The goal of this research is to develop a learning model by which a humanoid robot can learn a generalized form of a given gesture, such that the robot is able to stochastically generate its own novel instances of the gesture while still ensuring that the generated gesture is recognizable.