Physically Based Sound: Progress Report

1
Physically Based SoundProgress Report

• COMP259 Nikunj Raghuvanshi

2
Two Existing Approaches

• FEM deformable simulationO'Brien, J. F. et. al.,
  Synthesizing Sounds from Physically Based
  Motion. SIGGRAPH 2001.
  
• FoleyAutomatic (Modal Synthesis)Kees van den
  Doel et. Al., FoleyAutomatic Physically-based
  Sound Effects for Interactive Simulation and
  Animation. SIGGRAPH 2001.

3
A bit of detail
FEM Simulation Tetrahedral Mesh
FoleyAutomatic Modal Synthesis
4
Issues

• FEM Simulation Accurate, general, but very slow
• Modal Synthesis Fast, Good quality sound
• But extracting modes of general objects
  difficult
  
• Need measurements on actual physical object

5
Proposed Approach

• FEM is too costly. Use a simpler model.
• Particle Systems
• All PDEs become ODEs.
• Elasticity of Material maps to K
• K108 what about stability?
• Analytical solution!

6
A bit of math
K is a constant matrix
After the math, sound sample at time t, s(t) is
Sum of damped sinusoids, just like Modal
Synthesis! Gains take into account the listen
er position etc.
7
Efficiency

• Exponentiation of K has to be done only once for
  a given model, O(n3).
  
• Finding ak for a given scene, O(n2)(?)
• Generating sound once ak are known, O(n)
  Runtime Compexity.
  
• The constant is calculating the sine and
  exponential of a float
  

8
Preliminary Results

• 1D string
• n particles can capture n harmonics of a string!
• Lower harmonics accurate, higher inaccurate

9
Preliminary Results
1D string, 10 particles, K2108, B100, plucked
near center
10
Goals

• Semester Goal Harmonics of a 1D guitar string,
  with analysis of accuracy and efficiency of
  proposed method
  
• Similar things for a non-trivial shape in 2D for
  which the harmonics are known to compare against
  (for example, a disc)
  
• Long Term Goal Incorporating this technique into
  a simple rigid body simulator. The technique
  right now is quite limited and cannot handle
  rigid bodies.