Automatic Lip-Synchronization Using Linear Prediction of
Download
Report
Transcript Automatic Lip-Synchronization Using Linear Prediction of
Human Muscle Modeling using Generalized
Cylinders for Volume Considerationss
SK Semwal
Bill Watson
Debra McCullough
University of Colorado, Colorado Springs
Topics of Presentation
Introduction and Background
Generalized Cylinders
Volume Considerations
Results
Conclusions
Introduction
Need:
Long standing research problem
Generalized cylinders: simple and intuitive
Volume considerations for muscles
Intersection with adjacent muscles/bones lead
to suitable deformations
Previous Work
Since 1968
Chen-Zeltzer - Biomechanical
Badler’s work – human body
Nadia and Daniel Thalmann – human body
Semwal and Dow’s GC Muscle models
Generalized Cylinders
Shani and Ballard
Set of cross sections
Set of generalized axis
Dow and Semwal
Model upper and lower arm using GCs
Extensions
Leg Musles
Polygons. NURBS, Shades choices
Animation sequences – leg exercises
Tension on the muscles
Speech recognition front-end
Models contraction/deformations using
volume
Models
Femur or thigh bone – longest and
heaviest bone – hip to tibia
Tibia or shin bone – next heavy bone
transfers the weight to ankles from hip
Fibula – parallel to tibia on the outside
lateral – attached to several muscles –
acts a pulley to tendons behind ankle
Generalized Cylinders
Model 2D planar contours from Medical Books
(Tortara, Gardner and Cated
Define these 2D planar contours along GC axis
NURBS defined using n points on contours
Rendered on SGI/OpenGL code
Intersection Testing and
Intersection Resolution
Use cross sections
Move the intruding point away from the adjacent
muscle/bones polygonal area between contours
Volume
Two cross sections Aavg = (Ai + Ai+1)/2
Distance between the two GC-axis point for the two
cross sections
Volume between two cross sections = Aavg * d
Repeat for all cross sections pair for that GC
Deformation
pct_chg = (curr_vol - init_vol) / init_vol
rel_chg = (cum_sum (curr_vol init_vol))/cum_sum
rel_change acts as a guide based upon
tolerance in changing the cross section
points
Points next to bone and other muscle not
modified
Results
Precise timing can be
achieve
Smoothing introduces
“lag”
Results
Summary
GC model provided a good method for
modeling bones and muscles
Volume considerations allow good
deformation effects
Biomechanical analysis and animation
Future Work
Model animations
Realistic biomechanical based rendering
Automatic detection from CT data and
creating GCs
End