THE RECURRENT ALGORITHM FOR INTERFEROMETRIC SIGNALS PROCESSING BASED ON MULTI-CLOUD PREDICTION MODEL

M. A. Volynsky, I. P. Gurov, P. S. Skakov


Read the full article 

Abstract

The paper deals with modification of the recurrent processing algorithm for discrete sequence of interferometric signal samples. The algorithm is based on subsequent reference signal prediction at specifying a set (“cloud”) of values for signal parameters vector by Monte Carlo method, comparison with the measured signal value and usage of the residual for enhancing the values of signal parameters at each discretization step. The concept of multi-cloud prediction model is used in the proposed modified algorithm. A set of normally distributed clouds is created with expectation values selected on the base of criterion of minimum residual between prediction and observation values. Experimental testing of the proposed method applied to estimation of fringe initial phase in the phase shifting interferometry has been conducted. The estimate variance of the signal reconstructed according to estimated initial phase from initial signal does not exceed 2% of the maximum signal value. It has been shown that the proposed algorithm application makes it possible to avoid the 2π-ambiguity and ensure sustainable recovery of interference fringes phase of a complicated type without involving a priori information about interference fringe phase distribution. The usage of the proposed algorithm applied to estimation of interferometric signals parameters gives the possibility for improving the filter stability with respect to influence of random noise and decreasing requirements for accuracy of a priori filtration parameters setting as compared with conventional (single-cloud) implementation of the sequential Monte Carlo method.


Keywords: interferometric signals analysis, Bayesian filter, sequential Monte Carlo method, multi-cloud prediction
Copyright 2001-2017 ©
Scientific and Technical Journal
of Information Technologies, Mechanics and Optics.
All rights reserved.

Яндекс.Метрика