped that this research can be used to produce virtual

environments in which audio plays a more meaning-

ful role in the user experience.

ACKNOWLEDGEMENTS

Special thanks to Dr. Graham McAllister and Prof.

Michael Alcorn for their supervision and guidance.

Thanks to Mr Chris Corrigan for technical assistance.

Funding for this project was provided by a SPUR

scholarship.

REFERENCES

Akenine-Moeller, T. and Haines, E. (2002). Real-time Ren-

dering. A K Peters.

Allen, J. B. and Berkley, D. A. (1979). Image method for

efﬁciently simulating small room acoustics. Journal

of the Acoustical Society of America, 65(4):943–950.

Assarsson, U. (2003). A Real-Time Soft Shadow Volume

Algorithm. PhD thesis, Department of Computer En-

gineering, Chalmers University of Technology.

Bencina, R. (2003). Portaudio and media synchronisation

- it’s all in the timing. In Proceedings of the Aus-

tralasian Computer Music Conference, pages 13–20.,

Perth. Australasian Computer Music Association.

Blinn, J. (1988). Me and my (fake) shadow. IEEE Comput.

Graph. Appl., 8(1):82–86.

Crow, F. C. (1977). Shadow algorithms for computer graph-

ics. In SIGGRAPH ’77: Proceedings of the 4th an-

nual conference on Computer graphics and interac-

tive techniques, pages 242–248, New York, NY, USA.

ACM Press.

Environmental Audio Extensions 5.0 (2006).

http://developer.creative.com/.

Farag, H. H. A. (2004). The Psychoacoustics of Sound-

Source Occlusion. PhD thesis, Faculty of Engineer-

ing, Alexandria University.

Funkhouser, T., Tsingos, N., and Jot, J.-M. (2003). Sur-

vey of methods for modeling sound propagation in in-

teractive virtual environment systems. Presence and

Teleoperation.

Kopuz, S. and Lalor, N. (1995). Analysis of interior

acoustic ﬁelds using the ﬁnite element and the bound-

ary element methods. Applied Acoustics, 45:193–210.

Krockstadt, A., Strøm, S., and Sørsdal, S. (1968). Calculat-

ing the acoustical room response by the use of a ray

tracing technique. Journal of Sound and Vibrations,

8(18):118–125.

Lokki, T., Svensson, P., and Savioja, L. (2002). An efﬁcient

auralization of edge diffraction. In Proceedings of the

AES 21st International Conference, pages 166–172.

Audio Engineering Society.

Martens, W. L. (2000). Efﬁcient auralization of small clut-

tered spaces: Simulating sonic obstructions at close

range. In Proceedings of the 7th Western Paciﬁc

Regional Acoustics Conference, pages 317–320, Ku-

mamoto, Japan.

Martens, W. L., Herder, J., and Shiba, Y. (1999). A ﬁltering

model for efﬁcient rendering of the spatial image of

an occluded virtual sound source. In Proceedings of

the Forum Acusticum, Berlin, Germany.

Murphy, D. T. and Beeson, M. J. (2003). Modelling spatial

sound occlusion and diffraction effects using the dig-

ital waveguide mesh. In Proceedings of the AES 24th

International Conference, pages 207–216. Audio En-

gineering Society.

OpenAL (2006). http://www.openal.org/.

OpenGL (2006). http://www.opengl.org/.

Robbins, N. and Kilgard, M. (2006). The

OpenGL Utility Toolkit (GLUT).

http://www.xmission.com/ nate/glut.html.

Russell, M. K. (1997). Studies in Perception and Action IV,

chapter Acoustic Perception of Aperture Passability,

pages 88–91. Lawrence Erlbaum Associates.

Savioja, L. and Lokki, T. (2001). Digital waveguide mesh

for room acoustic modeling. In Acoustic Rendering

for Virtual Environments: ACM SIGGRAPH and Eu-

rographics Campﬁre, Snowbird, Utah, U.S.A.

Smith III, J. O. (1987). Musical applications of digital

waveguides. CCRMA Report No. STAN-M-39, Stan-

ford University.

Svensson, U. P., Fred, R. I., and Vanderkooy, J. (1999). An

analytic secondary source model of edge diffraction

impulse responses. Journal of the Acoustical Society

of America, 106:2331–2344.

Takala, T. and Hahn, J. (1992). Sound rendering, proceed-

ings,. In Computer Graphics (Proceedings of SIG-

GRAPH ’92), pages 211–220. ACM Press.

Tsingos, N. and Gascuel, J.-D. (1997). Soundtracks for

computer animation : Sound rendering in dynamic en-

vironments with occlusions. In Graphics Interface’97,

Kelowna B.C., Canada, pages 9–16.

Williams, L. (1978). Casting curved shadows on curved

surfaces. In SIGGRAPH ’78: Proceedings of the 5th

annual conference on Computer graphics and interac-

tive techniques, pages 270–274, New York, NY, USA.

ACM Press.

REAL-TIME SIMULATION OF SOUND SOURCE OCCLUSION

199