OPTIMIZATION OF AN ADAPTIVE KALMAN FILTER BASED ON INFORMATION-THEORY

An adaptive Kalman filter is described which can be optimized, even if the location of the required overlap-free region is not known a priori. The optimization is based on maximizing the information yield from the filter procedure, and the results with the highest precision and accuracy are found when the information is at a maximum. The information-based optimization is also used to demonstrate that more precise filtering is accomplished with higher point densities, and that an optimal filter window exists, depending on the spectral resolution and the sampling frequency. Simulations based on a grid search and on simplex optimization give similar results.