METHOD FOR SEPARATING ELECTROMAGNETIC EMISSIONS FROM A
PLURALITY OF EMITTERS
5 The subject of the invention relates to a method and a system
making it possible to separate electromagnetic emissions from a plurality of
emitters or sources in the field of telecommunications or even in the field of
radars.
One of the technical problems when in the presence of a plurality
io of emitting sources is how to separate them and then group together the
individual detections by emission, that is to say produce a segmentation of
the measurements by unsupervised classification, better known by the term
"clustering". It is therefore necessary to know how to separate the emissions
originating from a plurality of emitting sources and received on a plurality of
is sensors or an antenna array.
The prior art describes a number of methods for separating signals
originating from a plurality of sources or emitters.
For example, the blind separating of sources consists in estimating
a set of N sources from a set of P observations which are mixes of these
20 sources. The term blind means in the present context that there is no a priori
knowledge of the signals.
It is also known practice to perform the separation of signals by
using the parameters such as the frequency, the measurement instants, the
directions of arrival. The method implemented in the prior art is known as
25 "unsupervised classification".
The techniques used in the prior art give good results. However,
they notably present certain drawbacks including an inability to separate two
similar electromagnetic emissions even if they are of different polarization. In
practice, since they do not make use of the polarization measurements,
30 confusion is possible between two emissions of difference polarization, if they
otherwise exhibit relatively similar characteristics.
2
The object of the invention is notably to offer a method for
separating detections originating from two close or similar emissions , if their
polarization is different, and do so by exploiting the polarization of the
electromagnetic emissions received on a sensor.
5 The invention relates to a method for de-interleaving
electromagnetic signals received on an array of sensors positioned on a
carrier, said array of sensors being associated with a signal processing
module comprising , in combination, at least the following steps:
® acquiring electromagnetic signals from a plurality of distinct sources,
10 *determining a similarity criterion from the measurement of the
polarization of said electromagnetic signals,
® using said similarity criterion in an unsupervised data classification
method.
According to one embodiment, corresponding to a fixed sensor
15 configuration , the similarity criterion on the polarization is determined from
two measurements of detection Di and D2 of the electromagnetic signals
received on the array of sensors by using the Mahalanobis statistical
distance
d2mr =(V-V f(Mi+M;^'(V -V')
20 where V„ V, are the parameters of polarization of the detections D;, Dj; and
M,M, are the associated covariance matrices;
the polarization measurements corresponding to D;, Dj are given by a
complex polarization vector (a, (p):
5
V; =
Vi =
polarization parameters of the detection D;
(Pil
ra I
polarization parameters of the detection Dj
L(;
3
the exponent T is the transposed sign and the index m is the Mahalanobis
significance
and then by computing a measurement of similarity in polarization by the
formula:
5
s(Vi,V2)= max}0, 1- ,6. (V,-V2)T(MI +M2)-i(T; -V2)}
According to another embodiment, corresponding to a slightly
io mobile sensor, the variations of attitude position of antenna array remaining
low with regard to the intercepted emitter, the similarity criterion is
determined by introducing a model noise b =
01
0
between two polarization
measurements VI and V2 separated by a time T, taking into account the
trend of the inclination of the ellipse between the two measurements:
is by determining the covariance of the model noise
6b
0
Mb [ 0
01 covariance of the model noise
with a standard deviation
2 2 2 2
6b = 4 LCl^m T +6,
and by computing the extended Mahalanobis statistical distance by
20 introducing the covariance Mb of the model noise:
s(V,V2) =max}p a. (V - V2 )T (M1 + M2 + Mb)-'(VI - V2 )}
According to a third embodiment, suited to the situations in which
25 the sensor is greatly mobile , a first step for determining the similarity criterion
4
consists in correcting the measurement of inclination of the polarization
ellipse by taking into account the navigation information of the carrier, by
performing a change of reference frame of the polarization plane, by a
rotation in the polarization plane making it possible to change from the
5 reference frame [uh uv] linked to both the calibrated antenna array and the
trim to the reference frame [u'h u'v] linked to the horizontal of the place, and,
during a second step, in computing
s(V,V2)= max}0, 1- ,8. (Vi-V2) (MI +M2)-'(V-V2)}.
10 The method according to the invention is used, for example, for
telecommunication signals to be de-interleaved or for radar signals.
The invention also relates to a device making it possible to deinterleave
electromagnetic signals received on an antenna array, said
antenna array being associated with a signal processing module suitable for
1s executing the steps of the method having the abovementioned.features.
Other features and advantages of the device according to the
invention will become more apparent on reading the following d3scription of
an exemplary embodiment given as an illustrative and nonlimiting example
with appended figures, in which:
20 ® Figure 1 represents an exemplary architecture making it possible to
implement the method according to the invention,
• Figure 2 represents a representation of the modelling of the polarization
in reception,
® Figure 3 represents a modelling of the inclusion of the polarization
25 measurement according to the invention,
• Figures 4 and 5 represent an exemplary implementation of the method
according to the invention in a COMINT segmentation processing.
In order to better understand the object of the invention, a
description given below relates to a method for separating electromagnetic
so emissions for telecommunication signals.
5
Figure 1 schematically represents an exemplary system making it
possible to implement the method according to the invention. A plurality of
sources emit signals in different directions. A sensor or an array of sensors
associated with a data processing device receives the signals emitted by the
s different sources 10i or E.
An exemplary general architecture of the emission interception
and emitter locating equipment according to the invention is represented in
Figure 1. It comprises an antenna; array 1, a reception device 2, a signal
acquisition circuit 3, a data extraction and processing module 4. The antenna
io array is, for example, arranged on a carrier 5 that can be identified by its
position and its attitude.
The antenna array 1 receives a set of signals originating from
emitters that have to be differentiated. The objective is notably to separate
the signals received on the antenna system, to group together the individual
is detections by emission, that is to say, to produce segmentation of the
measurements by unsupervised classification, better known by the term
"clustering".
The idea implemented by the method according to the invention
consists notably in using the measurement of the polarization as an
20 additional similarity criterion in the unsupervised classification process,
known to the person skilled in the art.
In order to better understand the object of the invention, a review
of polarization measurement will be given in Figure 2.
The polarization measurement is an estimation to a complex factor
25 close to the Jones vector J, well known to people skilled in the art. This
vector makes it possible to characterize the behaviour of the electrical field E
of a monochromatic planar wave in the wave plane 20 (polarization plane)
orthogonal to the wave vector k.
Generally, the end of the vector E describes, as a function of time
30 in the polarization plane, an ellipse 30, Figure 3, characterized in a reference
frame [uh u„], by an inclination a, a half small axis a, a half large axis b and a
5
direction of travel. The estimation of the Jones vector to within a complex
factor can be characterized by a unitary vector p of components ph and p„
such that:
Ph = cos a
P„ = since'"
where a, abs((p) (absolute value of (p) and sign ((p) respectively define the
inclination, the angle of ellipticity, and the direction of travel of the polarization
ellipse in the reference frame [uh u„ ]
The polarization measurement is assumed given by the estimation
io of the unitary vector p = [Ph p„]' or, equivalently, by the complex ratio P„ / Ph
(=tanae J(I) or even by the pair (a , (p) (or by the vector v = [a cp]').
The accuracy of the polarization measurement is assumed described by a
covariance matrix on the parameters (a , (p).
The choice of the base [uh u„] as geometric reference frame is
15 arbitrary but linked to the reference frame of the antenna or antenna array.
By considering that uh characterizes the linear polarization being propagated
in the plane [uh k] and that u„ characterizes the linear polarization being
propagated in the plane [k u„], it is observed that the vector p represents the
decomposition of the field E in the orthonormed base [uh u„]. By a misuse of
20 language, the polarizations Ph and P, are respectively called horizontal and
vertical components, although they are truly that only when the reference
frame of the antenna or antenna array coincides with the horizontal and the
vertical of the place (that is to say, in an onboard context, when the attitude
of the carrier is horizontal) and only when the wave is being propagated in
25 the horizontal plane.
Figure 3 is a schematic representation of the use of the
polarization measurement in the de-interleaving of the signals received on
the receiver of the Figure 1, of which'a number of examples will be detailed
hereinbelow.
30
7
Figure 4 describes an exemplary implementation of the method in
a segmentation processing operation COMINT and a way of taking into
account the additional polarization criterion derived from polarization
measurements.
5 A typical segmentation processing operation COMINT consists in
producing a frequency histogram on the detection-individual goniometry data
(individual blocks), 400, then in separating the emissions of fixed frequency
type (FF), 401, from the emissions of pulsed type, or "bursts", EB, 402, on a
frequency occupancy ratio criterion. The blocks FF are then transmitted to a
io first processing subsystem FF which will produce, for each of the active
frequencies, an azimuth histogram 404 making it possible to consolidate the
grouping together on direction of arrival DOA criterion, for example, and to
validate the tracks 405.
The emissions of EB type are transmitted to a second processing
15 subsystem, 402, which will segment the frequency evasion emissions EVF,
406, as well as the bursty emissions isolated on the basis of direction of
arrival DOA criteria (azimuth histogram), then of estimation of the jump
frequencies 407, and of the level durations 408, in order to complete the
segmentation in terms of duration and synchronization using histograms. The
20 classes obtained are then merged and synthesized, to generate synthesis
blips of type FF, EVF and EB respectively 410, 411, 412 by methods known
to the person skilled in the art.
The taking into account of the polarization criterion can be done on
the classes obtained as a result of the azimuth histograms by introducing, for
25 example, a polarization histogram computation, or else a partitioning based
on the measurement of polarization similarity (a definition of the similarity
criterion is given later in the description), in order to group together in one
and the same class only the polarization-compatible polarization detections.
The addition of this step of segmentation on polarization criterion 504b, 506b,
30 is illustrated in Figure 5.
The polarization of the emitter (that is to say of the emission
8
source) is assumed constant. It is best, however to take into account the
variations of the polarization measurement between successive
measurements occasioned by the possible movements of the carrier. The
variation between two polarization measurements of one and the same
5 emission performed at two instants substantially spaced apart is caused both
by the changing of the geometric configuration (angular movement of the
carrier of the antenna array resulting in a variation of the incidence of the
wave vector and therefore a variation of the perceived polarization which can
be seen as a projection of the polarization of the emitter in the received wave
1o plane) and by the changes of trim of the carrier (resulting in a rotation of the
reference frame [uh u„] in the polarization plane, which affects the
measurement of inclination of the polarization ellipse).
According to a first variant embodiment, the method is used on
fixed emitters in the case of receivers or interceptors of the emitted signals
15 that are fixed on the carrier. The method uses a "rough" polarization
measurement.
The taking into account of the polarization measurement to
compare two detections translates into considering a polarization similarity
criterion based on the Mahanalobis distance, which is well known to people
20 skilled in the art.
If one considers two detections D1 and D2 of the electromagnetic signals on
the antenna array, the polarization measurements corresponding to these
detections are given, for example, by a complex polarization vector (a, (p)
V1=
w J
polarization parameters of the detection D1
25 V2 = a2
Vz
polarization parameters of the detection D2
The covariance matrices associated with the polarization measurements
(assumed Gaussian) are determined by methods known to the person skilled
in the art. Let
M1 be the covariance matrix associated with (V1)
9
M2 be the covariance matrix associated with (V2)
If the measurements on the inclination and eccentricity of the polarization
ellipse are decorrelated , the covariance matrix amounts to the associated
standard deviations ca, a9.
5 By definition in this first case, the relative variations of position and
of attitude of the antenna array of the receiver with respect to an emitter are
assumed negligible compared to the less good value of the accuracies of
estimation of the polarization inclination max(aa1,aa2). This makes it
possible to consider that the relative incidences of the received waves, and
io therefore the wave vectors , are identical.
In the case of a fixed sensor configuration, the polarization
similarity criterion is expressed by using the statistical Mahalanobis distance
known to the person skilled in the art:
d2M =(V.-V' )T(M,+MJ`(V -Vi)
is where V„ Vj are the polarization parameters of the detections D;, Dj ; and
M;,Mj are the associated covariance matrices;
the exponent T is the transposed sign and the index m is the Mahalanobis
significance.
This statistical Mahalanobis distance then makes it possible to define a
20 measurement of polarization similarity given by the formula:
s(V,,V2 ) = max}0, 1- fl. (V -V2Y(M1 +MO I (V1-V } in which
s(V,V2)= max^, 1-l, d2na }
25 where R is an adjustment coefficient which is fixed by choosing a threshold
on the Mahalanobis distance. This threshold is defined generally from a
probability of non-association of measurements relating to the same entity
(the threshold is computed on the basis of this probability d by using the
Mahalanobis distance distribution function which follows a chit law).
1 0
This measurement of polarization similarity is then used in the
method of separating and of grouping together electromagnetic emissions as
an additional criterion for the unsupervised classification.
It can be used, for example , either by contribution to the definition
s of a global multicriteria similarity on the detections which will then include the
polarization , or simply as a new criterion implemented on the classes
obtained after unsupervised classification based on the conventional criteria
(cascading of criteria ). In this latter case , if the polarization measurement
accuracies are almost mutually constant and uniform , it may be
io advantageous to limit the algorithmic complexity introduced by the new
partitioning criterion by prioritizing the recourse to histograms on the
polarization measurements for which the pitch is fixed on the basis of the
accuracy standard deviations.
According to a second variant embodiment , the variations of
15 position and of attitude of the antenna array are small, but not negligible,
compared to the polarization measurement error. These variations
correspond , for example the variations of position and of attitude of the
antenna array remain small with respect to the intercepted emitter. In this
case , the method will introduce a model noise to produce comparisons
20 between polarization measurements.
Considering a maximum angular speed of movement damax, the
variation of trim of the antenna array can be modelled by a centred white
noise of angular standard deviation a.. This variation of trim will essentially
affect the inclination measurement a.
25 Between two polarization measurements V1 and V2 separated by
a duration T, the method introduces a model noise b that takes into account
the trend of the inclination of the ellipse between the two measurements:
11
Mb =
0] 0
covariance of the model noise
with a standard deviation
2 1 2 2 z
6G - 4 Clam.'T + 6W
eb is determined from the maximum angular speed of movement of the
s carrier and from the standard deviation on the inclination.
The polarization similarity criterion will then use a statistic
Mahalanobis distance extended by introducing the covariance Mb of the
model noise:
1 0 s(V,V2)= max{0, 1-a. (Vi-V2 Y(MI +M2 +Mb)-'( - V2 )}
If a partitioning method based on histograms on the polarization
measurements is preferred over a partitioning method using the similarity
measurement based on the Mahalanobis distance, the taking into account of
15 the model noise is reflected in an increase in the pitch of the histogram
deduced from the standard deviation 6b.
According to a third variant embodiment suited to the situations
where the sensor is greatly mobile , for example when the value of the
observed polarization is greatly modified by the variations of position and/or
20 of attitude of the carrier.
The method for this third variant implementation consists in limiting
the effects of the variations of position of the carrier by limiting the use of the
polarization measurement to the de-interleaving processing operations which
occur over a limited horizon, typically one sensor cycle (over the duration of a
25 sensor cycle , which is of the order of a second , the variation of position of the
carrier can be considered to be negligible compared to the measurement
accuracies).
12
With regard to the variations of attitude of the antenna array,
considered as potentially important in this latter case (even over the horizon
of one sensor cycle ), the method relies on the use of navigation information
delivered, for example , by a navigation unit known to the person skilled in the
5 art and situated on the carrier , and on the choice of a reference frame [u'h
u'„] of the polarization plane which is aligned on the horizontal of the place
and therefore independent on the attitude of the carrier.
The choice of such a reference frame entails correcting the
measurement of inclination of the polarization ellipse by taking into account
10 the navigation information by performing a change of reference frame of the
polarization plane (rotation in the polarization plane making it possible to
change from the reference frame [uh u„] linked to the calibrated antenna and
therefore to the trim, to the reference frame [u'h u '„] linked to the horizontal of
the place). This reference frame then makes it possible to compare the
is corrected measurements when the variation of the position of the carrier
relative to the emitter remains small.
The corrected polarization measurements are determined by the
following formula , by using the values of V1 , V2, corrected by changing the
reference frame.
20 s(V,VZ) = max}0, t-,(3.IV- k (M1+M2)-'(V -V2)}
If a partitioning method based on histograms is preferred, the corrected
measurements are counted directly with a histogram pitch based on the
measurement standard deviations.
25
13
CLAIMS
1 - Method for de-interleaving electromagnetic signals received on an array
(1) of sensors positioned on a carrier (5), said array of sensors being
s associated with a signal processing module (4) comprising, in combination, at
least the following steps:
acquiring electromagnetic signals from a plurality of distinct sources,
determining a similarity criterion from the measurement of the
polarization of said electromagnetic signals,
10 a using said similarity criterion in an unsupervised data classification
method.
2 - Method according to Claim 1, characterized in that the similarity criterion
on the polarization is determined from two measurements of detection D1 and
15 D2 of the electromagnetic signals received on the array of sensors by using
the Mahalanobis statistical distance
din, = (V, -Vjy, (Mi + MY (V,. -Vj )
where V, V, are parameters of polarization of the detection D;, Dj; and
M„M, are the associated covariance matrices;
20 the polarization measurements corresponding to D;, D1 are given by a
complex polarization vector (a, (p):
V=
a,1
polarization parameters of the detection D;
Vi =
'P, j
[ai
(P1
polarization parameters of the detection Dj
the exponent T is the transposed sign and the index m is the Mahalanobis
25 significance
and then by computing a measurement of similarity in polarization by the
formula:
14
s(V,V2)= max}0, 1 /3. (Vi VZ)T (M1+M2)-'(VI -V2)}
3 - Method according to Claim 1, characterized in that the similarity criterion
5 is determined
by introducing a model noise
0
0 between two polarization
measurements VI and V2 separated by a time 7, taking into account the
trend of the inclination of the ellipse between the two measurements
by determining the covariance of the model noise:
10 Mb =
0 0
covariance of the model noise
with a standard deviation
2 1 2 2 2 Ch =4da,„..T + 17, ,and
by computing the extended Mahalanobis statistical distance by introducing
the covariance Mb of the model noise:
15
s(VI,V2)=max }0, 1-a4( VI-V2)7(M1+M2+Mb)-'(V -a2)}
4 - Method according to Claim 2, characterized in that, to determine the
similarity criterion,
20 ® a first step consists in correcting the measurement of inclination of the
polarization ellipse by taking into account the navigation information of
the carrier, by performing a change of reference frame of the
polarization plane, by a rotation in the polarization plane making it
possible to change from the reference frame [uh u„] linked to both the
1 5
calibrated antenna array and to the trim to the reference frame [u'h u'„]
linked to the horizontal of the place, and
during a second step, the following is computed
s s(Ti,V') = ma41 1- '8. Jl;-V2y(Mi+M2)-i(V -Vi^j.
5 - Method according to one of Claims 1 to 4, characterized in that the
electromagnetic signals to be de-interleaved are telecommunication signals.
io 6 - Method according to one of Claims I to 4, characterized in that the
electromagnetic signals to be de-interleaved are radar signals.
7 - Device making it possible to de-interleave electromagnetic signals
received on an antenna array (1), said antenna array being associated with a
is signal processing module (4) suitable for executing the steps of the method
according to one of Claims 1 to 6.