Study Of Measurement Noise And State Fusion Estimation Algorithms For Multi-Channel System With Multiplicative Noises

One of the important problem in stochastic signal processing is optimal estimation of signal under certain meanings, or obtaining useful information from signals polluted by noises. The optimal estimation algorithm of measurement noise and the optimal state fusion algorithm for multi-channel system with multiplicative noises are mainly researched in this dissertation. This work is supported by NSFS for Tianyuan Mathematical Fund Grant #A03 24676 and the Ministry of Education of China Grant #02131.At present the study of multi-channel system with multiplicative noises is not perfect. In the field of noise estimation, previous studys are only limited at dynamic noise estimation, but the measurement noise estimation has also important theory value and practical significance. This dissertation focuses on the optimal estimation of measurement noise under the condition that the multiplicative noise is in the form of a general stochastic matrix, which means a genuine multi-channel system. Furthermore, with the rapid development of computer technology there has been a considerable interest in the multi-sensor system because of its high accuracy and high reliability. Accordingly, the multi-sensor information fusion technique has received deep researches and has been used in military systems, oil seismic exploration and underwater remote targets detection, etc. Based on the former works, this dissertation also discusses the optimal fusion estimation algorithms.The main researching contents of this dissertation are as follows:1. In this dissertation an optimal filtering algorithm and an optimal smoothing algorithm of measurement noise are developed for multi-channel system with multiplicative noises under the conditions that the additive noises are independent white noises and the dynamic noise correlates itself in one-step and correlates with the measurement noise at the present step as well as one past step. These algorithms are optimal in the sense of linear minimum-variance.2. For the multi-channel system with multiplicative noises under multi-sensor observation an optimal fusion estimation algorithm is proposed on the conditions that the multiplicative noise is in the form of a general stochastic matrix and the measurement noises of each sensor are correlated with the dynamic noise. This expands the application of multi-sensor systems.3. The above algorithms are tested by many simulations, which show the validity of these algorithms.