Estimating orientation with gyroscopes and accelerometers

Many systems for recording human movement need some reference from beacons near the subject, such as video cameras. Our goal is to measure human kinematics with sensors that are placed on the segments of interest. This way, experiments in which human movement is recorded are not restricted to a lab. Our inertial sensor-unit consists of a little box with three miniature gyroscopes (Murata ENC05E) and three linear accelerometers (AD xl05) that measure 3D angular velocity and linear acceleration, respectively. Both the gyroscope and accelerometer signals contain information about the orientation of the sensor. The sensor orientation can be obtained by integration of the angular velocity signals obtained from the gyroscopes [1]. This operation introduces drift in the estimated orientation. Accelerometers do not only measure the acceleration of the sensor, but also the gravitational vector. This gravitational component not only has a bigger magnitude for many human movements but also always points downwards. This knowledge can be used to make an estimation of the tilt. The tilt is the angle between the sensor axes and the vertical. This tilt estimation is not very precise but does not suffer from drift. The abstract describes a way to fuse both sensors (gyroscopes and accelerometers) to obtain an estimate of the orientation that is both accurate and is limited in integration drift.