MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE Motion Pattern Preserving IK Operating in the Motion Principal Coefficients Space Schubert R. Carvalho, Ronan Boulic and Daniel Thalmann Presented by Schubert R. Carvalho MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE Overview • • • • • Introduction Related Work A Model for Editing Experiments Conclusions and Future Work WSCG 2007 MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE Introduction 1 • Challenges of Motion Editing Techniques – Preserve the characteristics of the original animation – Impose continuity between frames • Motivation – How to edit a motion by constraining just one pose without adding discontinuities? 2 WSCG 2007 MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE Related work • Traditional Motion Editing • [Lee et. al. 1999] • [Gleicher 2001] • [Kulpa 2005] • [Le Callennec and Boulic 2006] • [Liu et. al. 2006] • Principal Component Analysis in Computer Animation • [Alexa and Müller 2000] • [Safanova et. al. 2004] • [Glardon et. al. 2004] • [Urtasun and Fua 2004] WSCG 2007 MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE A Model for Editing • Three basic issues – Motion Parameterization – Constraining a Single Posture in the Motion PC Space – Impose Continuity WSCG 2007 WSCG 2007 MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE Motion Parameterization • Principal Components Analysis (PCA) Ψ1, Ψ2 , ..., ΨM Pose ΨM = [Θt , ..., Θt , ..., Θt ] 0 Motion Database k PCA 1 PC PCs 0≤ k ≤ 1 Θt = [Φ1, P1, Φj, ..., Φn] k root Joints angles m ΨD ≈ Ψo + ∑αDiEi I=1 m≤M MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE WSCG 2007 Constraining Formulation • Constraining a Single Pose in the Motion PC Space Δα = J(α)tεΔx p=3; m=8; n=93; D=12276 WSCG 2007 MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE Imposing Continuity • Continuity is enforced through the PCA parameters. Algorithm’s Convergence While not converged do k ε 1: Δα J(α)t Δx 2: αD 3: ΘD αD + ΔαD Ψot + αDEtk k end while 4: ΨD PC (Et ) m Ψo + ∑ αDi Ei I=1 Time (s) Nb. Of Iterations 2 40000 19196 3 110000 50214 4 42000 19929 5 9000 4194 6 1625 757 7 0.500 235 8 0.484 221 9 0.531 247 WSCG 2007 MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE Experiments: Editing and Style (1) 1 2 Input Motion 0 2 -4 -3 -2 -1 1 3 4 5 -5 2nd Component 3 4 5 Golf Swing Database -5 0 5 10 15 20 25 1st Component 30 35 40 MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE WSCG 2007 Experiments: Editing and Style (2) damping = 10; num. of iterations = 150 tracked error = 0.01m. MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE Experiments: Visual Comparison (1) f = 93 f = 94 WSCG 2007 WSCG 2007 MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE Experiments: Visual Comparison (2) PIK N. of Constraints Fr. Constrained Presented Artifacts Computation Time 5 11 Our Approach 1 1 Feet Sliding, Balance and Arms’ Inter Penetration Nothing Detected 1 sec. 100 sec. MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE WSCG 2007 Conclusions • If the motion needs just an adjustment of one key event, our approach can provide faster and more realistic results than current approaches. • Our approach provided quality results with less tuning from the user side. Future work • Extend the current method to treat multiple constraints. • Extend the current method to constrain multiple frames (if necessary). MOTION PATTERN PRESERVING IK OPERATING IN THE MOTION PRINCIPAL COEFFICIENTS SPACE Thanks four your attention! WSCG 2007