HUMAN FIGURE ANIMATION

1
HUMAN FIGURE ANIMATION

• Norman I. Badler
• Center for Human Modeling and Simulation
• University of Pennsylvania
• Philadelphia, PA 19104-6389 USA
• http//www.cis.upenn.edu/badler

2
Outline

• Generative Methods and Tradeoffs
• Re-use and Retargetting
• Parametric Models and Controls
• Real-time or Offline? (i.e., Hard problems)
• Model Availability
• Body Requirements and Assumptions
• Object Virtual Human Interactions

3
The Realism Ceiling for Human Models
Time to create Behavioral Realism
Real-time
Visual Realism
Inanimate objects
Life-like virtual humans
4
What Approaches Push the Curve Toward More
Realism?
Motion Capture
Time to create Behavioral Realism
Parameterization (face, gait, gesture)
Real-time
Visual Realism
Inanimate objects
Life-like virtual humans
5
Why is Realism Still Hard?
Motion Capture
Time to create Behavioral Realism
What are the right parameters?
Difficult to generalize
Parameterization (face, gait, gesture)
Real-time
Visual Realism
Inanimate objects
Life-like virtual humans
6
Trade-Offs Advantages
Disadvantages
Procedures Use models Generalize control Fast Find models Specific code Integration w/others
Motion Capture Natural Nuance Physics Actors Acting Motion specificity Retargetting
Dynamics (Physics-based simulation) Physics Apparent interlimb coordination Free fall (rag doll) Control laws Model complexity Retargetting Collision response
7
Human Movement Categories

• Voluntary (task-oriented, reach, look-at)
• Involuntary (breathing, balancing, blinking)
• Subconscious
• Low level motor functions (fingers, legs, lips)
• Communicative acts (facial expressions, limb
  gestures, body posture)
• What generative techniques work for these?

8
Human Movement Implementations

• Voluntary (procedures IK, mocap)
• Involuntary (procedures physiological or
  dynamic models)
• Low level motor functions (procedures, mocap,
  dynamics)
• Communicative acts (mocap, procedures)
• category X technique animation paper

9
MoCap Retargetting

• Joint angles dont neatly map to different size
  or structured figures.
• Need to maintain constraints (IK) to avoid feet
  slipping and sliding.
• Need to add IK to change reach and look-at
  direction.
• Moving interactions may fit but change dynamics
  (e.g., figure skating pairs).

10
Parametric Models (Procedures)

• IK
• Gait
• Faces
• Eyes
• Muscles (Deformations)
• Gestures (BEAT/MagiCster)
• EMOTE

11
IK at Work
Eye view Note attention Control
Instructions translated to parameterized actions.
12
Walking (Sun and Metaxas)

• Gait from parameterized biomechanical model,
  original movements from motion capture and
  interpolated based on terrain angle.
• movie

13
Gesture Control EMOTE

• EMOTE A real-time motion modification system.
• Defines qualities of movement with 8 parameters.
• Based on Effort and Shape components of Laban
  Movement Analysis.

14
Inputs EMOTE Output
4 Efforts 4 Shapes
Key Poses
End Effector Goals
Frame Rate Poses
Procedures
Inverse Kinematics Interpolation
Motion Capture
15
Motion Qualities are Important

• Movements with EMOTE qualities give insight into
  the agents cognitive state.
• When EMOTE qualities spread from limbs to body,
  movements appear more sincere.

16
Dont Animators Know This?

• Good key pose animators know this and use
  character deformation to great advantage
• EMOTE allows this concept to be applied
  parametrically (under program or interactive
  control) to articulated (jointed) figures.

17
An Actor Trying to Act (Apologies to W.
Shakespeare)
Well manually match these movements in our
virtual human model, then vary the motion
qualities via EMOTE.
18
Manually Matching the Motions
19
The Actor with Neutral Efforts (A
Politician?)
20
Actor with Less Rising Shape (Not quite as
excited?)
21
Moving the Shapes Inward (Woody Allen?)
22
With Light and Sustained Efforts
(More solemn and serious?)
23
Without Torso Movements (Used Car
Salesman?)
24
Manner Variants (adverbs) HIT (with real dynamic
response)
Hit the ball
softly.
forcefully.
25
DI-GuyTM Being combined with EMOTE (mocap
parameters)
26
Procedural Eye Movement Model from Human
Performance Data
Source Eyes fixed ahead Eyes moved by
statistical model Full MPEG-4 face
27
Important Problems to Solve in Real-Time

• Constrained reach planning/execution
• Coordination and consistency
• Dynamics on demand (integrated with
  constraints/goals)
• Individuation (and its specification)
• Collision response
• Deformable surface
• Clothing

28
E.g. Real-Time Constrained Reach Planning
29
Model Availability

• Standard H-Anim (seamless) avatars (VRML /
  WEB3D / MPEG-4)
• DI-Guy (Joint angle control)
• Jack (Toolkit API plug-ins)
• Application-specific models (Extempo, Ananova,
  )
• Custom models in Maya, 3DSMax,
• Issues Support, Platform, API, Body Swap

30
Body Requirements and Assumptions

• Enough joints, correct placement, DOFs, limits
• Skin surface shape (deformations)
• Torso shape and flexibility
• Physiology breathing, itching, eating,
• Rendering skin, hair, moisture, vascular
  changes
• Human or caricature?
• Resemblance to someone (or not)
• Clothing

31
Object Human Interactions Challenges

• Grasping (one / two hands, or with other body
  parts)
• Multi-person actions (carrying furniture)
• Smart objects Compliant motion (constraints
  applied to existing motions)
• Knowledge of how things operate, fit together,
  disassemble

32
The Agent needs to Know
Action choices cause undesirable effects
Actions changed to prevent errors
33
Questions!