The invention relates to a method and to a-system'for detecting one or more electromagnetic emissions emitted in a given area, using a distributed sensor architecture including at least two antennae.
Hereinafter, the term 'sensor3 will be applied to any system capable of digitizing an electromagnetic signal and that is equipped with a processor for processing the digitized signal and with a digital emission system. A sensor may be a software-defined radio receiver, for example. The invention is notably used in the field of radio monitoring.
Locating electromagnetic sources is a significant problem in the field of monitoring authorized or unauthorized emissions in a given space.
In order to detect electromagnetic emissions, it is known, from the prior art, to use the principle of blind detection, which may be implemented using several techniques. _
It is thus possible, in order to locate one or more emitter sources, to estimate the time difference of arrival (TDOA) and the hyperbolic location in accordance with a method described in patent US2005/0073459. However, the main drawback of the location technique using the TDOA is that of having to transmit, on the radiocommunication emissions, the digitized signal of each detected emission to the unit performing the intercorrelations between the signals from the various sensors in order to estimate the time difference of arrival, and then the position of the emitter. The location performance depends on the product band factorxsignal time (i.e. on the number of samples), this meaning that, in order to achieve good performance, a large number of samples have to be exchanged and therefore a high communication throughput of the sensors is necessary. The limitation on tfie throughputs does not allow all of the emissions to be systematically located.

Another technique consists in estimating the angle of arrival (AOA) of the emissions and the MAOA (multiple angle of arrival) triangulation using a moving sensor or with a plurality of sensors. One example is given in the publication by Don J. Torrieri, entitled 'Statistical Theory of Passive Location Systems3, IEEE 1984. MAOA location techniques implement unambiguous networks of antennae and have complex antenna networks having more than two antennae. These networks may prove difficult to integrate into a carrier, in particular for the low frequencies imposed by networks of large size, given the wavelengths and the angular precision required,
MAOA location techniques implement PhDOA (phase difference of arrival) goniometry, allowing location ambiguities to be removed by utilizing the measurements performed by the sensor during the trajectory of the carrier.
This multichannel location technique leads to ambiguous estimations of directions of arrival, whose ambiguities are not able to be resolved if the sensor is static or if the periods of the emissions that are observed are too short in terms of the time necessary to move the sensor so as to remove the ambiguities.
Patent US8248210 describes the use of two antennae and the calculation of the differential phase in order to estimate angles of arrival, with ambiguities, of a plurality of RFID (radiofrequency identification) identification tags so as to distinguish them spatially. These tags are distinguished spatially by calculating the phase differences linked to each of the RFID tags (SD-PDOA for space"division - phase difference of arrival technique).
Patent USH2224H1 uses the calculation of the phase difference of arrival, or PhDOA, of the TDOA and of the frequency difference of arrival FDOA between two single-channel receivers to locate the emitter.
Another way consists in estimating the power difference of arrival or PwDDA. However, this location technique has the drawback of having location performance linked to the propagation models. With the propagation channel not being able to be estimated in a blind location system, the

location performance is uncertain and is often unacceptable, in particular in an urban environment or in a wet environment.
For example, patent US9316719 uses a PwDOA technique to estimate the position of an emitter using-a mobile sensor and measurements of the power difference of arrival
Patent" application WO98/47019 presents a location system in which the cellular network through which the emitters communicate will locate the latter.
The various techniques from the prior art do not make it possible, in particular and at the same time, to solve the following problems:
• The short emission periods of the emitters to be located (TDOA, PhDOA with moving MAOA and unambiguous AOA, moving MAO A),
• Systematic location of all of the communications intercepted by sensors with wide-band instantaneoos reception (TDOA),
• _Radiometry networks of antennae that are complex and difficult
to integrate (AOA),
• Fixed-position emitters to be located and fixed-position sensors (FDOA),
• Good-performance operation of the blind location (estimation of the propagation channel for the PwDOA or order to force the emitters to emit),
• Using the same software-defined radio (SDR) reception means to observe the emissions to be located and receive information from the other sensors contributing to the location,
• Separating co-channel emission situations (multipath or a plurality of sources in the same channel) from single-signal situations (PwDOA, PhDOA and TDOA).
In the remainder of the description, the"letter Pi denotes a power value and Pl denotes an estimated value.

The. invention relates to a method for locating at least one electromagnetic source S- within a communication network comprising at least two sensors Ri, R2, said sensors being separated by a given distance, -each sensor'comprising a network of antennae comprising at least two reception antennae A1? A2, characterized in that it comprises at least the following steps: •.
• For each sensor R1f R2,
• Estimating the powers P± and P2 of the received signal Sr coming from a source S and received at the sensors Ri, R2,
• For the signal Sr arriving at each of the two reception antennae, measuring the phase difference Acp on the basis of the value of the phase difference Acp and of information on the position and the orientation of the network of antennae Ai, A2l determining a set of potential directions of arrival DOA for a given source S, and then
• Merging the power values 1\, P2 received at the sensors Ri and R2l the
-potential direction of arrival DOA values and the geographical
coordinates of the networks of antennae of the sensors Ri and R2| in
order to calculate the coordinates of the emission source by executing
the following steps:
• On the basis of the power values Pll ?2 estimated by the two sensors Ri, R2, of a propagation model and of the value of the ratio of these powers BjY ^, defining, for each ratio value, an elementary location area Ze in which the source is present, and then a potential location area Zs corresponding to the intersection of the various elementary location areas for an emission source S,
• On the basis of the geographical coordinates of the sensors and of the possible directions of arrival, calculating the possible half-lines of sight and their points of intersection P(Dk),
• Retaining only those points of intersection P(Dk) of the half-lines of sight that belong to the potential location area Zs,

• Selecting the points of intersection P(Dk) for which the geographical
density is highest, and determining the coordinates of the emission
source on the basis of these points of intersection.
....-.■ , ■■-. The-step of-merging the data, estimated-power--values, directions
of arrival DOA and coordinates of the networks of antennae is executed in
the processor of each sensor of the communication network.
According to another variant embodiment, the step of merging the data is executed in a centralized processing device communicating with the sensors that are present in the communication network.
The method may comprise a step of auto-calibration of the network of antennae of each sensor by using the known position of the other sensors while they are being used in emission mode.
The method comprises, for example, a step of time segmentation and frequency segmentation of the signal received at the antennae, and a step of detecting the number of sources present by calculating the covariance matrix of the signal at each time/frequency cell and by comparing the two eigenvalues of the covariance matrix with the noise level
The propagation model used to determine an elementary location area of the source on the basis of the estimated power values is a Tree space' propagation model.
The measurement periods on the signal received (detection, various estimations known under the term 'sensing') from the sensors use the time division multiple access (TDMA) method, and a specific TSLT slot is used to execute the steps of the method according to the invention.
The invention also relates to a system for locating at least one emission source in a communication network comprising at least two sensors Ri, R2 communicating with one another by way of a radio link, each sensor comprising a network of antennae comprising at least two reception antennae Ai, A2j a sensor comprising a processor designed to process the received signals so as to determine the power values of the signal received at the antennae, the potential directions of arrival and the phase offset value

of the signal received at the two antennae, the location system comprising a processor designed to execute the following steps:
• Merging-the power values received and estimated at the sensors, the . • potential direction of arrival values and the coordinates of the-network of antennae, in order to calculate the coordinates of the emission source by • • executing the following steps:
• On the basis of the power values P^, P? estimated by the two sensors Ri, R2J of a propagation model and of the value of the ratio of these powers PJ P2, defining, for each ratio-value, an elementary location area Ze in which the source is present, and then a potential location area Zs corresponding to the intersection of the various elementary location areas for an emission source S,
• On the basis of the geographical coordinates of the sensors and of the possible directions of arrival, calculating the possible half-lines of sight and their points of intersection P(Dk),
• Retaining only those points of-intersection P(Dk) of the half-lines of sight that belong to the potential location area ZSs
• Selecting the points of intersection for which the geographical density is highest, and determining the coordinates of the emission source on the basis of these points.
The invention is applied to a method and to a system implementing software-defined radio receivers.
Other features and advantages of the present invention will become more clearly apparent on reading the description of exemplary embodiments alongside the figures, in which:
• Figure 1 shows an example of a system according to the invention,
• Figure 2 and Figure 3 show an example of the steps implemented by the method according to the invention,
• Figure__4 shows a depiction of the arrival of a wavefront at two antennae,

• " Figure 5 and Figure 6 show two depictions of the definition of a
possible location area,
• Figure 7 shows a three-dimensional histogram of the number of
• instances of intersections - of the lines of sight as-a function of their
abscissae and ordinates,
• Figure 8 shows an exemplary application in the case of a
communication using the TDMA waveform
To facilitate understanding of the subject matter of the invention, an example is given for a communication network comprising two sensors Ri, R2, communicating by virtue of radiofrequency links.
Each sensor Rk, R1f R2 comprises at least two antennae A1 and A2, each linked to a reception channel and comprising an acquisition system. The two sensors are positioned at a given distance with respect to one another. These two acquisition channels 10, 11 have to be synchronous. The samples acquired (corresponding in particular to the signal emitted by the source to be located) are sent to a processing module 12, such as a processor. Each sensor also comprises a geolocation system 14, giving its geographical position and its orientation, as well as a wireless communication module 15 that allows the sensors Rs< that are present in the communication network to exchange data with one another. The sensors do not have to be collocated, but there is not a minimum distance to be complied with.
Without departing from the scope of the invention, the system could comprise a remote centralized processing module that is not a sensor involved in the communication system.
In the example, each sensor comprises two antennae, but could comprise a number M of. antennae greater than two. The processing of the data that is explained hereinafter will then be carried out, for example, considering the antennae in pairs.
Figure 2 and Figure 3 detail the steps implemented by the invention. The steps described in Figure 2 are performed in each sensor. In

the example that will be described in Figure 3, the steps that are outlined are
performed by each sensor, but could, without departing from the scope of the
invention, be executed by a processing device separate from the N sensors
-of the system, which will-receive the various measurements carried out by
each of the sensors. • -
In Figure 2, each sensor R1? R2 receives the signals on the instantaneous band to be listened to, and samples the band on its two synchronous reception channels 10, 11 (Figure 1), 201 The sampling step is performed by each sensor, in_accordance with a principle known to those skilled in the art. The received signal Sr, corresponding to the two signals Sn, Sr2 received at each antenna Ai, A2, is time-segmented and frequency-segmented, for example using a bank of polyphase filters, 202.
For each time/frequency cell, 203, the processor of the sensor enumerates the number of co-channel sources contained in the received signal Sr. From the two reception channels, the processor will determine the number of sources that are potentially present in the signal: zero sources, one source or more than one source.
If there are zero sources or more than one source, 204, the processing of the time/frequency cell stops there, and the method continues to process the following time/frequency cells, 205.
If just one source S has been detected, 206, the processor of the sensor will execute the following steps:
• It estimates the received power P(S) of the source S, 207,
• It estimates the differential phase Acp of the signal Sr between the two reception channels of the sensor (between the two antennae) using a technique known under the abbreviation PhDOA ('phase difference of arrival5), 208,
• On the basis of the differential phase Acp estimated in the previous step"" and of information on the position and the orientation of the network of antennae, the processor estimates the potential directions of arrival (DOA), 209. Among these directions of arrival are the

direction of arrival DOA corresponding to the source S to be.located
- and one or more ambiguities, 210. The higher the ratio d/A, the higher the number of ambiguities, where d corresponds to the distance between the two reception- antennae and A corresponds to the wavelength of the processed signal,
* The sensor will then transmit, 300, the following' data to the other
sensors:
• the estimation of the power that it has received,
• the estimations of the potential directions of arrival, DOA,
■ • its location coordinates (longitude, latitude)'and the orientation of its network of antennae, which data are known for example by virtue of a navigation system 211,
• These data (powers, directions of arrival, coordinates and orientation
of the network of antennae) are merged (combined), by the processor
of the sensor, with the data calculated in the other sensors that are
present-in the system, in the example given a second sensor, and thus
allow calculation of the locations of the radio emitter(s) using the steps
described in Figure 3.
These data are transmitted by virtue of the wireless communication module of each sensor.
In the example, the steps of Figure 3 are executed in each of the two sensors. Without departing from the scope of the invention, these steps could be executed in a device (external to the sensors) whose role would be to centralize the processing of the data.
The sensor R<| recovers the data associated with the second sensor R2 (the value of the estimated received power V2 aRd the measured directions of arrival DOA, 301, that were obtained by executing the steps described for the sensor R2). Reciprocally, the sensor R2 recovers the data associated with the sensor R<|.
On the basis of the first power value ?± estimated ?± by the first sensor R1 and of the second power value P2 estimated P2 by the second sensor R2l of a

10
propagation model and of the value of the ratio of these two powers, Ri/ R2, the processor determines an elementary location area Ze and then deduces from this a potential location area Zs of the emission source S (Zs is the intersection of all of the. elementary areas Ze determined by taking all of .the 5 possible pairs of sensors. In the case, in the example, where there are only two sensors, Zs=Ze). The processor calculates the intersections of half-lines of sight, 303, on the basis of the possible direction of arrival values and of the latitude and longitude of the antenna network of the sensors R-i and R2.
TheJb.llowing step consists in eliminating the intersections of the 10 half-lines that are not located in the location area Zs, 304, so as to eliminate some of the points of intersection P(Dk) corresponding to ambiguous lines of sight.
Next, the processor 12 will determine the densest cloud of points, 305, and calculate, for example, the barycentre of the intersections in the 15 cloud in question, 308. This barycentre corresponds to an estimation of the location of the emitter source, 307.
These steps are executed in each sensor of the communication system, so that each sensor is able to locate the emitter source S.
The steps that have just been described may, without departing
20 from the scope of the invention, be performed in a centralized device
(external to the sensors) that receives the aforementioned data from each of
the sensors, and that processes said data in a similar way so as to locate the
electromagnetic source(s).
The broadcasting of the measurements and of the measurement 25 results may be performed in multicast mode so as to distribute the data to all of the sensors that are present in the system and allow each of them to estimate the position of one or more electromagnetic sources locally.
If the measurements performed in the first processing phase are transmitted to a centralized processing device^ it is the latter that will estimate 30 the position of the electromagnetic source or sources that is or are present in an area.

11
In order to determine whether the signal received at a sensor and
processed by the latter ■ corresponds to a single emission source, the
processor of this sensor calculates for example the matrix of covariance of
. the received signal Rxx (2x2 matrix) on each time/frequency cell obtained at
5 the end of the time/frequency division step. The two eigenvalues of Rxx are
then calculated and compared with the noise level in order to deduce
therefrom whether there are zero, one or more than one source(s). The noise
level is estimated using statistics regarding the eigenvalues of all of the
channels, on the instantaneous band processed, using a_principle known to
10 those skilled in the art. Next, the processor deduces the eigenvalues of the
matrix Rxx by using methods known to those skilled in the art.
The differential phase between the two channels or antennae is deduced directly from the eigenvector associated with the eigenvalue corresponding to the signal (the eigenvalue that is detached from the noise 15 level).
Using the conventions of Figure 4, the phase difference between
the two antennae of a sensor has the value A

) of the emitter source by the two sensors-.-On the
■5 basis of the value of the ratio of these powers ?1/P2, the elementary location
area Ze of the source Is deduced by taking, as propagation model, for
example, a tree space' model with an attenuation proportional to the inverse
distance squared. If this model were to be exact, the elementary area Ze
-would be a circle. £p is the upper bound of the error applied to the estimation
10 of the power ratio ?±/P2- This means that:
5_[5L_ ^U 1
P2E[p2 £PT2 + £pJ "
where Pj_ and P? are? respectively, the estimations of Pt and P2.
The value of £p is chosen to be high5 so as to ensure that an upper bound of
the error is indeed involved, regardless of the actual propagation mode!
compared to the chosen propagation model.
15 The propagation model under consideration may also be a
propagation model that takes obstacles and the terrain into account. In any case, the propagation model is known to those skilled in the art, and will not be detailed.
Two circles C1, C2, respectively corresponding to ^-£p and to
p2
20 -^ + £p, are thus defined. The elementary area Ze is then the area contained
p2
between these two circles, outside the circle defined by -^ - £p and inside the
P2
p~~~ circle defined by ^+ £p, as indicated in Figure 5. In this case, the elementary
area Ze is a finite area.
Depending on the position of the two sensors and on the value of 25 £pj there are cases where the elementary area Ze is outside the circle defined

13
by -^ — £p and" outside the circle defined by ^ + £p, as indicated in Figure 6.
In this case, the elementary area is an infinite area.
The PwDOA algorithm determines whether the elementary area is
inside or outside the circle defined by ^ + £n. To this end, it observes
1 P2 p .
5 whether or not the centres CM, Cr2 of the two circles Ci, C2 are situated on
the same side of the two sensors (the centres of the circles are to the right of
the rightmost sensor or the two centres of the circles are to the left of the
leftmost sensor).
In practice, to determine the location area ZSl the geographical
10 area in which the sensors are moving is covered by a grid defined in an abscissa and ordinate reference frame. For each pair of sensors Ri, R2, the cells of the grid are contained within the elementary area Ze when at least one of their four corners is in the elementary area Ze. The location area Zs corresponds to the cells that are in ail of the elementary areas Ze (that is to
15 say in as many elementary areas_as there are pairs of sensors).
This principle may be extended, without departing from the scope of the invention, to other propagation models, such as for example propagation models utilizing a digital terrain model, in order to reduce uncertainties and therefore the potential location areas.
20 To calculate the intersections of the half-lines of sight (actual ones
and ambiguities), the processor of a sensor considers the geographical coordinates of the sensors and the estimation of the possible directions of arrival DOA. The equations for the half-lines of sight (half-line starting at the sensor and pointing in the direction of the DOA) are calculated, followed by
25 all of the intersections of these half-lines.
In order to locate the emission source(s), the method will detect the densest cloud of points (area with the highest concentration of points of intersection). In order to detect the area where the density of points is highest, histograms with two variables (abscissae and ordinates of the points
30 of intersection) are calculated using a method known to those skilled in the

14
art. The bins of the histogram are therefore rectangular parallelepipeds
whose height corresponds to the number of points of intersection contained
in the area defined by the area of the parallelepiped. The cloud of points in
: question corresponds to the points in the highest biR-of the histogram and in
5 the adjacent bins (so as to dispense with the problem of the centre of the
bins). The width of the bins is adaptive. The algorithm starts with a very fine
pitch and increases iteratively until the height of the bins is close to the height
expected in theory (number of times that N straight lines intersect, N being
the number of sensors).
10 Finally, to estimate the location of the emitter source, the
processor will consider, for example, the barycentre of the points of the densest cloud.
Figure 8 shows two exemplary implementations of the method according to the invention when the sensors are of software-defined radio
15 receiver type. The reception of the signals on the band to be inonitored may be performed in specially allocated TDM A time slots TS\ TS1 to TS5. Pseudorandom offsetting of the time slot(s) makes it possible in particular to avoid a stroboscopic effect on the observation of the communication system that has the same TDMA frame period as the device of the invention. Other
20 ways of proceeding may be contemplated, such as using the period between the increments of emissions with frequency hops.
The power of arrival (POA) and the differential phase (PhDOA) are calculated and estimated by detecting situations of single-signal emission through calculation of the eigenvalues of the intercorrelation matrix on the
25 two channels of each sensor. The eigenvector associated with the signal is then used for the calculation of the differential phase. These estimations may be performed over short periods of the order of 160 ps for a division of the instantaneous reception band into channels of 25 kHz. This short TSLT (time slot look through) observation period, and the synthetic data~that it gives
30 (PhDOA and PwDOA), make it possible to obtain an amount of information to be transmitted by sensor and thus to systematically locate all of the

intercepted emissions. This short observation period makes it possible to combine the 'sensing' function with the radiocommunication function necessary for the communication service, ensuring the exchanges of data for the benefit of the- 'sensing' function and the- communication needs of the users.
According^ to one variant embodiment, the method makes it possible to carry out automatic calibration of the two antennae of a sensor by using the known position of the other sensors that are present in the communication system and that are considered to be emitter sensors. The.. processor of a sensor then determines the calibration correction to be applied during the processing of the data.
The steps that have just been described apply to a system in which the sensors are software-defined radio receivers.
The method and the system according to the invention perform a distributed location of one or more radio wave-emitting sources. They make it possible in particular:
• to dispense with the ambiguities of dual-channel differential-phase goniometry,
• to locate emitters with short observation periods and small volumes of data to be exchanged between sensors, thus enabling systematic location of the intercepted emissions and a 'simultaneous' service formed of the 'sensing3 service and of the radiocommunication service.
The implementation of the invention does not require any scenario involving stationary of "the' position of the emitters and of the sensors. The invention may be implemented in existing software-defined radio receivers having two synchronous channels (MlMO or multiple input multiple output capability). The invention makes it possible to facilitate integration into a carrier by utilizing a network of two reception antennae having spatial reception diversity (antennae spaced apart by between less tharTone wavelength and several wavelengths).

CLAIMS

1 - Method for locating at least one electromagnetic source S within a communication network comprising at-least two sensors-Ri, R2, said sensors being separated by a given distance, each sensor comprising a network of antennae comprising at least two reception antennae Ai, A2, characterized in that it comprises at least the following steps:
• For each sensor R1s R2,
• Estimating the powers Px and P2 of the received signal Sr coming from a source S and received at the sensors R1f R2,
• For the signal Sr arriving at each of the two reception antennae, measuring the phase difference Acp on the basis of the value of the phase difference Acp and of information on the position and the orientation of the network of antennae Ai, A2l determining a set of potential directions of arrival DOA for a given source S, and then
• Merging the estimated power values Pj_, ?2 received at the sensors Ri
and R2, the potential direction of arrival DOA values and the geographical
coordinates of the networks of antennae of the sensors Ri and R2j in
order to calculate the coordinates of the emission source by executing
the following steps:
• On the basis of the power values f\3 V2 estimated by the two sensors Ri, R2, of a propagation model and of the value of the ratio of these powers Pj_, f2i defining, for each ratio value, an elementary location area Ze in which the source is present, and then a potential location area Zs corresponding to the intersection of the various elementary location areas for an emission source S,
• On the basis of the geographical coordinates of the sensors and of the possible directions of arrival, calculating the possible half-lines of sight and their points of intersection P(Dk),
• Retaining only those points of intersection P(Dk) of the half-lines of sight that belong to the potential location area Zs,

• Selecting the points of intersection P(Dk) for which the geographical density is highest, and determining the coordinates of the emission
source on the basis of these points of intersection.
2 - Method according to Claim 1, characterized in that the step of merging the data, estimated power values, direction of arrival DOA and coordinates of the antennae is executed in the processor (12) of each sensor of the communication network.
3 - Method according to Claim 1, characterized in that the step of merging the data is executed in a centralized processing device communicating with the sensors that are present in the communication network.
4 - Method according to Claim 1, characterized in that it comprises a step of automatically calibrating the network of antennae of each sensor by using the known position of the other sensors while they are being used in emission mode.
5 - Method according to one of the preceding claims, characterized In that it comprises a step (202) of time segmentation and frequency segmentation of the signal received at the antennae, and a step (203) of detecting the number of sources present by calculating the covariance matrix of the signal at each time/frequency cell and by comparing the two eigenvalues of the covariance matrix with the noise level.
6 - Method according to one of the preceding claims, characterized in that it uses, as propagation model to determine an elementary location area of the source on the basis of the estimated power values, a Tree space' propagation model.

7 - Method according to one of the preceding claims, characterized in that the measurement periods on the signal received from the sensors use the TDMA access method, and in that a specific TSLT slot is used to execute the steps of the method accordin-g to the invention. ■ —-
8 - System for locating at least one emission source in a communication network comprising at least two sensors Ri, R2 communicating with one another by way of a radio link, each sensor comprising a network of antennae comprising at least two reception antennae A1? A2) a sensor comprising a processor (12) designed at least to process the received signals so as to determine the power values of the signal received at the antennae, the potential directions of arrival and the phase offset value of the signal received at the two antennae, the location system comprising a processor designed to execute the following steps:

• Merging the power values received at the sensors, the potential direction of arrival values and the coordinates of the network of antennae, in order to calculate the coordinates of the emission source by executing the following steps:
• On the basis of the power values Pj_, P2 estimated by the two sensors R1, R2? of a propagation model and of the value of the ratio of these powers Pj_, B?3 defining, for each ratio value, an elementary location area Ze in which the source is present, and then a potential location area Zs corresponding to the intersection of the various elementary location areas for an emission source S,
• On the basis of the geographical coordinates of the sensors and of the possible directions of arrival, calculating the possible half-lines of sight and their points of intersection P(Dk),
• Retaining only those points of intersection P(Dk) of the half-lines of sight that belong to the potential location area Zs,

• Selecting the points of intersection for which the geographical density is highest, and determining the coordinates of the emission source on the basis of these points.
9 - Location system according to Claim 8, characterized in that a sensor is a software-defined radio receiver.