Forward-backward cost-reference particle filtering-based instantaneous frequency curve estimation method

Abstract

The invention discloses a forward-backward cost-reference particle filtering-based instantaneous frequency curve estimation method. The method comprises the steps of firstly, constructing a state equation and an observation equation for non-linear frequency modulation signals; secondly, defining particle costs by a user according to the state equation and the observation equation; thirdly, generating initialized particles, and figuring out particle risks based on the particle costs; fourthly, calculating the re-sampling weight according to the particle risks; fifthly, re-sampling according to the re-sampling weight and iteratively updating the particle costs to obtain particle-cost sets at L moments, and obtaining the forward-state estimation with a minimum cost as the estimation rule; sixthly, constructing a state equation and an observation equation for the signals of a backward dynamic system; seventhly, inputting observed data in an inverse sequence into the backward dynamic system to obtain the estimation of a minimum cost state (imgfile='DDA 0000754751940000011. tif'wi =78'he ='66' / ); eighthly, obtaining the estimation of an instantaneous frequency curve based on the estimation of the minimum cost state (imgfile='DDA 0000754751940000011. tif'wi =47'he='71'. According to the technical scheme of the invention, the stability of the signal state estimation is improved, and the estimation error is reduced. Therefore, the method can be used for the target state estimation in a non-linear dynamic system.

Description

technical field [0001] The invention belongs to the technical field of signal processing, and in particular relates to a method for estimating instantaneous frequency curves based on forward-backward cost reference particle filtering, which can be used for target state estimation in nonlinear dynamic systems. Background technique [0002] In the field of radar and sonar, the detection of weak maneuvering targets in noise or clutter is always a challenging problem. Due to the difference of target motion state and radar working mode, the target echo model can be divided into two categories, namely parametric model or non-parametric model. When the observation time is short and the target motion is stable, the echo signal can be simulated by a parameter model containing a small number of unknown parameters. An example is a chirp signal with unknown amplitude, initial phase, and Doppler shift. When the observation time is long and the target motion is complex, the amplitude an...

AI Technical Summary

Problems solved by technology

However, most particle filters and their improved algorithms require the statistical characteristics of the dynamic system to be known, which is difficult to meet in practical applications.

For example, the observation noise in the maneuvering target detection problem in the residual clutter and noise background, and the system noise in the dynamic system describing the instantaneous frequency curve of the unknown FM signal, will cause the statistical properties of the dynamic system to be unknown, which makes the traditional particle The filtering method will produce a large estimation error or even fail

Images