Real-Time Tracking and Display of Human Limb Segment Motions Using Sourceless Sensors and a . . .

Maximum 200 words.) Recently, a new impetus has been given to work to produce a whole body human/computer interface system, which encumbers the wearer to a minimum degree and operates over long distances. The input portion of such an interface can be derived by appropriate processing of signals from a nine-axis sensor package consisting of a three-axis angular rate sensor, a three-axis magnetometer, and a three-axis linear accelerometer. This paper focuses on data processing algorithms for MARG sensors. Since human limb segments are capable of arbitrary motion, Euler angles do not provide an appropriate means for specifying orientation. Instead, quaternions are used for this purpose. Since a quaternion is a four-dimensional vector, and a MARG sensor produces nine signals, this data processing problem is "overspecified". This fact can be used to discriminate against sensor noise and to reduce the effects of linear acceleration on measurement of the gravity vector by accelerometers. Detailed computer simulation studies and the results of preliminary experiments with a prototype body tracking system confirm the effectiveness of the quaternion body-tracking filter