See discussions, stats, and author profiles for this publication at: [Link]

net/publication/216813562

Simulation, visualization, and virtual reality based modeling of room

acoustics

Conference Paper · January 2007

CITATIONS READS

4 317

5 authors, including:

Eduard Deines Frank Oberste-Dommes

CD-adapco IG Metall
22 PUBLICATIONS 272 CITATIONS 22 PUBLICATIONS 178 CITATIONS

SEE PROFILE SEE PROFILE

Martin Hering-Bertram Jan Mohring

Bremen University of Applied Sciences Fraunhofer Institute for Industrial Mathematics
85 PUBLICATIONS 1,023 CITATIONS 70 PUBLICATIONS 963 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Jan Mohring on 31 May 2014.

The user has requested enhancement of the downloaded file.

 19th INTERNATIONAL CONGRESS ON ACOUSTICS
MADRID, 2-7 SEPTEMBER 2007

SIMULATION, VISUALIZATION, AND VIRTUAL REALITY BASED

MODELING OF ROOM ACOUSTICS

PACS: [Link]

Deines, Eduard1,4; Michel, Frank1; Hering-Bertram, Martin2; Mohring, Jan2; Hagen, Hans3
1
IRTG Kaiserslautern
2
ITWM Kaiserslautern
3
University of Kaiserslautern
4
e_deines@[Link]

ABSTRACT
We present an audio-visual Virtual Reality display system for simulated sound fields. This
system combines the simulation of acoustics inside closed rooms, different approaches for
visualizing the sound propagation as well as the auralization of the resulting signals. The room
acoustic simulation is performed by use of a particle tracing algorithm, the Phonon Tracing,
applied at middle and high frequency range, combined with a wave-based approach (FEM
solver) utilized for low frequencies. For a realistic representation of the simulation results we use
a stereoscopic back projection system for rendering as well as a professional sound system for
audio reproduction. For auralization purposes we developed a soundfield synthesis approach
for accurate control of the loudspeaker system.

INTRODUCTION
Acoustics is important for planning of rooms for musical or speech presentations, such as
theatres, concert halls, and lecture rooms. Today a computer aided simulation of acoustical
behavior inside closed rooms is possible. There exists a variety of simulation algorithms which
can be classified into two groups: wave-based and geometric approaches. The former
numerically solve the wave equation. They provide accurate results, but the complexity
increases drastically with the highest frequency considered. Hence, the wave model is suitable
for low frequencies, only. Thus, in room acoustics mostly geometric approaches are applied.
Two classical representatives of geometric acoustics are the image source method [1] and the
ray-tracing method [16, 17]. An approach making use of advantages of image-source method
and ray-tracing is introduced in [25]. Here the visibility check of the image-source algorithm is
performed via ray-tracing. Beam-tracing methods [9, 10, 21] overcome the aliasing problem of
classical ray-tracing by recursively tracing pyramidal beams, implying the need for highly
complex geometric operations. Approaches utilizing the photon mapping algorithm [12] for
acoustic simulation also exist [13, 4].
The question comes up for an appropriate visual representation of the simulation results.
Stettner et. al. [24] visualize room acoustic quality properties such as clarity or spatial
impression by use of specific icons. Furthermore, they color the room boundaries according to
the pressure level. Khoury et. al. [14] represent the sound pressure levels inside the room by
means of color maps. Additionally, the authors analyze the precedence effect (or law of the first
wavefront) by using isosurfaces. Funkhouser et. al. [9] use visualization of source points,
receiver points, pyramidal beams, reverberation paths etc. in order to understand and evaluate
their acoustic modeling method. The system also provides the presentation of acoustic
measures like power and clarity. Existing commercial systems 1 , 2 , 3 provide some tools for
visualizing the computed acoustic metrics throughout the audience.
Virtual Reality environments provide an immersive representation of computer-generated
scenes. Integrating acoustic simulation, visualization, and auralization into the design process

1
[Link]
2
[Link]
3
[Link]
aims at interactive design and immediate exploration of virtual models. Several approaches to
integrate acoustic simulation into Virtual Reality systems have been proposed in recent years.
The DIVA system [23] provides sophisticated synthesis and spatialization algorithms. Here set
of either perception-based or physics-based parameters for auralization is defined. In [22] an
audio-rendering system for use in immersive virtual environments, the blue-c, is proposed. It is
optimized for efficient rendering of moving sound sources. Since the applications of this system
are restricted to fictitional scenes no physical-based acoustic simulation is provided. In [20] the
authors describe a four wall VR system for immersive visualization of sound rays propagation
inside closed rooms. A real-time audio system applied in a CAVE like environment with four
loudspeakers is proposed in [19]. A real-time simulation and rendering of moving sources and
listeners is possible.
In the present work we describe an audio-visual Virtual Reality system, which allows computer
aided simulation, visualization and auralization of acoustic behavior of a room model. For
auralization purposes we combine a FEM approach and phonon tracing [8] in order to obtain a
realistic impression of the sound perceived at given listener positions. The wave field synthesis
approach then allows a correct auditive output of the convolved signals on professional sound
equipment. With our system, a walkthrough of the listener positions is possible, such that a
visual and auditive impression of the scene can be provided.
From our visualization, the effect of different materials on the spectral energy distribution can be
observed. The first few reflections already show whether certain frequency bands are rapidly
absorbed. The absorbing materials can be identified and replaced in the virtual model,
improving the overall acoustic quality of the simulated room. A wave front is the pressure wave
corresponding to a unit pulse. In order to visualize the wave fronts spread out from the sound
source we use both particles (spheres) and surfaces, color coded by use of energy spectra of
the corresponding virtual sound particles called phonons. After a sufficient time period, a great
number of reflections have occurred, such that individual wave fronts cannot be identified,
anymore. The darker the color of spheres or surfaces gets, the more energy has been absorbed
by the different materials. Applications of our work include (but are not limited to) the acoustic
improvement during architectural design, and equipment phases of class- and congress rooms.

ACOUSTIC VIRTUAL REALITY SYSTEM

In this section we describe our visual and auditive virtual reality display system, see Figure 1. As
input our system requires the geometry model of the room, the absorption/reflection properties
of the room surfaces, and sources and listener positions. The module “Acoustic simulation”
computes the modal state space model (low frequencies) by use of the Finite Element Method
(FEM), and the room impulse response (RIR) and the phonon map (middle and high
frequencies) by means of phonon tracing algorithm. The simulation approaches are presented
in the next section. On the basis of this information the anechoic source signal is modified for
sound synthesis in the “Acoustic rendering” module. Thereafter, the sound field synthesis is
performed utilizing the acoustic hardware. The “Acoustic Visualization” module provides the
visualization of wave propagation from the sound source as well as the visualization of the
resulting phonon map, described further on in the text, following by the rendering step (“Visual
rendering” module).
For auralization purpose we used our synthesis method (see SYNTHESIS). The loudspeakers
of our audio system are driven separately with the corresponding signals. The audio playback is
administrated by the master computer and is started in a new process (thread).

Figure 1 - Modules of the audio-visual Virtual Reality system.

19th INTERNATIONAL CONGRESS ON ACOUSTICS – ICA2007MADRID

SIMULATION
For the simulation of the acoustics inside closed rooms we apply a combined method. Since in
the middle and high frequency range the diffraction is less important we use the phonon tracing
algorithm described in [4] for the simulation at these frequency bands. A recent improvement of
this algorithm [8] uses Gaussian basis functions with approximate partition of unity to represent
wave fronts. This is obtained by constructing the basis functions on the unit sphere around a
sound source and dilating them with the traversed distance of the associated phonons. In
contrast to the original approach where quadratic attenuation is used because of the spatial
particle density, the improved phonon trace facilitates both linear and quadratic attenuation for
the simulation of energy and pressure, respectively. Since the sign of energy is always positive,
we consider pressure in order to correctly represent annihilation effects. This is obtained by
scaling the basis functions according to the wave front propagation. Due to partition of unity,
both simulated fields are continuous and smooth.
Phonon tracing or any other methods based on geometric acoustics (ray tracing, mirror image)
fail in the low frequency range for two reasons:
1. Wavelengths are of the order of typical dimensions of the room. Hence, diffraction and
interference can no longer be neglected.
2. Damping is typically low at low frequencies and reverberation times become too long to
be represented by a convolution kernel of reasonable length.
Therefore, we have to fall back on wave acoustics to simulate the low frequency part of the
sound field. For closed rooms the wave equation is preferably solved by the finite element
method (FEM), which approximates the wave equation by a large system of ordinary differential
equations (ODEs) the unknowns of which are the pressures at grid points covering the room. In
general, there are by far too many unknowns to solve these systems of ODEs in real time.
Hence, we need to reduce the system to a concise statespace model with similar input-output
behavior in the frequency range of interest. There are many different approaches to model
reduction [2]. The common observation is that system dynamics can often be represented quite
well by a superposition of a few (generalized) eigenmodes. The coefficients of these modes are
the unknowns of the new reduced system. Finally, assuming samplewise constant input (e.g.
acceleration of the loudspeaker membrane) the continuous state-space model is transformed
into a discrete one, which can be solved in real time. For a detailed description of this approach
we refer to [8].
In practice, a low pass filter is used to split the input signal into a low frequency part and a
remainder. The low frequency part ( < 300 Hz) is handled in the way described in [8] and the
remainder by phonon tracing.

SYNTHESIS
There are two major sound synthesis approaches: wave field synthesis [5] and binaural
synthesis [3]. The latter approach uses either headphones or loudspeakers utilizing the cross-
talk cancellation [18] to reproduce the sound field simulated at the ears of the virtual listener.
The advantages are low cost equipment and low number of channels to be computed. However,
the listener needs to wear either headphones or a tracking device. The sound field is only
reconstructed for two positions inside the ears, filtered by head-related transfer
functions (HRTFs) obtained from psycho-acoustic measurements. For a greater range of
accuracy we apply wave field synthesis. The underlying theory is based on Huygens' principle
implying that solutions of the Helmholtz equation correspond to distributions of sound sources
on the surface of the domain (Kirchhoff-Helmholtz integral, [11]). This continuous distribution is
approximated by a finite number of loudspeakers. The higher the number of speakers and the
lower the frequency the greater is the sweet spot where the approximation is accurate. Usually,
only the sound field of a 2D listening plane is synthesized as this requires already quite large
number of speakers. For instance, Fraunhofer IDMT furnished a cinema with 192 speakers to
achieve good reproduction at all seats [15]. For our purpose it is enough to have a good
reproduction close to the proband's head the position of which is, except for small movements,
assumed to be fixed. The present system comprises currently 4 full-band speakers and a bass.
In a future system additional speakers will be added. In the following we describe our synthesis
approach.
Let ~
p ∈ 2 nb be a section of the digital signal simulated in point ~
i x ∈ 2 of the horizontal
i
measuring plane cutting the virtual room. nb is the block length. We choose a power of 2 to
speed up later Fourier transforms. The np points are distributed equally on a small circle of

19th INTERNATIONAL CONGRESS ON ACOUSTICS – ICA2007MADRID

radius r about the virtual head position. The radius has to be of the same order as the wave
c
length corresponding to the highest frequency to be reproduced: r ≈ λmin = 0 .
f max
For f max = 20kHz and a sound velocity of c0 = 343m / s the radius is only 12 mm. The exact
radius depends on the number of loudspeakers nl . To get an idea, imagine that an arbitrary
sound field of frequency f max has to be approximated by a polynomial with nl coefficients which,
for small nl , will be valid only within a small radius. Choosing r this small we avoid artefacts in
reproducing the high frequency part of the sound field while, at lower frequencies, the
approximation will still be good also for larger radii.
The system for sound field synthesis is made to reproduce sound in n p fixed positions on a ring
of radius r about a point in the projection room. As the proband may change the line of vision in
the virtual room these points cannot be directly associated with the n p points from the fixed grid
where sound is simulated. Hence, the simulated samples ~ p of time step n are interpolated i,n
and evaluated at the rotated real positions xi using cubic splines in the angle.
The resulting signals pi ∈ 2 nb are first windowed by a lifted cosine and then subjected to a fast
Fourier transform:
⎡ ⎛ π (n + 0.5) ⎞⎤
pˆ i = fft ( piw ) , piw, n = 0.5⎢1 − cos⎜⎜ ⎟⎥ pi , n (Eq. 1)
⎟
⎢⎣ ⎝ nb ⎠⎥⎦
Synthesizing the sound field we assume that the loudspeakers induce plane waves in the
central listening position and that time delay and amplification are the same for all
loudspeakers. These conditions are met fairly well as we use an anechoic projection room and
the speakers are furnished with FIR filters and placed on a circle. However, in a later version
the plane-wave assumption will be replaced by using measured impulse responses. So far, the
Fourier coefficients qˆ j , k of the signal of the j-th speaker are computed to satisfy:
nl ⎛ ω ⎞
pˆ i , k = ∑ qˆ j , k exp⎜⎜ − i k v j , xi ⎟
⎟ (Eq. 2)
j =1 ⎝ c0 ⎠
ωk = π f s k / nb is the k-th angular frequency, f s the sampling rate, and v j the radiation direction
of the j-th speaker. This leads to a linear system for each frequency:
Ak qˆ., k = pˆ ., k (Eq. 3)
For stability reasons the number n p of reproduction points is chosen greater than the number
nl of speakers and (Eq. 3) becomes overdetermined and is solved in a least-squares sense.
Moreover, in order to keep speaker signals bounded we introduce a small regularization
parameter α :
qˆ., k = Tk pˆ ., k , [
Tk = Ak* Ak + αI ]
−1
Ak* (Eq. 4)
The matrices Tk are computed once for each frequency when initializing the system. As the
speaker signals will be real we have qˆ., k = qˆ.,*nb − k and only half of the systems need to be
solved. Next, the q̂ j are retransformed into time domain by the inverse Fourier transform to get
q wj . The superscript w indicates that these speaker signals will reproduce only the windowed
measurements. To get the final speaker signals, the first half of q wj is added to the last half of
the q wj computed for the last block:
q j = q wj ,1, b...n + q wj ,,nb −+11...2 n (Eq. 5)
b b b

Finally, q wj ,,nb +1...2 n is stored for adding in the next step.

b b

19th INTERNATIONAL CONGRESS ON ACOUSTICS – ICA2007MADRID

VISUALIZATION
In this section we give a brief overview of the visualization approaches of the phonon tracing
results. The phonon map characterizes the acoustic behaviour of a scene considering the
location of a specific sound source. It consists of the reverberations of a unit pulse, coming from
different directions with different time delays and specific pressure and energy distributions.
First we visualize the spatial propagation of sound waves from the source [4]. The
corresponding wave front traverses the room and is reflected on surfaces, altering its intensity
and energy spectrum. We visualize these sound waves by rendering small spheres
representing the phonons. The spheres are colored by means of their spectral energy.
Therefore, we use the RGB components, such that blue corresponds to the average of the
energy by 40, 80, 160, 320 Hz, green corresponds to the average of energy by 640, 1280, 2560
Hz, and red to the average by 5120, 10240, 20480 Hz. When sliding through time, the spheres
follow the simulated phonon paths. To improve the visualization of wave fronts, we use
polygonal surfaces with phonons as vertices. Fig. 4 (a+b) shows an example of the described
visualization approach.
In another work [7] we present different visualizations of the phonon map. The first three
techniques visualize the wave fronts on the scene surfaces independent of listener position.
First of all we rendered phonons as color coded spheres at their positions on reflecting surfaces
(see Fig. 4 (d)). Furthermore, we visualize the outgoing direction of different particles by use of
cones. In the second method we visualized the sound wave fronts reflecting from different
materials by use of triangulated surfaces (see Fig. 4 (e)), which are deformed according to the
traversed distance of the phonons contributing to this wave front. In the third visualization
method we represented the energy distribution on a given scene surface by means of scattered
data interpolation (Fig. 4 (c)). All three methods provide the option to render the energy
considering the overall frequency spectrum or the energy of only one selected frequency band.
Energy is color coded using the RGB color space in the first case and the HSV color space in
the second case. Additionally to the methods described above, we visualized the energy
received at a given listener position (Fig. 4 (f)). Therefore we render a color coded sphere at the
listener position deformed according to the direction and amount of the arrived energy. These
visualization approaches allow the visual representation of the phonon map on the scene
surfaces as well as the representation of the energy received by a listener.

(a) (b) (c)

(d) (e) (f)

Fig. 4. Visualization of the simulation results calculated using the phonon tracing. (a)Sound
particles (phonons) rendered as small colored spheres. (b)Representation of wave fronts
by use of triangulated surfaces. (c)Visualization of the energy on the floor at 20 msec for
the 160 Hz band using scattered data interpolation. (d)Visualization of particular phonons.
(e)Clustered wave fronts. (f)Deformed spheres representation for 1280 Hz band.

19th INTERNATIONAL CONGRESS ON ACOUSTICS – ICA2007MADRID

 CONCLUSIONS
We presented an audio visual Virtual Reality system for room acoustics. For acoustic simulation
purposes we combined a geometric approach the Phonon Tracing and FEM solver. The
auralization is performed over professional sound equipment using our synthesis approach.
Furthermore we introduced different visualization techniques of the simulation results (the
phonon map) on a stereoscopic display including visualization of individual wave fronts and
listener based visualization. The system enables the exploration of acoustical behaviour inside
a room visual as well as aural allowing the feasibility to improve the acoustics of the room
during the design process. Future work includes the improvement of human interaction
feasibilities with the system. Another interesting aspect is an automatic optimization of room
geometry and/or surface materials. By means of well defined acoustic metrics a room can be
evaluated if it is suitable for speech or music presentations. The challenge is to find an
appropriate formal description which concludes how the room shape affects the given metric, to
be able to start optimization algorithms for room improvement.

ACKNOWLEDGEMENTS
This work was supported by the Rheinland Pfalz Cluster of Excellence on Dependable and
Adaptive Systems and Mathematical Modeling (DASMOD).

References:
1. Allen JB, Berkeley A (1979) Image method for efficiently simulating small-room acoustics. In: J. Acoust. So. Amer.,
65(4):943.950.
2. Antoulas A, Sorensen D (2001) Approximation of large-scale dynamical systems: An overview. Technical report,
Rice University.
3. Assenmacher I, Kuhlen, T, Lentz T, Vorländer M (2004) Integrating real-time binaural acoustics into VR
applications. In: Eurographics Symposium on Virtual Environments, Grenoble 129-136
4. Bertram M, Deines E, Mohring J, Jegorovs J, Hagen H (2005) Phonon tracing for auralization and visualization of
sound. In: IEEE Visualization, Minneapolis, MN.
5. Berkhout A, de Vries D, Vogel P (1993) Acoustic control by wave field synthesis. J. Acoust. Soc. Am. 93(5) 2764-
2778
6. Cook PR, Essl G, Tzanetakis G, Trueman D (1998). In: International Conference on Auditory Display (ICAD98),
Glasgow, UK.
7. Deines E, Michel F, Bertram M, Hagen H, Nielson GM (2006) Visualizing the phonon map. In: Proceedings of
EuroVis, pages 291–298, Lisboa.
8. Deines E, Bertram M, Mohring J, Jegorovs J, Michel F, Hagen H, Nielson GM (2006) Comparative visualization for
wave-based and geometric acoustics. In: Transactions on Visualization and Computer Graphics (TVCG),
Proceedings of IEEE Visualization, volume 12, pp 1173–1180.
9. Funkhouser TA, Carlbom I, Elko G, Pingali G, Sondhi M, West J (1998) A beam tracing approach to acoustic
modeling for interactive virtual environments. In: Computer Graphics (SIGGRAPH 98), pp 21-32, Orlando, FL.
10. Funkhouser TA, Min P, Carlbom I (1999) Real-time acoustic modeling for distributed virtual environments. In
Computer Graphics (SIGGRAPH 99), pp 365-374, Los Angeles, CA.
11. Hackbusch W (1989) Integralgleichungen, Theorie und Numerik. Teubner Verlag.
12. Jensen HW (1996) Global illumination using photon maps. In: Rendering Techniques ’96 (Proceedings of the 7th
Eurographics Workshop on Rendering), pp 21–30.
13. Kapralos B, Jenkin M, Millios E (2004) Sonel mapping: Acoustic modeling utilizing an acoustic version of photon
mapping. In: IEEE International Workshop on Haptics Audio Visual Environments and their Applications (HAVE
2004), Ottawa, Canada.
14. Khoury S, Freed A, Wessel D (1998) Volumetric visualization of acoustic fields in cnmat’s sound spatialization
theatre. In: IEEE Visualization ’98, pp 439–442 & 562.
15. Klehs B, Sporer T (2003) Wave field synthesis in the real world part 1: In the living room. In: 114th AES Convention,
Amsterdam
16. Krockstadt U (1968) Calculating the acoustical room response by the use of a ray tracing technique. In: Journal of
Sound and Vibrations, 8(18).
17. Kulowski U (1984) Algorithmic representation of the ray tracing technique. In: Applied Acoustics, 18:449-469.
18. Lentz T, Behler G (2004) Dynamic cross-talk cancellation for binaural synthesis in virtual reality environments. In
Proceedings of the 117th Audio Engineering Society Convention, San Francisco, USA.
19. Lentz T, Schröder D, Vorländer M, Assenmacher I (2007) Virtual Reality system with integrated sound field
simulation and reproduction. In: EURASIP Journal on Advances in Signal Processing, Volume 2007.
20. Lokki T, Nenonen V (2006) Immersive visualization of room acoustics. In: Joint Baltic-Nordic Acoustics Meeting,
Gothenburg, Sweden.
21. Monks M, Oh BM, Dorsey J (1996) Acoustic simulation and visualization using a new unified beam tracing and
image source approach. In: Convention of the Audio Engineering Society. ACM.
22. Naef M, Staadt O, Gross M (2002) Spatialized audio rendering for immersive virtual environments. In: Symposium
on Virtual Reality Software and Technology, ACM, pp 65-72
23. Savioja, L, Lokki, T, Huopaniemi, J (2002) Auralization applying the parametric room acoustic modelling technique -
the diva auralization system. In: International Conference on Computer Auditory Display, Kyoto, Japan, pp 219-224
24. Stettner A, Greenberg D (1989) Computer graphics visualization for acoustic simulation. In: International Conference
on Computer Graphics and Interactive Techniques, pp 195–206.
25. Vorländer M (1989) Simulation of the transient and steady-state sound propagation in rooms using a new combined
ray-tracing/image-source algorithm. In: J. Acoust. So. Amer., 86(1):172-178.

19th INTERNATIONAL CONGRESS ON ACOUSTICS – ICA2007MADRID

View publication stats