University of Houston

Control mechanisms to drive realistic and accurate real-time character motions are difficult to design. The most prevalent methods typically use button and joystick interfaces to control character movement. Systems that allow for more detailed motion control are typically expensive and complex making them unsuitable for home use. This paper explores character control through the use of accelerometers, devices that when moved in 3D space measure their acceleration vectors. Motion reconstruction is done by creating a database of poses that have been preprocessed to include synthesized acceleration data for each control joint. At runtime, the K nearest poses in the database are matched using a query metric based on the real-time user driven human accelerometer feed and the previous pose. These K nearest poses are then used to reconstruct the motion of the user. By developing a system based entirely on passive sensors, the user is no longer constrained by external sensor devices, such as motion capture camera systems.