Motion Abstraction and Mapping with Spatial Constraints

A new technique is introduced to abstract and edit motion capture data with spatial constraints. Spatial proximities of end-e ectors with tagged objects during zero-crossings in acceleration space are used to isolate signi cant events and abstract constraints from an agent's action. The abstracted data is edited and applied to another agent of a di erent anthropometric size and a similar action is executed while maintaining the constraints. This technique is speci cally useful for actions involving interactions of a human agent with itself and other objects.