The present invention relates to a radar signal deinterleaving process, the method comprising: - the reception of electromagnetic signals by a receiver and extracting the pulses of the received signals, and - pulse train formation consisting of at least three spaced pulses of the same pulse repetition interval, each pulse train being defined by the pulse repetition interval. The present invention also relates to a corresponding deinterleaver. One of the challenges of electronic warfare is to intercept radio transmissions from detection systems such as radar transmitters. The presence of many transmitters that the intercepted signals are interlaced, that is to say that signals from a radar transmitter interest are interfered with by other signals from the ambient electromagnetic environment. It is therefore necessary to deinterlace the intercepted signals to separate the different signals from different radar emitters. However, the signals transmitted by the same radar transmitter may have features, defining a waveform, variables, especially in terms of range of pulse repetition signals or carrier frequency of the signal pulses, making the process complex deinterlacing. A great wealth of waveforms of the electromagnetic world corresponds to a wide variety of treatments deinterleaving to extract pulses of the same waveform of the ambient electromagnetic environment. More particularly, the technical field, the object of this method relates to the deinterleaving waveform whose pulse repetition interval is medium or short (up to several hundreds of micro-seconds). These waveforms are generally constituted of multiple trains of pulses. It is known to use extractors radar signals implementing a signal deinterlacing process radar in two stages. The first step consists of pulse trains of the formation from all intercepted pulses. The second step is to combine the pulse trains formed to obtain deinterleaved radar signals. The first pulse train forming step uses the statistical information of the intercepted signals such as frequencies of the carriers of pulse signals, the pulse repetition interval and the arrival directions of the pulses. Second pulse trains regroupment step regroups pulse trains trained following proximity to form deinterleaved signals. However, existing algorithms pulse train grouping does not offer the same level of maturity as those of pulse trains training. In particular, the pulse trains of the grouping step is generally approached as a problem of "clustering" or "classification of data", wherein each train is compared to another at a single distance criterion. However, the waveforms formed by the pulse train are of great diversity, some waveform families may have totally antagonistic characteristics therebetween. Therefore, a single distance criterion can lead to erroneous reconstruction of signals. The technical problem relates to the grouping of pulses from the same dense electromagnetic environment in radar signal trains where several distinct waveforms can occur simultaneously, the difficulty of not perform bad pulse train combinations. US 201 1/0150053 A discloses a method and a detection signal of a radar apparatus. The method includes collecting a plurality of pulses based on a reference signal. The method also comprises the classification of the pulses into groups based on the similarity of pulse widths. Article of Tuesday At HK entitled "New technologies for the deinterleaving of repetitive sequences" published on 1 st August 1989 in IEE Proceedings F Communications, Radar & Signal Processing, Institution of Electrical Engineers, Volume 136, Number 4, Part F, page 149-154, describes an algorithm for deinterleaving quickly and accurately several repetitive signals. There is therefore a need for a radar signal deinterlacing process for grouping pulse with improved reliability trains, limiting the risk of obtaining a false interlaced signal, while being quick implementation. To this end, the invention relates to a radar signal deinterleaving method of the aforementioned type, wherein the method further comprises: - the pulse trains grouping having the same interval of repetition of the pulses according to a predefined grouping of law to form pulse levels, and - the combination of the following pulses of levels at least one predetermined law of association to obtain deinterleaved radar signals formed from the concatenation of pulse trains of pulses associated bearings. According to particular modes of implementation, deinterlacing method includes one or more of the following characteristics, taken individually or in all technically possible combinations: - each pulse train is also defined by at least one element selected from a group consisting of: time of arrival of the first pulse of the pulse train arrival time of the last pulse of the pulse train , the frequency of the pulse train of pulses, the pulse train of the pulse duration and the direction of arrival of the pulse train pulses. - the method comprises prior to the combining step, a pulse train classification step according to their carrier frequency to produce two pulse trains classes: a class containing fixed carrier frequency pulse trains and the other class containing the variable carrier frequency pulse trains, the combining step is implemented for each of the two classes of pulse trains and for obtaining single-frequency pulse levels from the class of mono-frequency pulse trains and agile pulse frequency levels from the class of agile pulse train frequency. - the associating step comprises a phase of consolidation bearings of pulse repetition intervals of the different pulses which are linked in time to obtain pulses in switching levels of groups. - the associating step comprises a phase of consolidation of fixed carrier frequency pulses bearings, having repetitions of identical pulses and intervals superimposed over time for groups of bearings pulse recovery. - each of the laws of grouping and association is implemented by at least one algorithm for obtaining groups from elements, the elements designating pulse trains at the grouping step and bearings of pulse during the association step, groups designating pulse level during the step of grouping and groups of bearings pulse during the association step, the algorithm comprising: o selecting a reference from a set of elements, o Removal of the reference member of the set of elements and adding, in a plurality of groups, a reference group with the reference element, o selection in the set of elements, elements compatible with the reference group against a set of criteria for a set of candidate elements, o Evaluation of the distance between the reference group and each member of the set of candidate elements, o the annexation of part of the set of candidate elements minimizing a distance to the reference group and the removal of the annexed part of the set of elements, o repeating the selection phases, evaluation and annexation as the set of candidate elements comprises elements, and o the repetition of the preceding steps until the set of elements contains elements. - the reference element is the element of the set of elements of which the arrival time of the first pulse is the smallest. - the set of criteria to evaluate the compatibility of the elements of the set of elements with the reference group based on one or more characteristics, the characteristics being selected from a group comprising: the direction of arrival of the elements, the chronological overlapping of the elements, the carrier frequency of the elements, the elements of the pulse width, the elements of pulse repetition interval, the phase of elements and the number of pulse elements. - the criteria are selected according to statistics on the characteristics of radar waveforms of a database. - for the consolidation law, the distance is the difference in time between the last pulse of the reference group of the first pulse of the candidate member of the set of candidate elements, and the law of association, distance is a recovery rate between the reference group and the candidate of the set of candidate elements or score element to select from the set of candidate elements element sharing the most common characteristics with the group reference. - the method comprises prior to the combining step, a step of discarding inconsistent pulse trains in terms of the pulse repetition interval. - the method comprises prior to the combining step, a pulse train rejection step whose pulse repetition interval is greater than a pulse repetition interval threshold and whose number of pulses is less than a pulse number threshold. - associating step comprising at least one grouping of phase selected from: o a phase of consolidation of repetition intervals pulses bearings of different pulses and which are linked in time to obtain groups of bearings pulse switching, and o a phase of consolidation fixed carrier frequency pulse bearings having repetition intervals of identical pulses and in overlapping time for groups of pulse recovery levels. The invention also relates to a radar signal deinterleaver comprising: - a clean receiver receiving electromagnetic signals, - a digital processing unit of the own signal to extract the pulse signals received by the receiver, and - a readable information carrier on which is stored a computer program comprising program instructions, the computer program being loadable on a data processing unit and adapted to drive the implementation of such a method as described above when the computer program is implemented on the data processing unit. Other features and advantages of the invention will become apparent from reading the following description of the invention embodiments, given by way of example only with reference to the drawings, which are: - Figure 1 a schematic representation of a radar signal deinterleaver for the implementation of a deinterleaving method of the invention, - Figure 2, a flowchart of an exemplary implementation of a deinterleaving method of the invention, - Figure 3, a schematic representation of three pulse trains, - Figure 4, a schematic representation of a pulse train and quantities defining the pulse train, - Figure 5 is a diagrammatic representation of a signal and a pulse bearing resulting from the combination of three pulse trains of this signal, - Figure 6 is a diagrammatic representation illustrating the grouping of two pulse trains at a grouping step of de-interleaving method according to the invention, - Figure 7, a flowchart of the operation of an algorithm implemented in the deinterleaving process of the invention, - Figure 8, a schematic representation of an elimination phase of incompatible elements with a reference group, - Figure 9 is a diagrammatic representation of a meeting phase of a compatible component with a reference group when the distance between the element and the reference group is minimal compared to other compatible components, - Figure 10 is a diagrammatic representation of a group of bearings pulse formed from the pulse levels of one form of FMICW type of wave, - Figure 1 1, a schematic diagram illustrating the combination of three levels of pulses to form pulse groups bearing at another associating step of deinterleaving method according to the invention, - Figure 12 is a diagram illustrating the association of eight levels of pulses to form two pulse levels of groups in an association step of the deinterleaving method according to the invention, and - Figure 13, a schematic representation of the radar signals deinterleaving process from receiving signals by a receiver to obtain deinterleaved signals. A device 10 for radar signals deinterleaving is shown in Figure 1. The device 10 of radar signals deinterleaver is adapted to implement a deinterleaving method of radar signals. Data input 10 of deinterleaving device are measurements of the characteristics of pulses received by the device 10. The features are, for example, the carrier frequency of the pulses, pulse width, pulse power, the direction of arrival pulses or periods of pulse repetition. Such pulse characteristics measurements are taken, for example: - the reception of radar signals by a receiver, then the scanning signals and the extraction of pulse signals by a digital signal processing unit, - the reception of analog signals generated in the laboratory to simulate reception of radar signals and scanning signals and the extraction pulse signals by a digital signal processing unit, - generating, via software, digital data simulating the reception and digitization of radar signals, and the extraction pulse signals by a digital processing unit of the signal, or - generating, via software of pulses which are then recorded on a receiving interface. As illustrated in Figure 1, the device 10 comprises a receiver 12 of electromagnetic waves, a computer 14 and a computer readable information carrier 16 interacting with the computer 14. The receiver 12 is adapted to receive electromagnetic signals from, for example, radio-detecting systems such as radar. The electromagnetic signals are, for example, from issuing radars are analog signals created in laboratory and simulated radar signals. The receiver 12 is connected to the calculator 14 and is adapted to send signals received by the receiver 12 to the computer 14. The receiver 12 is, for example, an antenna. For example, the receiver 12 is an antenna element, a network antenna, reflector antenna, a circularly polarized antenna, a waveguide antenna, an active antenna, a shortened antenna, a broadband antenna , a patch antenna, a frame or a loop antenna or an antenna system consisting of one or more of the preceding antennas. The calculator 14 is adapted to receive signals from, for example, receiver 12 or digital data from simulation software. The computer 14 is a computer having a processor 18 and, optionally, a man-machine interface 20 and a display unit 22. The calculator 14 further comprises a digital signal processing unit 23 and optionally an interface 32. The processor 18 includes a data processing unit 24, memory 26 and a reader 28 of the information carrier. The reader 28 of the information carrier is suitable for receiving and reading the readable information carrier 16. The drive 28 of information carrier is connected to the data processing unit 24.The readable information carrier 16 is a medium readable by the reader 28 of the information carrier. The readable information medium 16 is a medium suitable for storing electronic instructions and capable of being coupled to a bus of a computer system. For example, the readable information carrier 16 is a floppy disk or disk (the English name floppy disk), an optical disk, CD-ROM, a magneto-optical disk, a ROM, a RAM , EPROM memory (acronym of the English Erasable Programmable Read-Only memory), an EEPROM (acronym for EEPROM), a magnetic card or an optical card. On the readable information carrier 16 is stored a computer program product comprising program instructions. The computer program is loadable on the data processing unit 24 and is adapted to drive the implementation of a radar signal deinterleaving method according to the invention. The man-machine interface 20 is, for example, a keyboard. The display unit 22 is, for example, a screen. The digital signal processing unit 23 is configured to digitally process the signals received by the computer 14. More specifically, the digital signal processing unit 23 is configured to digitize the received signals, extracting the signals of the pulses and measuring characteristics of each extracted pulse. Alternatively, the digital signal processing unit 23 is also configured to generate digital data simulating the receipt and scan radar signals or to directly generate pulses. The interface 32 allows on one hand to store the pulses resulting from the digital signal processing performed by the digital signal processing unit 23 and on the other hand to receive previously stored pulse or receiving pulses generated by a simulation software. The operation of the radar signal deinterleaver is now described with reference to Figure 2 which is a flowchart of an exemplary implementation of a radar signal deinterleaving method according to the invention. The deinterleaving method comprises a receiving step 100 a plurality of electromagnetic signals by the receiver 12, scanning of the received signals and extracting pulses h, I m electromagnetic signals digitized by digital signal processing. Digitizing the signals and the extraction pulse h, I m are performed by the digital signal processing unit 23 and are optionally stored in the interface 32. Such pulses h, I m electromagnetic originating in, or representative, for instance, signals transmitted by radio detection systems such as radar. Alternatively, the pulses \ l m is already stored in the interface 32. The steps of de-interleaving method described in the following are implemented by the computer 14 by interaction with the readable information carrier 16. The deinterleaving method then comprises a step 1 10 of ΤΊ pulse train formation, ..., T n from the \ I pulses m received by the receiver 12 and recorded in the interface 32. During the step of formation 1 10, at least three pulses h, l 2 are combined to form a pulse train T x . The training test pulse trains T \ , ..., T n consists in grouping in a single train of pulses T x , the pulses h, I m having the same repetition interval PRI pulses. It is understood by the term "repeat interval pulses," the period between two consecutive pulses. In other words, the pulses h, I m grouped together in one pulse train T x are such that any two consecutive pulses of the pulse train T x are spaced the same repetition interval PRI pulses that two pulses any consecutive pulse train T x . Pulse trains T 1; ..., T n formed constitute a set S T of pulse trains T 1; T n . Pulse trains T 1; T n formed are defined by the repetition interval PRI pulses between two consecutive pulses of the pulse train T 1; T n . It is illustrated in Figure 3, three pulse trains T 1; T 2 and T 3 having repetition intervals of the pulses, PRI 1; PRI 2 and PRI 3 , possibly different from each other. As illustrated in FIG 4, a pulse train T x is also defined by the arrival time of the first pulse of the pulse train T x , called TOA déb and the arrival time of the last pulse the m of the pulse train T x , called TOA end . In addition, each pulse being T x is optionally defined by at least one element selected from a group consisting of: pulse frequency h, I m of the pulse train T x , the pulse duration h, I m of pulse train T x and the direction of arrival of the pulses h, I m of the pulse train T x . The deinterleaving method then comprises a step 120 of rejection of ΤΊ pulse trains, T n inconsistent. The rejection step 120 verifies the consistency by repetition interval PRI ΤΊ pulses of pulse trains, T n formed during the step of formation 1. 10 Indeed, some pulse trains T x may have been formed by the mixture of at least two repetition intervals PRI close yet different pulses. A pulse train T x incoherent in terms of the pulse repetition interval PRI is, for example, detected by a statistical test based on the intervals between all pulses of the train T x . A khi2 test can for example be used. The pulse trains T x deemed inconsistent in terms repetition interval PRI pulses are eliminated from the set S T of pulse trains During this rejection step 120, optionally, the consistency of ΤΊ pulse trains, ..., T n in terms of carrier frequency of the pulses is also verified. It is understood by the term "carrier frequency of a pulse," the carrier frequency of the pulse, the carrier being a modulated wave by an input signal. Indeed, some T pulse trains x may have been formed by mixing at least two carrier frequencies near but nonetheless different. A pulse train T x incoherent in terms of carrier frequency is, for example, detected by a statistical test based on the frequencies of the pulse train T x . A khi2 test can for example be used. T pulse trains x deemed inconsistent in terms of the carrier frequency pulses are eliminated from the set S T of pulse trains T 1; T n . In the case where the repetition interval PRI pulses, carrier pulse frequencies respectively, are modeled as Gaussian and independent variables therebetween, Consistency repetition interval PRI pulses in carrier frequency of the pulses respectively, is evaluated by a statistical test of chi-square. It is understood by the term "chi test two," abbreviated as "test χ 2 " or "khi2 test ', a statistical test to test the suitability of a series of data to a family probability laws or test the independence between two random variables. The deinterleaving method then comprises a step 130 of pulse trains of the discharge T x does not belong to the category of waveforms of high typical or average repetition frequency. A waveform includes ΤΊ pulse trains, ..., .tau η with common characteristics in terms, in particular, repetition interval PRI pulse carrier frequency pulses and number of pulses. For example, all ΤΊ pulse trains, ..., .tau η having repetition intervals PRI short pulses and a large number of pulses constitutes a waveform of high or medium recurrence frequency category. AT reverse, the set E T of ΤΊ pulse trains. ,. , Τη having high repetition intervals PRI pulses and a number of low pulses is another waveform of low repetition frequency category. The rejection step 130 comprises defining thresholds characteristics of a waveform: a repetition interval threshold S pulses PR | and a number of S pulses of threshold | MPU | If ons- The repetition interval threshold S pulse PR | and the number of S pulses of threshold | MPU | If ons are defined from a database representative of a number of radar waveforms of interest. In an alternative embodiment, the database of waveforms is stored in the memory 26 of the processor 18. In another alternative embodiment, only the thresholds determined from the database are stored in the memory 26 of the processor 18. The threshold S PR | is selected to be greater than the values observed predominantly in the database of waveforms. The repetition interval threshold S pulses PR | is, for example, understood broadly between 1 microseconds (με) and 1 milliseconds (ms) to test membership in forms of high type waves or average repetition frequency. The threshold S | MPU | If ons is selected to be less than the values observed predominantly in the database of waveforms. The number of pulses Simpuisions threshold is, for example, including in the broad sense between 1 pulse and 100 pulses for testing belonging to the waveforms of high or medium recurrence frequency category. In this case, such repetition interval thresholds S pulses PR | and number of pulses S | MPU | If ons allow the exclusion of pulse trains T 1; ..., T n low repetition rate guy. The rejection step 130 then includes the rejection of pulse trains T 1; T n whose repetition interval PRI pulses is greater than the repetition interval threshold S pulse PRL The rejection step 130 also includes the rejection of Ti pulse trains, ..., T n the number pulse is less than the pulse number threshold If MPU isions. The repetition interval threshold S pulses PR | and the pulse number threshold If MPU i S ions thus form a template so that any pulse train T x out of the template is rejected. Thus, the pulse trains T x rejected are removed from the set S T of ΤΊ pulse trains, ..., T n and are further treated by another specific deinterleaving process. The deinterleaving method then comprises a step 140 of classification of pulse trains T 1; T n . The classification step 140 consists of a sort of pulse trains T 1; T n of the set E T of ΤΊ pulse trains. ,. , Τη following the carrier frequency of the pulses of each pulse train T 1; T n . In this classification step 140, the pulse trains T 1; T n are divided into two classes, C 2 pulse trains T 1; T n . The first class includes pulse trains T 1; ..., T n fixed carrier frequency, that is to say the ΤΊ pulse trains, T n formed of pulses having the same carrier frequency to a measurement uncertainty near. Measurement uncertainty is, for example, equal to plus or minus 5 percent (%) of the carrier frequency. The ΤΊ pulse trains, ..., T n fixed carrier frequency are also called ΤΊ pulse trains. ,. , Τη single frequency. The second class C 2 of ΤΊ pulse trains, ..., T n gathers the ΤΊ pulse trains, T n variable carrier frequency, that is to say the ΤΊ pulse trains. ,. , Τη containing pulses of different carrier frequencies. The ΤΊ pulse trains, ..., T n variable carrier frequency are also called pulse trains T 1; ..., T n frequency agile. The carrier frequency of the pulses of pulse trains T 1; T n frequency-agile is generally random or pseudo-random. The deinterleaving method comprises then a grouping step 150 pulse trains T 1; T n of the set E T of pulse trains When grouping step 150, pulse trains T 1; T n having the same repetition interval PRI pulses are grouped according to at least one predefined grouping law to form pulse levels P 1; P p . It is understood by the term "pulse level", a set of pulses from at least one pulse train formed in step 150 of clustering. It is illustrated in Figure 5, an example of received signal R by the receiver 12 and a Pi pulse trains bearing formed by the grouping of pulses T 4 , T 5 and T 6 of this signal. Similarly, there is illustrated in Figure 6, a pulse level P 2 from the pulse trains grouping T 6 and T 7 . Specifically, the step 150 of clustering of ΤΊ pulse trains, T n is implemented on the one hand, to the pulse trains T \ , ..., T n fixed carrier frequency during a first sub-step 160 to obtain pulse levels P 1; P p of fixed carrier frequency. The grouping of law used is called "fixed combination of law." The step 150 of clustering of pulse trains T 1; T n is implemented, secondly, to pulse trains T 1; T n variable carrier frequency during a second sub-step 170 to obtain pulse bearings PP p of variable carrier frequency. The grouping of law used is called "variable grouping of law." Pulse trains T 1; T n grouped in the clustering step 150 have not been put together to form a single train of pulses T 1; ..., T n in the training step 1 10 For such pulse trains T \ , ..., T n exhibited, for example, mitages M, that is to say pulses missing due to technical reasons. These technical reasons depend in particular on receiver's listening limitations 12, the quality of the measurements from the digital signal processing unit 23, the proximity of the receiver 12 relative to the transmitter, the receiver 12 relative positioning to the transmitter and electromagnetic disturbances. The mitages are, for example, shown in the pulse trains T 6 and T 7 . Missing pulses are reconstructed during the formation of the pulse level P 2 . In general, a grouping law consists in an algorithm whose general structure is described with reference to Figure 7. Initially, the algorithm comprises a phase 150A choosing a reference e ref from a set of elements E. The algorithm comprises, then, 150B a phase deletion of the e reference ref of the set of E and additional elements in a set of groups E g , a reference group g ref comprising e reference ref . The algorithm comprises, then, 150C a phase selection in the set of elements E, material consistent with the reference group g ref according to a set of criteria C. elements compatible with the reference group g ref form a set of candidate elements E c . Criteria C are, for example, chosen according to statistics on the characteristics of radar waveforms of the database of waveforms. For example, Figure 8 illustrates a group G 3 already formed and a group G 4 being formed which is therefore at this point the reference group g ref of the algorithm. Ei elements, e 2 , e 3 and e 4 are members of the set I of the algorithm elements. As shown by a cross in the figure 8, the elements ei and e 3 are inconsistent with the group G 4 training. Conversely, as shown by the check mark in this figure 8, the elements E 2 and E 4 are compatible to be combined with the group G 4 training. The set of candidate elements E c of the algorithm therefore has two components: E 2 and E 4 . The algorithm comprises, then, a stage 150D evaluation of the distance between the reference group g ref and each element of the set of candidate elements E c . Then, the algorithm comprises a phase 150E annexation of an element of the set of candidate elements E c to the reference group g ref . The element attached is removed from the set of elements E. The element of the set of candidate elements E c together with the reference group g ref is the element with the smallest distance d. It is understood by the term "annexation of a part of a group", the connecting element to the group: after annexation the final group consists of the original group and the attached element is joined to the original group. For example, Figure 9 illustrates the determination of the element to meet at footnote g Group ref , that is to say to the group G 4 training, among the set of candidate elements E c . FIG 9 includes the same groups and elements as those of FIG 8, but shown in a different way. Thus, the group G 3 already formed comprises three components represented by three rounds. The group G 4 being formed already comprises four elements represented by four rounds. As shown in Figure 9, the distance d 2 between the element e 2 of the group G 4 is less than the distance d 4 between the element e 4 of the group G 4 . Therefore, the element of the set of candidate elements E c which minimizes the distance to the reference group g ref is the element e 4 . The element e 4 will be reunited with the group G 4 training. The phases of selecting 150C compatible elements of distance evaluation 150D and 150E annexation are then repeated with the remaining elements of the set of elements E, so that all the candidate elements E c calculated in the selection phase of elements 150C compatible, includes elements. Finally, phase 150A choice, delete and add 150B, 150C selection of compatible elements of distance evaluation 150D, 150E and annexation are repeated until all elements of E contains elements . The fixed combination law applied in the first sub-step 160 is specifically described in what follows. Each element of the algorithm is a ΤΊ pulse train T n of first class Ci ΤΊ of pulse trains, T n fixed carrier frequency and each group of the algorithm is a pulse Pi bearing, P p fixed carrier frequency. The reference group g ref of the algorithm is a Pi pulse bearing, P p . In addition, the initial set of elements E of the algorithm is the set of pulse trains T 1; T n of the first class Ci T pulse trains 1; T n of fixed carrier frequency. In Phase options 1 50A, the item selected from the set of elements E is the element whose arrival time of the first pulse is the smallest. In other words, the selected element is the element of the set of elements E arriving on the first receiver 12 or the recorded member on the interface 32 with the arrival time of the first pulse TOA smallest. The set of criteria C comprises the following criteria: a criterion of direction of arrival, a temporal criterion, a frequency criterion, a criterion of pulse width, a pulse repetition interval criterion and a phase criterion. Only the elements of the set of elements E satisfying all criteria of the criterion set C are compatible with the g Reference Group ref and are added to the set of candidate elements E c . Alternatively, the set of criteria C does not include all of the criteria or include different criteria from those statements. The criterion of direction of arrival is used to test the compatibility of each element of the set of elements E with the reference group g ref depending on their direction of arrival DOA. The criterion of direction of arrival states that to be compatible, the g Reference Group ref and the element to be formed of pulses having the same direction of arrival DOA. The criterion of direction of arrival is checked for an element of the set of elements E when the element satisfies a comparison test of the arrival directions DOA of the element and the reference group g ref . The comparison test is based on a statistical model of the arrival direction DOA and is adapted to take into account the proportion of outliers. To check the comparison test, the element of the set of elements E must satisfy the following equation eqi: (DOA, - DOA 2 ) 2 $ 2 * * * (+) * where DOA, is the arrival direction of the reference group g ref , DOA 2 is the arrival direction of the element E of the set of elements to be tested, σ 2 is the variance of the direction of arrival of the pulses r is the proportion of direction of arrival measurements degraded, ni is the number of pulses present in the reference group g ref , n 2 is the number of pulses present in the element of the set of elements E to be tested, erf "1 is the reciprocal of the error function, the error function is given by the equation erf (x) = j = fe ~ t2 dt, and P (H 0 \ H 0 ) is the probability of detection of equality between ΌΟΑ chi and DOA 2 , H 0 denotes the hypothesis that the DOA values 1 and DOA 2 are equal, and P (H 0 \ H 0 ) denoting the probability of choosing H 0 knowing that it is in the case H 0. Alternatively, another comparison test consists of calculating and comparing, the average arrival direction DOA of the pulses of the reference group g D f with respect to the average arrival direction DOA of the pulses of each element of the set of elements E. Alternatively, yet another comparison test is to calculate and compare the median arrival direction DOA of the pulses of the reference group g ref with respect to the direction of arrival median DOA pulses of each element of the set of elements E. the use of the median rather than the mean avoiding the inclusion of strongly present outliers in measurements of DOA DOA. The temporal criterion to test the temporal compatibility of each element of the set of elements E with the reference group g ref . Temporal criterion stipulates that a member of the set of elements E which overlaps in time with the reference group g re f is inconsistent with the reference group g re f. Temporal criterion further stipulates that a gap too much time between an element of the set of elements E and the reference group g re f also results in a mismatch of the element with the reference group g re f. Indeed, a significant gap in time between the reference group g re f and the element can mean that there was no pulses. In this case, the reference group g re f and the element, even if they belong to the same transmitter, are distinct. The maximum difference in time between the reference group g re f a compatible member of the set of element E is, for example, a multiple of the repetition interval PRI pulses via the reference g Group ref . The difference in time is maximum, e.g., equal to twenty times the average value of the repetition interval of the pulses of the reference PRI g Group ref . The multiple of the value is set empirically, or from the database of waveforms. Alternatively, other multiple values or other threshold values not dependent on the PRI impulse repetition interval can still be envisaged. The frequency criterion used to test the compatibility of frequency elements of the set of elements E with the reference group g ref . The frequency criterion states that to be compatible two elements must be formed of pulses having the same carrier frequency f. The frequency criterion is verified for an element of the set of elements E when the element satisfies a comparison test of the carrier frequencies of the pulses of the element with the carrier frequencies of the pulses of the reference group g ref . The comparison test compares the average carrier frequency of the pulses of the reference group g ref with respect to the average carrier frequency of the pulses of each element of the set of elements E. Such a comparison test based on the fact that a frequency measurement follows a Gaussian model known variance. However, assuming the carrier frequency measurements are obtained with the help of frequency windows, the possibilities for excluding certain frequency values are expected. In this case, the distribution of measured carrier frequencies is no longer Gaussian, because partly truncated, and therefore the comparison of the mean is biased. Alternatively, particularly in the case where the frequency distribution is no longer Gaussian, a test of χ 2 is used. To check the frequency criterion, the element of the set of elements E must satisfy the equation eq 2 below: ^ ^ + 71,4 / 2,4 n + Π in X rt 1 + n 2 -l, PH 0 \ H 0 ) = 0 = 0 1 2 where f u is the carrier frequency of the order of pulse i of the reference group g ref , f 2 , i is the carrier frequency of the order of pulse i of the element of the set of elements E to be compared, ni is the number of pulses of the reference group g ref , n 2 is the number of pulses of the element of the set of elements E to be compared, is the variance of the distribution of carrier frequency measurements of the pulses, x iV is the quantile x p for the probability p of the distribution of chi2 to v degrees of freedom, that is to say that if a random variable X follows a law of chi2 to v degrees of freedom, the probability of X