Throwing Recognition Based on Magnetic Tracking and Trajectory Computation in an Augmented Reality System

In our augmented reality system, the throwing is taken as a main interactive method. So the system must have the ability of action recognition and quick response. The research analyzes the character of throwing, and finds out the law of its velocity changing, then uses convolution filtering to smooth dynamic data. Based on them, the fuzzy synthetic judgment algorithm can be used for recognizing the throwing action. Besides, object´s trajectory is affected by many factors, such as its shape, posture and material. Taking advantage of the principle of dynamics, their relationships are obtained. The experiments demonstrate the system can recognize various throwing actions, and compute the trajectory accurately.