Abstract
In this paper, a new method of finding the sequential state estimation algorithm is established as one of the techniques for overcoming the difficulty owing to the observation noise (or the background noise) which inevitably appears in the actual observation. It is derived by generalizing the well-known Kalman Filter under consideration of the proper characteristics appearing especially in the field of environmental noise and vibration. Its form of algorithm can be artificially preestablished form the viewpoint of practical usefulness. More concretely, the Bayes Theorem is expressed in its expansion form, matched with the preestablished algorithm form without losing the information on higher order moments, and a method of putting the information of sequentially observed data into this preestablished algorithm form is proposed. Furthermore, the theoretical legitimacy of this new method is established by proving that this algorithm and the Kalman Filter agree perfectly with each other in a special case. Then the methodological effectiveness is confirmed by applying it to two kinds of typical estimation problems for actual architectual acoustics.
