“A biologically inspired latent space for gait parameterization” by Sothern, Gho, liu and Zhang

The problem of character locomotion synthesis is notorious for its high dimensionality and the nonlinear relationship between dimensions. However, many human motion activities lie intrinsically on low dimensional manifolds [Safonova et al. 2004] leading to significant data redundancy. Linear and non–linear methods for dimension reduction have been applied to the problem, but none of the existing approaches for dimensional reduction provide a physically–justified explanation for selected dimensions, instead they use general methods which employ numerical error analysis.

1. Chen, V. F. H. 2007. Passive dynamic walking with knees: A point foot model. Master’s thesis, Massachusetts Institute of Technology.
2. Full, R. J., and Koditschek, D. E. 1999. Templates and anchors: neuromechanical hypotheses of legged locomotion on land. Journal of Experimental Biology 202, 23, 3325–3332.
3. Matsuoka, K. 1985. Sustained oscillations generated by mutually inhibiting neurons with adaptation. Biological Cybernetics 52, 1, 345–353.
4. Safonova, A., Hodgins, J. K., and Pollard, N. S. 2004. Synthesizing physically realistic human motion in low-dimensional, behavior-specific spaces. In ACM SIGGRAPH 2004 Papers, ACM, New York, NY, USA, SIGGRAPH ’04, 514–521.