Research Of Filtering Method Based On Characteristic Function

Filtering technology refers to the separation of specific band frequencies in a specific signal.It is an effective method for suppressing interference commonly used in modem industrial electronics,aerospace,and image processing.Therefore,it has received great attention from experts and scholars at home and abroad.In practical applications,these harsh constraints are met only in very few cases.In order to expand the application of filtering technology,scholars have studied the filtering algorithm for a class of nonlinear systems with Gaussian noise.The research on nonlinear system filtering methods for non-Gaussian noise is relatively limited.This paper studies the filtering methods of such complex systems.This paper mainly focuses on the extension of the filtering algorithm based on characteristie function.(1)Analyzing the limitation of the existing characteristic function filtering.In the case where the filtering algorithm based on multiple input multiple output is applied to multiple outputs,only a numerical approximation gain can be obtained,and there is an error due to local minimum.The original filtering algorithm is just suitable for the system whose state equation is linear and observation is nonlinear.In this paper,using Taylor expansion make the filtering is suitable for the system whose state equation is also nonlinear.(2)When the characteristic function filtering is applied to multi-dimensional observation system,the filter gain matrix is solved by the fixed points iteration.The polynomial fitting of the result is carried out.The convergence of the algorithm is analyzed from the numerical method,and compared with the previous algorithm,the effectiveness of the method is proved.Finally,the effectiveness and applicability of the proposed method are demonstrated by a measurement model simulation laboratory of an industrial device.(3)In order to obtain the analytical solution of the filter gain under multi-dimensional observation,two types of filters,parallel and sequential,are designed.The former needs to integrate the information of all the sensors,so the information is more reliable;the latter considers that there are time differences between reaching the fusion center due to the sensors being in different places,and the sensor may also have packet loss phenomenon.Sequential filtering is also considered.The problem of information inheritance is reached.