Development of a magnetic measurement system for mandibular movement using neural networks

A high performance measurement system using neural networks for mandibular movement is proposed, and its applicability is examined. The problem of estimating position and orientation of a magnet from magnetic field is a kind of inverse problem. It is difficult to design high speed and accurate measurement systems using traditional methods for solving this problem. The proposed system makes use of backpropagation neural networks to estimate the position and direction from the magnetic field. According to computer simulation, average errors were 7.2 μm for position and 0.002 degree for orientation. Processing time was 4.2 ms/sample. Furthermore, availability was confirmed by simplified two dimensional model experimentally. These results show the proposed system is applicable to practical measurement of a mandible, such as range of motion and conditions of occlusion