WO2002049230A1

WO2002049230A1 - Interference suppression method, in reception, of a radio signal transmitted in spread-spectrum band

Info

Publication number: WO2002049230A1
Application number: PCT/FR2001/003928
Authority: WO
Inventors: Franck Florin
Original assignee: Thales
Priority date: 2000-12-15
Filing date: 2001-12-11
Publication date: 2002-06-20
Also published as: AU2002217246A1; FR2818470A1

Abstract

In order to eliminate spurious signals in satellite radio navigation (GPS, GNSS1, GNSS2, GNSS3, GALILEO, GLONASS ) conventional techniques use frequency estimation of the sensed signal, estimation of interfering frequencies and elimination of said frequencies by filtering. The invention provides a different approach based on estimation of the time correlation matrix and on direct filtering of the signal without explicit estimation of interfering lines. Said technique offers the advantage of minimising adaptation time and the amount of calculation required. It proves to be well adapted to suppression of continuous wave (CW) and frequency modulated (FM) spurious signals. The direct filtering is advantageously carried out with a linear-phase adapted filter which minimises distortion products of the radio navigation signal.

Description

METHOD FOR DEWORING, RECEIVING, A RADIOELECTRIC SIGNAL BROADCAST EMITTED

The present invention relates to the deworming, on reception, of a radio signal transmitted in spread band by means of modulation by pseudo-random binary codes, such as the signals transmitted by the satellites of a GNSS radio navigation system. (acronym taken from the Anglo-Saxon "Global Navigation Satellite System"). These GNSS radionavigation systems are made up of three segments: a network of traveling satellites whose orbits and positions on their orbits are known at all times with precision and which emit radionavigation signals making it possible to estimate their distances and radial speeds with respect of a receiver evolving on the surface or around the terrestrial globe, of the ground stations ensuring the control of the network of satellites and of the receivers which receive the signals of radionavigation of the satellites and make it possible to deduce therefrom, by triangulation operations, the positions and velocity vectors of their carriers. The signals of radionavigation of the satellites which are emitted by on-board transmitters of low power and which have to travel long distances of the order of 20,000 Kms, are received with very low power levels, in an electromagnetic environment which can be congested with interfering signals, that is to say occupying the same bands, which are much more powerful because they originate from powerful ground transmitters (broadcasting transmitters, television transmitter, DME transmitter) often close to receivers.

The problem of deworming consists in eliminating as much as possible the interference affecting reception, the radio navigation signals emitted by the GNSS satellites to avoid masking of the signals of the satellites rendering the processing of the receiver inoperative.

The known solutions are essentially of two types: spatial filtering and spectral filtering with variants mixing the two solutions. Spatial filtering consists in playing on the differences between the directions of origin of the useful signals and interference signals, and more especially, in taking advantage of the fact that the useful signals come from sources satellites with a high site angle while the interference signals come from sources on the ground or close to the ground with a low site angle. It consists in modeling the shape of the radiation pattern of the receiving antenna to place the sources of interference in the holes of this diagram and the sources of useful signals in the main lobes. This technique is implemented using a directive reception antenna having a network of radiating elements associated with a more or less sophisticated pointing system.

Spectral filtering consists in analyzing the frequency spectrum of the received signal, identifying in this spectrum the frequencies affected by interference and eliminating or reducing them as long as the interference lasts. The identification of the frequencies affected by interference is based on the expected form of the useful signals which are signals transmitted in spread band obtained by modulation of a carrier by means of pseudorandom binary sequences and which have the property of having in the band of frequency they occupy or useful band, the properties of white noise with a flat frequency spectrum and a level of the order or lower than that of thermal noise due to the low emission powers and the distances traveled. Spectral lines exceeding, in the received signal, an arbitrary threshold close to the thermal noise level, are considered to be due to interference and therefore to attenuate or even to eliminate.

Spectral filtering is effective for narrow band interference, on the order of less than twenty percent of the wanted signal band, or for short-term interference, which is often the case with unintentional interference from 'human activity.

The object of the present invention is to weaken and even eliminate a large category of interference affecting the reception of a radionavigation signal by satellites which is constituted by parasitic signals situated in the band of the radionavigation signal but not having like him a spectrum of white noise.

It relates to a deworming method, on reception, of a radioelectric signal transmitted in spread band by modulation of a carrier by means of pseudo-random binary codes remarkable in that it consists in subjecting said radio signal received on reception to a spectral whitening filtering rendering it its characteristics of white noise in the occupied band, by means of a spectral filter whose coefficients are deduced from an estimate of the coefficients of its matrix autocorrelation.

This deworming process makes it possible to reduce, in large proportions, the interference caused, on a radio signal having the characteristics of white noise in the useful band as is the case with a radio navigation signal by satellites, by all parasitic signals which, not having a white noise spectrum, have a partly predictable shape.

Advantageously, the whitening spectral filter used is a transverse filter of order p, with coefficients a _pi corresponding to an autoregressive modeling of order p of the captured signal represented by a series of samples {* „}: nf, n, p / J neither nor

(ε _fnp being a modeling error of the sample s _n comparable to a white noise if the model is faithful and the coefficient a _pfi being equal to 1 by definition) which are determined so as to minimize the mean squared modeling error .

Advantageously, the coefficients a _pi of a transverse spectral filter of order p performing the bleaching are obtained by solving the system of equations:

",? _" = 0 fke [l, p]

(= 0

c _k _ _ι being the coefficient of the ith row and the kth column of the autocorrelation matrix of order p of the picked up signal which depends only on the absolute value of the difference ki, the picked up signal being stationary, and which is estimated iteratively using the relation: ^C l = ∑C nl ^X n ^X n π = 0

/ being equal to: k -i, the a _nl being weighting coefficients,

* designating the conjugate complex,

N being the instant of correlation estimation, the initial instant 0 being chosen arbitrarily, x _n and x _{π + k} _. being samples of the signal to be suppressed, and the system of equations being solved with the constraint: a _pfi = 1

Advantageously, the bleaching filter is a transverse filter of order p with symmetrical coefficients.

Advantageously, when the whitening filter is a transverse filter of order p with symmetrical coefficients with an odd number of coefficients indicated by indices: ^a -M ' ^a -M +1> ^a -M +'" ^• ) ^a - \ X ^a 0, ^α i> - » ^a Ml ' ^a M

(being an integer such that: p = 2M +1), its coefficients being obtained by solving the system of equations:

M

Σ ^a pΛ ki + C k + i) = 0 V £ e [l,] ι = 0

^c _k -i ^{and c} _{k +} i ® ^both coefficients of the autocorrelation matrix of order p of the signal to be suppressed which depend only on the absolute values of the difference ki or of the sum k + i, the signal considered being stationary and real, and which are estimated iteratively using the relation:

^ê l = Σ ^a nl ^X n ^X n * π = 0

/ being equal to: k -i in one case and ak + i in the other, the α _nl being weighting coefficients,

N the instant of correlation estimation, the initial instant 0 being chosen arbitrarily, x _n and x _{n + k} __ being samples of the received signal, and the system of equations being solved with the constraint: a ₀ = iζ

Other characteristics and advantages of the invention will emerge from the description below of an embodiment given by way of example.

This description will be made with reference to the drawing in which a single figure schematically represents a deworming circuit according to the invention.

In reception, a signal of radionavigation by satellites is received by a reception antenna, brought back in intermediate band or in base band by the stages of entry of a receiver to be able to be sampled and digitized at a reasonable rate without loss of information, sampled and digitized, then subjected to a digital processing having for object to extract information from it on the position and the speed vector of the carrier of the receiver. After sampling and digitization, it generally takes the form of a sequence x {n} ^ of complex digital samples succeeding one another at a rate sufficient to respect Shannon's theorem, each sample having two components from '' a quadrature demodulation, a component from the phase I demodulation channel and a component from the Q quadrature demodulation channel

The signal of radionavigation by satellites has the particularity of being transmitted in spread band by means of a carrier modulated by pseudo-random binary sequences, which gives its frequency spectrum, in the band which is allocated to it, the characteristics pseudo white noise, with a relatively constant spectral density implying that the signal in the useful band is difficult to predict.

As a radio navigation signal by satellites is transmitted with low power, the electric power available on board a satellite being counted and the distance traveled by strong waves, the signal received in reception is received at or below the noise level thermal. Due to its low power, its operation by a receiver is easily disturbed by parasitic interference in its useful band, caused by radioelectric signals due to human activity and present in the receiver environment at often much higher power levels. A large part of the parasitic interference which is found in the useful band of a radiocommunication signal by satellites causes a break in the uniformity of the spectral density of the signal received, which has led to the development of different deworming techniques consisting of restore, by spectral filtering, the spectral density uniformity of the signal in the useful band before seeking to use it to extract information therefrom on the position and the speed vector of the carrier of the receiver.

We propose here a deworming technique consisting in eliminating from the signal received, any predictable component which can only come from interference since the expected signal must have a uniform spectral density in the useful band. This deworming technique makes it possible to effectively combat any interference by a radio signal having more or less pure spectral lines such as a CW continuous wave signal or a FM frequency modulation signal.

An interfering signal s is supposed to be predictable when the value of one s _n of its samples can be deduced from a linear combination of the values of its np previous samples, that is to say that the sequence s {n} of his samples verify a prediction relation in the direct sense of the form:

sn = ε f _f , n, p - /> J a pi .s ni.

(= 1

where p is the order of the model, i.e. the number of previous samples taken into account in the prediction, {a _p , j} a series of p complex coefficients defining the model and ε _{/ î p} a prediction error in the direct direction, by the model of order p, of the sample s _n which must be comparable to a white noise if the model is faithful, or again: 'fnp = Σ p, ι n (2)

. = 0

with

^U P, O = ¹ The prediction relation (1) performs a so-called modeling

"autoregressive" because, in the absence of noise and measurement errors on the samples, it is all the more faithful the higher the order p of the model. We notice that we have:

P

___? __, ap „, ι, s n _, - i, = —sn„ - ε j _t , n „, p„ ι = l

so that the prediction coefficients {a _p , j} make it possible to define a transverse filter with finite impulse response RIF making it possible to eliminate from the signal received, the signal interfere predictable s. We also note that the modeling of the predictive interfering signal can be done from the signal received itself since the useful signal is comparable to a noise in the same way as the prediction errors so that we can say that the modeling is also that of the signal picked up, the samples {s _n } being able to be replaced in the prediction relation by the samples {x _n } of the signal picked up.

It remains to determine the prediction coefficients {a _Pι j}. A known method for determining these coefficients is the so-called least squares method.

The method of least squares is based on a minimization of the sum of the squares of prediction errors in the direct direction ε _{/ π p} for n varying from 0 to Np-1, the samples s _n being assumed to be zero outside the interval [1, N]. We have, according to equation (2):

To minimize this sum of squares of errors, it is necessary to cancel its derivatives with respect to the sequence {a _Pι ι} of the prediction coefficients, which results in the system of linear equations:

As the useful signal is considered as noise, one can replace, in the system of linear equations (3), the samples {s _n } of the predictable interfering signal by the samples {x _n } of the picked up signal:

Σ ^a _p l∑ ^x _n -i _j ) = 0 Vje [l, ..., p] (4) i = 0

The system of linear equations (4) can still be written, since a _Pι0 is equal to one by definition:

∑s _n s _n ^* _ _j = - a 5 „_, x V / e [l, ..., p] ι = l ^» being the conjugate operator.

By asking :

T being the transposition operator, and

-0 "1 'p-l

Ω = -0 - p-2

^C pl ^C p-2 we can rewrite the system of equations (3) in a matrix form and arrive at the normal Yule-Walker equation:

^Ω Λ: -C „

from which we derive the value of the vector A _p from the prediction coefficients of order p

A _p = -Ω ^~ ; C _p (5)

-1 being the inversion operator.

We note that Ω _p is both the autocorrelation matrix of order p of the predictive interfering signal s that we are trying to suppress and that of the signal picked up {x _n } since the wanted useful signal is comparable to white noise as well as the prediction error. The application of the method of least squares for the determination of the prediction coefficients and therefore of the transverse filter with finite response allowing to eliminate the predictable interferences of the signal of radionavigation by satellites thus passes by the estimation of the autocorrelation matrix of the received signal and its inversion.

The estimation of the coefficients of the autocorrelation matrix of order p of the captured signal can be done in the following iterative manner:

a _nk being weighting coefficients chosen to attenuate the edge effect due to truncation,

N the instant of estimation of the autocorrelation and 0 the initial instant arbitrarily chosen.

The figure illustrates a deworming circuit implementing the method which has just been described. It has three modules:

A first module 1 for estimating the autocorrelation matrix of the signal placed at the input and receiving the x _n samples of the signal to be suppressed, received and digitized by the input stages of a satellite navigation receiver,

A second module 2 for calculating the bleaching filter coefficients operating from the correlation matrix coefficients estimated by the first module 1, and

A third whitening filtering module 3 consisting of a digital filter which receives its coefficients from the second module 2 and processes the samples x _n of the signal to be suppressed before their use by the stages downstream of the satellite radionavigation receiver in order to extract therefrom information on the position and the speed vector of the receiver carrier.

The first module 1 for estimating the autocorrelation matrix receives the samples {x _n } of the signal picked up while it has been placed in the intermediate frequency band or in the base frequency band by the input stages of the receiver of radionavigation by satellites but before demodulation and despreading. It applies to these samples the iterative method of estimating the coefficients of the autocorrelation matrix described relative to relation (6) to generate an estimate of these coefficients intended for the second module 2 for calculating the bleaching filter coefficients.

The second module 2 for calculating the bleaching filter coefficients calculates prediction coefficients from the normal Yule-Walker equation by applying the relation (5) to the estimated coefficients of the autocorrelation matrix of the signal received, provided by the first module 1. To do this, it reverses beforehand the estimated autocorrelation matrix provided by the first module 1 for example by means of a Levinson-Durbin type matrix inversion technique. The third whitening filtering module 3 performs, on the samples {x _n } of the signal received, the operations of a spectral filter of the transverse type with a finite response FIR taking as coefficients the prediction coefficients provided by the second calculation module 2 whitening filter coefficients. The filtered signal it delivers is applied to the downstream stages of the satellite navigation receiver, directly under digital form, either after switching to high frequency including digital-to-analog conversion and up-frequency, followed by down-frequency and a second analog-to-digital conversion.

As a variant, it is possible to impose on the filter implemented by the third whitening filtering module 3 a condition of symmetry on its coefficients giving it a linear phase law which is particularly appreciated for the exploitation of the radionavigation signal by satellites after the operation filtering since a linear phase law filter behaves as a simple delay from the point of view of the distortion brought to the radio navigation signal.

In the case of a bleaching filter with an odd number of coefficients, the condition of symmetry is better highlighted with the following indexing of the coefficients:

^a -M ' ^a -M + l' ^a -M +2 " ^{• • •} > ^a - \ '^ ^a 0' ^a i '" ^• > ^a M- \' ^a M

(M being an integer such that: p = 2M +1). It then becomes:

_. = a,

Its taking into account during the application of the method of least squares to the research of the prediction coefficients, leads to the resolution of a system of linear equations modified compared to that mentioned previously in (4). This new system becomes:

Σ * pl∑ ^X ni ^X : - _J + Σ ^X n + i ^X lj) = 0 V * 6 [Uf] (= 0 nn J with the constraint: a ₀ = y ₂ or again:

M M. .

∑ _p c _H + c _{J + i} ) = 0 Y / e [l, M] i = 0 always with the constraint a ₀ =

c _H and c _{j + i} being coefficients of the autocorrelation matrix of order p of the modeled signal which only depend on the absolute values of the difference ji or of the sum j + i

The solution of this system of linear equations is obtained in the same way as that of the system of linear equations (4), as solution of a matrix equation of Yule-Walker type combining matrices respectively of Toepliz and Hankel type obtained using estimates of 2M + 1 autocorrelation coefficients of the signal received.

The deworming method which has just been described applies to the treatment of any type of parasite or high level radio interference for the reception of radioelectric signals having a uniform spectrum in their useful band, as soon as the parasites or interference have somewhat predictable forms.

To facilitate understanding, the deworming treatment has been represented and described as being executed by three separate modules, a module for estimating the coefficients of the autocorrelation matrix, a module for determining the coefficients of the whitening filter and a module for whitening filtering. It should not be concluded that this is necessarily the case in practice. Indeed, as we are dealing with digital signals, all the processing and functions carried out in these modules are done by means of one or more processors controlled by software, the distribution of tasks between one or more processors coming under the competence of the skilled person.

Claims

1. A method of deworming, on reception, of a radio signal transmitted in a spread band by modulation of a carrier by means of pseudo-random binary codes, characterized in that it consists in subjecting said radio signal received on reception to filtering. whitening spectral rendering it its characteristics of white noise in the occupied band, by means of a spectral filter whose coefficients are deduced from an estimation of the coefficients of its autocorrelation matrix.

2. Method according to claim 1, characterized in that the spectral bleaching filtering is carried out by means of a transverse filter of order p, with coefficients a _pk corresponding to an autoregressive modeling of order p of the captured signal represented by a series of samples {x _n }: χ „= ε f, n, P / ^a p, i ^X n- ι = l

(ε _{/ Λp} being a modeling error of the sample x _n and the coefficient a _{p ϋ} being equal to 1 by definition) which are determined so as to minimize the mean square error of modeling.

3. Method according to claim 2, characterized in that the coefficients a _pi of the transverse filter of order p carrying out the bleaching filtering are obtained by solving the system of equations:

c _k _. being the coefficient of the ith row and the kth column of the autocorrelation matrix of order p of the picked up signal which depends only on the absolute value of the difference ki, the picked up signal being stationary and real.

4. Method according to claim 3, characterized in that the coefficients c _k __ of the autocorrelation matrix of the sensed signal are estimated iteratively by means of the relation:

^C l ⁼ _j ^a nl ^X n ^X n * π = 0

l being equal to: k - i, the a _nl being weighting coefficients,

N being the instant of estimation of the correlation, the initial instant 0 being chosen arbitrarily, x _n and x _{n + k} _. being samples of the signal to be suppressed, and the system of equations being solved with the constraint: a _{p 0} = 1

5. Method according to claim 2, characterized in that the spectral bleaching filtering is carried out by means of a transverse filter of order p with symmetrical coefficients.

6. Method according to claim 2, characterized in that the bleaching filtering is carried out by means of a transverse filter of order p with symmetrical coefficients with an odd number of coefficients identified by indices: ^a - _M ' ^a - _M + ι ' ^a - _{M +2} '•••> ^a -ι, 2a ₀ , a _x , ..., a _M _ _l , a _u

(M being an integer such that: p = 2M +1), its coefficients being obtained by solving the system of equations:

M

∑tf _{p ,.} fe-A <A,) = 0 \ / ke [l, M]

1 = 0

^c _k -ι ^{and c} _{k +} i being coefficients of the autocorrelation matrix of order p of the sensed signal which only depend on the absolute values of the difference ki or of the sum k + i, the sensed signal being stationary and real .

7. Method according to claim 6, characterized in that the coefficients c _k _, and c _{k + i} of the autocorrelation matrix of the sensed signal are estimated iteratively by means of the relation:

/ being equal to: ki in one case and to k + i in the other, the α _nl being weighting coefficients,

N being the instant of estimation of the correlation, the initial instant 0 being chosen arbitrarily, χ _n and x _{n + lc} _. being samples of the received signal, and the system of equations being solved with the constraint: α _{p 0} = Vi.

8. Method according to claim 1, characterized in that the processed radio signal is a radio navigation signal by satellites.