A Novel Magnetometer Calibration Approach with Artificial Data

This paper proposes two methods for calibrating triaxial magnetometers. Both of them calibrate these sensors with more general assumption of noise on three axes than previous state-of-the-art methods. The first method estimates bias and rotation parameters more accurately and the second method yields a better estimate for the scaling parameter than the state-of-the-art method subMLE. The computational time of the latter is also 43 times faster than subMLE, which allows this method to be applied in devices with low-computational resources (e.g. smartphones). Furthermore, the second method yields more robust heading angle estimates compared to subMLE. This result implies that the second method can be applied in light-weight inertial measurement systems, for which the orientation of the device is vital information for pedestrian dead reckoning system.