Simplification of Scene Geometry for Sound Propagation

1
Simplification of Scene Geometry for Sound
Propagation

• Sachin Patil

2
Existing Sound Propagation Approaches
Limitations

• Several methods available for sound propagation
• Too slow for real-time sound propagation taking
  into account the increasing scene complexity,
  dynamic environments and multiplicity of
  sources/listeners
• Acceleration techniques Culling structures like
  BSP-trees, SIMD hacks, GPU implementations are no
  longer sufficient
• Most VR systems do not allocate a large chunk of
  resources for handling sound
• Solve the sound propagation problem for
  simplified scenes
• Introduces the need for defining a metric for
  measuring acoustic fidelity

3
Commercial acoustic simulation software

• Glue Surface option provided by ODEON
  (www.odeon.dk) for manipulating CAD models for
  acoustic simulations

4
Its a jungle out there

• Taxonomy of Geometry Reduction techniques 1

5
Criterion for geometry reduction for acoustics

• Absorption coefficients for polygonal faced in
  the scene controlling error while blending and
  simplification
• Rough surfaces account for diffuse reflection and
  scattering
• Volume and total absorption area must be
  preserved as various parameters such as
  reverberation time heavily dependent on these
  criteria
• For handling reflections correctly, surface
  normals must be preserved as accurately as
  possible
• Features comparable to the wavelength of the
  sound must be preserved

6
Issues

• Tradeoffs between
• Cost of simplification and sound propagation
• Offline simplification versus real-time
  simplification
• Ease of implementation of the proposed approach
• Local reduction of features versus global
  simplification

7
Project goals

• Given a simplified scene, how different is the
  sound quality as compared to the original
• Could be statistical, acoustic metrics or
  perceptual user studies
• Planning to use an industry standard benchmark
  for acoustic measurements such as CATT, ODEON
• Given such a metric, how and by how much can one
  simplify a given scene geometry without
  perceptual differences in quality
• Progressive simplification and LOD Providing
  frame-rate guarantees given the resources
  available for sound simulation and propagation

8
Project roadmap and deliverables

• Use existing libraries to simplify scene geometry
  and carry out initial tests using CATT to
  determine an error metric for measuring acoustic
  fidelity
• Study existing methods for geometry
  simplification and choose an existing approach/
  propose a novel approach to solve this problem
• Integration with existing work Sound
  propagation using ray-frusta tracing / adaptive
  rectangular decomposition

9
References

• Geometry Reduction in Room Acoustics Modeling
  Samuel Siltanen
• Significant Facet Retrieval for Real-Time 3D
  Sound Rendering in Complex Virtual Environments
  C.Joslin, M.Thalmann
• Instant Sound Scattering N.Tsingos,
  C.Dachsbacher, S.Lefebvre, M.Dellepiane
• Interactions of Model Detail Level and Scattering
  Coefficients in Room Acoustic Computer
  Simulation L.Wang, J.Rathsam, J.Rindel