We describe a method for calibrating an electromagnetic motion tracking device. Algorithms for correcting both location and orientation data are presented. In particular, we use a method for interpolating rotation corrections that has not previously been used in this context. This method, unlike previous methods, is rooted in the geometry of the space of rotations. This interpolation method is used in conjunction with Delaunay tetrahedralization to enable correction based on scattered data samples. We present measurements that support the assumption that neither location nor orientation errors are dependent on sensor orientation. We give results showing large improvements in both location and orientation errors. The methods are shown to impose a minimal computational burden.