Synthetic brain imaging: grasping, mirror neurons and imitation
Neural Networks - Special issue on the global brain: imaging and modelling
Synthesis of complex dynamic character motion from simple animations
Proceedings of the 29th annual conference on Computer graphics and interactive techniques
Synthesizing animations of human manipulation tasks
ACM SIGGRAPH 2004 Papers
Online knowledge acquisition and general problem solving in a real world by humanoid robots
ICANN'10 Proceedings of the 20th international conference on Artificial neural networks: Part III
| Hi-index | 0.00 |
We propose a new method for interpolation and extrapolation of motion patterns in proto-symbol spaces. The proto-symbol space is a topological space which abstracts motion patterns by utilizing continuous hidden Markov models, and the mimesis model that recognizes/generates known/unknown motion patterns by using this topological space. An interpolation algorithm for the proto-symbol space has already been proposed, but it had a mathematical problem. Furthermore, extrapolation of motion patterns was not defined, and could not be calculated. In the new method, the synthesis of proto-symbols is done separately for state transition probabilities and output probabilities, and the synthesis of the state transition probabilities is done in the time domain. Experiments in a simulation environment demonstrate the feasibility of this method.