Stepwise fuzzy correction of the algorithm filters of random signals

• Main Authors: A. A. Lobaty, A. S. Radkevich
• Format: Article
• Language: English
• Published: Belarusian National Technical University 2019-07-01
• Series: Sistemnyj Analiz i Prikladnaâ Informatika
• Subjects: discrete process; estimation; probability density; membership function; fuzzy logic
• Online Access: https://sapi.bntu.by/jour/article/view/252

The task of estimating the information contained in random signals from various sources – meters. It is assumed that the gauges are discrete and are described, like the original process assessed, by a discrete mathematical model in the form of difference equations. As an estimation algorithm, we consider a discrete Kalman filter, which, in the general case, when mathematical models are inadequate to real processes, can give distorted information. To improve the accuracy of estimation, it is proposed to apply the integration of all possible meters with the introduction of additional a priori information using a fuzzy logic system. At the same time, it is proposed to make a transition from the obtained probability characteristics of the estimated process to the membership functions of fuzzy logic based on the output filter parameters using the normalization of the posterior probability density. This approach allows to increase the accuracy of estimation, as it takes into account additional information and its complex processing.