A neural measurement system for a moving object using magnetic sensors

Measurement using magnetic fields is one of the most useful methods to gauge the movement of a living body etc. Estimation of the position and direction of a magnet attached to a object from flux density distribution around it is an inverse problem. Though analytical methods are used to solve these problems, they need a lot of calculations to get a convergent solution. In this paper, the authors apply the back propagation neural networks to solve this inverse problem, and their applicability and accuracy are examined. As a result of computer simulations, we obtain an accuracy reading of 0.91% for position error and 0.19/spl deg/ as an average value.<<ETX>>