Method, device and program for demodulating a data frame

Abstract

The invention relates to a method for demodulating a signal transmitted on a transmission channel on which is received (200) a frame of binary words modulated by a particular symbol from a dictionary of symbols, the method comprising steps of determining (201) a set of relative distributions representative of a difference between symbols and the frame; determining (202), for each relative distribution of the set, a partial relative distribution by relating a first relative distribution determined from symbols and with a second relative distribution determined from a symbol and the symbol; consolidating (203) one of the relative distribution by recombination of the determined partial relative distributions; and demodulating (205) the frame from at least a part of the set of consolidated relative distributions. Figure for the abstract: Figure 2

Description

Title of the invention: Method, device and program for demodulating a data frame Technical field

[0001] The invention generally belongs to the field of telecommunications and more particularly relates to a method for improving the demodulation performance of a data frame, in particular in communication systems which rely on a relative representation of the information for the purpose of its transmission and / or its processing. Prior art

[0002] Communication systems are known that rely on a relative representation of information for the purpose of its transmission and / or processing. A common example is the use of differential modulations.

[0003] Consider for example a frame of 5 symbols [xq, _x15 x2, Xy x4], each encoding k bits of information, and thus able to take 2^ values. A possible differential representation of the frame can be [x0 / i, x1 / 2, x2 / 3, X3 / 4 ] “ [x0- x{, x}-x2, x2-X4-X3]. In this example, it is therefore the difference between the values ​​that is coded, and a single absolute reference is sufficient to decode the complete frame. Of course, in such a configuration, the subtraction, or any other operation, is done in the finite field considered, here F^ (i.e. modulo k).

[0004] Such a representation can be used for signal transmission or processing purposes within the communication chain. For example, when the variations between consecutive samples are small, transmitting differences between samples can be particularly effective.

[0005] Such a differential representation of the frame may also prove useful in certain receivers, for example when receiving a signal modulated by the cyclic rotation of a root sequence of complex symbols, such as a CCSK modulation (for Cyclic Code-Shift Keying in English).

[0006] CCSK modulation proposes to modulate the data to be transmitted by the cyclic rotation / shift of a sequence of complex symbols called the root sequence. The root sequence is such that its shifted versions are orthogonal to each other, that is to say that it offers a good autocorrelation function. Thus, each binary word to be transmitted is associated with a particular cyclic shift of the same root sequence.

[0007] For example, by choosing a root sequence composed of the 4 symbols complex [a; b; c; d], CCSK modulation allows 2 bits of information to be modulated orthogonally via the rotations of this sequence. The binary word '00' can correspond to the transmission of the sequence [a; b; c; d]; the binary word '01' may correspond to the transmission of the sequence [d; a; b; c]; the binary word '10' may correspond to the transmission of the sequence [c; d; a; b], etc.

[0008] Knowledge of the root sequence allows the receiver to demodulate each symbol of the frame by cross-correlation with the root sequence. The demodulation can for example be carried out by cross-correlation of the received sequences with a particular sequence of the frame, for example a pilot sequence, the offset of which with the root sequence is known to the receiver. Thus, it is the difference between the respective offsets of the sequences of the frame which allows the demodulation.

[0009] It is thus necessary to have a good knowledge of the differences between the sequences of the same frame to carry out such demodulation. However, it happens that disturbances do not affect the symbols transmitted in the same frame in the same way, which can cause problems during its demodulation.

[0010] There is therefore a need for a method which makes it possible to improve demodulation performance, in particular when a communication is based on a relative representation of the information. Summary of the invention

[0011] For this purpose, a method is proposed for demodulating a signal transmitted on a transmission channel on which a frame comprising N information words is received, a word Pi with i G [Q, N - 1] being modulated by a particular symbol CPi from a dictionary of K symbols, the method comprising the following steps: - Determination of a set E of probability distributions, called relative distributions, a relative distribution P^j being representative of the respective probabilities of a plurality of possible values ​​of the difference between symbols CPi and Cpj from a plurality of pairs (CPi; CPj) formed from the symbols making up the frame, with i, je [O, N - 1] - For each relative distribution P^j of the set E; - Determination of at least one partial relative distribution by relating a first relative distribution determined from symbols CPi and CPm with a second relative distribution determined from a symbol Cp, and from the symbol CPm, with i, j * m, and me [0; N -1], - Determination of a consolidated relative distribution P, by recombination of the determined partial relative distributions p^, Demodulation of the frame from at least part of the consolidated relative distributions.

[0012] Generally speaking, the difference between symbols, or between values ​​associated with the symbols, corresponds in this document to a difference between digital representations associated with these symbols. The differences calculated between symbols may correspond to differences between indexes associated with the symbols or differences between the shifts of the sequences composing the symbols in the case of CCSK type modulation, or even to a difference in angle or phase in the case of QPSK modulation (Quadrature Phase Shift Keying). Such differences are calculated according to modular arithmetic, in a finite group equipped with an additive law and a neutral element, for example in the group mL) mKL.

[0013] It is thus proposed to determine a differential representation of the frame in which we associate with each pair of symbols that can be formed with the symbols of the frame, a vector similar to a discrete probability distribution, the size of which is equal to K, the cardinality of the dictionary of symbols and each element of which is a probability that the difference between two symbols is equal to a particular value. In this document, a relative distribution Pj[j corresponds to such a vector of K probabilities determined for a pair of symbols (CPi; Cpj).

[0014] The determination of the relative distributions for each pair of symbols (CPi; Cpp of a frame of N symbols, with i G [0; N -1] and / e [ 0; N - 1 ], results in a representation of the frame in the form of a matrix of the probability distributions of the differences between symbols of the frame:

[0015] Po / o ■” Pn-i / q' P = : : .Pq / ni "■ Pna / ni.

[0016] Such a representation can be calculated for any frame of symbols without modifying the invention, independently of the type of modulation (CCSK, QPSK for Quadrature Phase Shift Keying, or Quadrature Phase Modulation in French, QAM for Quadrature Amplitude Modulation, or quadrature amplitude modulation in French, etc.), and more broadly for any sequence of discrete random variables.

[0017] Consider for example the sequence of discrete random variables X = [X] X2 ... G of distribution law Dx = [ Dyt I)\ . ... Dx;V ( |. It is possible to express the matrix containing the set of differential random variables y F'^xA tc' O116

[0018]

[0019]

[0020]

[0021]

[0022] Yy = X^ - Xj E Fy and the matrix of corresponding distributions DY such that Dy . = Dy ® Dy where ® is the cross-correlation operator. Given the similarity of the notions of sum and difference in the context of modular arithmetic in a finite field, we can also define the matrix containing the set of random variables ye tc' 9ue Yq = Xj + X j E and the matrix of corresponding distributions DY such that DY. . = Dy *Dy where * is this time the convolution operator. In this document, we denote by "Relative distribution DY / y" the distribution of the random variable resulting from the difference (or sum) of the two random variables X and F. It should be further noted that by construction, the diagonal of the matrix is ​​expected to contain distributions whose maximum is reached for index 0 because the difference between the index of a particular symbol with respect to itself is zero. Moreover, a form of symmetry is present in the matrix in the sense that the shift of symbol i with respect to symbol j corresponds to the inverse shift of symbol j with respect to symbol i. It is then proposed to determine relative distributions called partial distributions determined on the one hand from the relative distributions determined for the symbols CPi and CPm and on the other hand from the relative distributions determined for the symbols Cp / and CPm. In other words, an index difference between a first symbol and a second symbol is determined by transitivity, that is to say from a difference between the first symbol and a third symbol and from a difference between the second symbol and the same third symbol. A plurality of estimates of the difference between the first and the second symbol based on different symbols are thus obtained. The different partial relative distributions are then combined by product in order to obtain a consolidated version of a relative distribution P^j. The frame can then be demodulated from the consolidated information. For example, when the first symbol of the frame is known to the receiver by convention, then the consolidated difference between the different symbols of the frame and the first symbol, the Pi / o for i E [1 ; N - 1], allows improved demodulation of the other symbols. The general concept of the invention is thus based on a combination of the probabilities of differences between digital representations associated with the symbols of the frame by relying on the relationships between variables inherent in such a representation of the information, with the aim of improving the quality of the estimation of this relative information. The method significantly improves the performance of the de modulation of a frame, particularly in communication systems which rely on a relative representation of information for its transmission and / or processing.

[0023] According to a particular embodiment, the method is such that a symbol CPi is a sequence of complex values ​​obtained by applying a particular cyclic shift to a root sequence of size P, a sequence CPi being transmitted on K sub-carriers of an OFDM symbol.

[0024] The information to be transmitted is modulated by a cyclic shift of a root sequence known to the receiver. This is, for example, a CCSK type modulation using a Zadoff-Chu type root sequence.

[0025] The CCSK sequences (or symbols) are integrated into a CP-OFDM frame so that each element composing a CCSK sequence, called a chip, is transmitted on a separate OFDM subcarrier. In its simplest form, we then obtain a time-frequency frame in which each of the N time steps contains a complete CCSK symbol of size K distributed over all the subcarriers in frequency, as shown in [Fig.l].

[0026] The root sequence (and therefore all its cyclic shifts) being known to the receiver, the transmitted CCSK sequences can then serve as pilot sequences for synchronization and channel estimation, while carrying information.

[0027] According to a particular embodiment, a relative distribution Pj]j is a vector of size K in which each element is a probability that a difference between the symbols CPi and CPj of the frame corresponds to the value k.

[0028] According to a particular embodiment, the method comprises at least one step of normalizing the probabilities.

[0029] It may be necessary to normalize the distributions obtained in the different steps of the method so that their sums are equal to 1 in accordance with their probabilistic nature.

[0030] According to a particular embodiment, the step of determining a partial relative distribution comprises a cross-correlation operation between a relative distribution Pi / m and a relative distribution P j / m, or between a relative distribution Pm / i and a relative distribution Pmjj with i, jm and me [0; N - 1].

[0031] In other words, for relative distributions P ai b associated with relative variables of the form (A - B), it is proposed to define partial relative distributions according to one or more of the following operations, where ® represents the cross-correlation and Pajb the mirror distribution of Pb / a on the finite field:

[0032] P aie ® P b / c Pc / b ® Pe / a Pay ® P'c / b Pc / b ® Pa / e Pe / a ® Pb / c P b / c ® P c / a Pe / a ® Pc / b Pb / C®Pa / e

[0033] When the relative distributions between symbols are expressed by index differences in a finite field (i.e. modulo K for a dictionary of K symbols), the relating of the indexes of the symbols can be carried out by a cross-correlation operation.

[0034] According to a particular embodiment, the step of determining a relative distribution partial fp”) includes a convolution operation between a relative distribution Pi / m and a relative distribution Pm / j, with 4 j * m and ni e [0; N -1],

[0035] Thus, it is also proposed to define partial relative distributions by one or more convolution operations among the following operations, where * is the convolution operator and Pa[h the mirror distribution of Pb / a on the finite field:

[0036] 'P a / c" P c / b p(c) — Pa / h Paic*Pb / c P^c / b Pcia^Pb / e

[0037]

[0038]

[0039] Thus, the relating of relative distributions is done by a sum in a finite field (i.e. modulo K for a dictionary of K elements), i.e. by a convolution operation. In a particular embodiment, the method is such that the recombination step does not take into account the partial relative distributions determined from a relative distribution P^j and / or a relative distribution P Such an arrangement helps to guard against certain "self-influence loops" by masking the data so as to avoid updating a relative distribution from data from this same relative distribution (or from the mirror distribution). This kind of phenomenon can occur in an iterative algorithm when updating the value of a variable involves the intrinsic information of this variable (i.e. information from the observation of the variable that we wishes to update) rather than extrinsic information (i.e. information derived from the observation of other variables linked to it).

[0040] It is thus proposed not to take into account the relative distributions Pp (nor the mirror distribution Pp ) to determine a partial relative distribution p1^ in order not to influence the result of the recombination. In other words, we avoid taking into account a distribution Pp (which could be calculated by mirror of P - operation which we note P^) nor Pj^ (which could be calculated by mirror of Pp - operation noted Pp). Indeed, there exists a form of symmetry in the matrix of relative distributions so that P is similar (but is not necessarily equal) to the mirror of Pp.

[0041] According to a particular embodiment, the steps of determining a partial relative distribution and of recombination are repeated, the method further comprising at each iteration, a step of recombination with the relative distributions of the set E determined initially.

[0042] In this way, the consolidated relative distributions are used in a subsequent iteration to further consolidate the relative distributions. This recombination then comprises a term-by-term product of the relative distributions of the set E and the corresponding relative distributions consolidated in the previous iteration, or the relative distributions updated in the current step.

[0043] According to another aspect, there is provided a device for demodulating a signal transmitted on a transmission channel on which a frame comprising N information words is received, a word Pi with i G [0; N - 1] being modulated by a particular symbol CPi from a dictionary of K symbols, the device comprising a processor coupled to a memory in which instructions are recorded adapted to implement the following steps, when they are executed by the processor: - Determination of a set E of probability distributions, called relative distributions, a relative distribution Pp being representative of the possible values ​​of the difference between symbols CPj and Cpj from a plurality of pairs (CPi; Cpj) formed from the symbols composing the frame, with 4 je [0, N - 1], as well as their respective probabilities, - For each relative distribution Pp of the set E-, - Determination of at least one partial relative distribution pP") by relating a first relative distribution determined from symbols CPi and CPm with a second relative distribution determined from a symbol Cpj and the symbol CPm, with i, j * m, and m G [O, N - 1], - Determination of a consolidated relative Pp distribution, by recom combination of determined partial relative distributions, - Demodulation of the frame from at least part of the set E of consolidated relative distributions.

[0044] The invention also relates to a radiofrequency receiver comprising a demodulation device as described above, as well as a communication terminal or a base station comprising such a receiver.

[0045] In a particular embodiment, the different steps of the demodulation method are determined by computer program instructions.

[0046] Consequently, the invention also relates to a computer program comprising instructions adapted to the implementation of the steps of a demodulation method as described above, when the program is executed by a processor.

[0047] This program may use any programming language, and be in the form of source code, object code, or code intermediate between source code and object code, such as in a partially compiled form, or in any other desirable form.

[0048] The invention also relates to a computer-readable information medium on which is recorded a computer program comprising instructions for executing the steps of a demodulation method as described above.

[0049] The information medium may be any entity or device capable of storing the program. For example, the medium may comprise a storage means, such as a ROM, for example a CD ROM or a microelectronic circuit ROM, a flash memory, or a magnetic recording means, such as a hard disk.

[0050] On the other hand, the information medium may be a transmissible medium such as an electrical or optical signal, which may be conveyed via an electrical or optical cable, by radio or by other means. The program according to the invention may in particular be downloaded from an Internet-type network.

[0051] Alternatively, the information carrier may be an integrated circuit in which the program is incorporated, the circuit being adapted to execute or to be used in the execution of the method in question.

[0052] The various embodiments or features mentioned above may be added independently or in combination with each other, to the steps of the demodulation method.

[0053] The devices, terminals, programs and information media have advantages similar to those conferred by the method to which they correspond. Brief description of the figures

[0054] Other features and advantages will become apparent upon reading a method of rea- preferred embodiment described with reference to the accompanying drawings among which: - [Fig.l] illustrates a time-frequency frame in which each of the N time steps contains a complete symbol of size K distributed over all the subcarriers in frequency, - [Fig.2] is a flowchart representing the main stages of a demodulation process according to a particular implementation, - [Fig.3] is a table showing the variables involved in the calculation of partial distributions in the case of a set of three random variables, - [Fig.4] is a table showing variables likely to intervene in the consolidation of a partial distribution, - Figure 5 illustrates an example of a calculation tree governing the execution of two iterations of the demodulation method with a view to consolidating a relative distribution ^i / o, in a particular embodiment, - Figure 6 illustrates a particular embodiment of a calculation tree capable of governing the execution of two iterations of the demodulation method with a view to consolidating a relative distribution P i / o, in which branches capable of being masked have been highlighted, - [Fig.7] is a diagram representing the architecture of a device suitable for implementing the demodulation method in a particular embodiment. Detailed description

[0055] In the following description, embodiments are described on the basis of non-limiting examples making it possible to explain the concepts on which the invention is based. In particular, although the examples and the terminology used may refer to certain well-known technologies or standards, these references are not limiting and other technologies may be adapted to implement the concepts of the invention. For example, the CCSK modulation technique referred to may be replaced by other modulation techniques, such as QPSK, or QAM (Quadrature Amplitude Modulation), without it being necessary to modify the invention.

[0056] [Fig. 2] illustrates the main steps of a demodulation method according to a particular embodiment. The method is for example implemented by a radiofrequency receiving device, such as a connected object, a terminal, user equipment, a base station, etc.

[0057] In a first step 200, the method comprises receiving a data frame transmitted by another device. In a particular embodiment, the frame is a CP-OFDM frame comprising a plurality of CCSK sequences, each of the N steps of time comprising a CCSK sequence of size K distributed over all the subcarriers in frequency, as shown in Figure 1. The received frame thus comprises N binary words of size P bits, each binary word P, of the frame being modulated by a sequence of complex values ​​CPi of size K = 2P, with 1 e [Oi N - 1],

[0058] In step 201, for each sequence CPi of the frame, a difference with each of the other sequences is estimated. The difference operator in question here depends on the modulation technique used. In the case of CCSK modulation, such a difference corresponds for example to the value of a cyclic shift of the sequence considered with respect to another sequence. In the case of QPSK modulation, the difference operator is adapted to estimate phase or angle shifts.

[0059] In a particular embodiment, the estimated difference is representative of a difference between index values ​​associated with the symbols in a symbol dictionary, according to an indexing system equal to the set {0, 1, ..., Kl] up to an isomorphism, and their respective probabilities. More particularly, the index values ​​associated with the symbols belong to a finite group of size K, in which a difference between two values ​​of the set is always also a value belonging to said index set.

[0060] Each symbol in the dictionary is thus associated with an index, typically a value from 0 to Kl when the dictionary includes K symbols. We can represent a vector P a (respectively Pg) as the probability distribution of symbol A (respectively of symbol B). For example, if K=4, we can have Pa = [0.1,0.1,0.1,0.7]. The indexing system makes it possible to determine that the corresponding symbol probably has the index symbol value 3. Thus, Paib = [0.1,0.7,0.1,0.1] means that the difference (modulo K=4) between the indices associated with A and B is probably equal to 1.

[0061] As indicated, such a difference is estimated in a finite group of size K, i.e. modulo the cardinality of the symbol dictionary when assuming an indexing in the form {0, 1, ..., Kl}.

[0062] In a particular embodiment, the index values ​​are defined by the group mZ / inKZ, considering a modular arithmetic on the remainders of the division by mK andm a strictly positive integer, Z being the set of relative integers.

[0063] For example, with m=2 and K=4, we obtain a set of indexes [0,2,4,6] such that any difference (or sum) modulo mK between elements of the set is also a value belonging to the set. Thus, in this example, we actually observe that 0 - 2 = -2 [8] = 6, 0 - 4 = -4 [8] = 4, 2 - 4 = -2 [8] = 6, etc...

[0064] We thus obtain a set of relative a priori distributions, that is to say a set noted E of probability distributions which express differences between numerical values ​​representative of the symbols, and which are estimated directly at from the symbols received.

[0065] In a particular embodiment, this processing results in a representation of the OFDM frame in the form of a matrix of probability distributions (or relative distributions in the sense of the invention) of the differences between symbols, i.e. of the relative offsets between the CCSK sequences of the frame:

[0066] PO / o ■■■ Pn-i / q' P = : : ,Po / Nl ■" Pn-I / N-1.

[0067] In this matrix, each index Pip thus contains a vector of size K (making it possible to represent all the possible symbols, which can correspond, in the case of a CCSK modulation to the K possible shifts of a sequence of size K), an element p^ of the vector containing the probability that the relative shift between the symbols CPi and Cpj of the OFDM frame corresponds to the value k,

[0068] For example, consider a frame containing 5 CCSK symbols of size 4 (each encoding 2 bits of information). The cardinality of the symbol dictionary is therefore equal to 4. If Py^ — [0.2, 0.3, 0.1, 0.4], then the most probable relative offset of symbol 3 with respect to symbol 5 is Argmax (P3 / 5) = 3. In other words, the values ​​of these two symbols are linked by the relation p3 = p5 + 3 mod 4.

[0069] As already noted, it is expected by construction that the diagonal of such a matrix contains distributions whose maximum is reached for the index 0 (the shift of a CCSK symbol with respect to itself is 0). We can also note a form of symmetry in the matrix, the shift of the symbol / with respect to the symbol j corresponding to the inverse shift of the symbol j with respect to the symbol i: = Pjl[K " k mod Æj- Thus, taking the previous example, P3 / 5=[0.2, 0.3, 0.L 0.4] and Argmax(P3 / 5) = 3, then it is expected that P5 / 3- [0.2, 0.4, 0.1, 0.3] and Argmax(P5 / 3) = 1, up to noise. In other words, the values ​​of these two symbols are linked by p5 — p^ - 3 mod 4 = p + 1 mod 4

[0070] Of course, it is possible to only determine a subset of these relative distributions without modifying the invention.

[0071] During a step 202, for each relative distribution P^j of the matrix determined in step 201, at least one relative distribution called partial pk$ is determined by relating a first relative distribution determined from symbols CPi and CPm with a second relative distribution determined from a symbol Cpj and the symbol CPm, with mi, mj, and m G [0; N - 1],

[0072]

[0073]

[0074]

[0075]

[0076]

[0077]

[0078]

[0079]

[0080] In this step, partial versions of all relative distributions are calculated using the pairs of relative distributions calculated in step 201. In a particular embodiment, these partial versions are obtained using a cross-correlation operation between the relative distributions: ^1,= This relationship between distributions is a direct consequence of the transitivity of the difference operation between the underlying random variables. In a particular embodiment, relative relationships based on sums rather than differences are considered, the cross-correlation operator is then replaced by the convolution operator without modifying the invention. Figure 3 is a table showing the variables involved in the calculation of the N3 = 27 partial distributions in the case of a set of three random variables, for example a CCSK-CP-OFDM frame of size N = 3. This table indicates the relative distributions x and from which a partial relative distribution p^ is obtained. Thus, for example, the partial relative distribution p($ is produced by cross-correlation between the distributions Pyi and 0 / i. According to a particular embodiment, the partial relative distributions pL™) determined from a relative distribution P^ and / or a relative distribution P^ are not calculated. The method comprises a step 203 during which a relative distribution Pa / b is calculated from the partial relative distributions p^ determined in step 202. For this, it is proposed to calculate a product of the partial relative distributions pC) calculated: £ ajb According to a particular embodiment, the method comprises a step 204 during which the distribution Pa / b obtained by the product of the partial relative distributions p^^ is recombined with the corresponding a priori distribution, that is to say with the relative distribution Paib calculated in step 201. This recombination allows us to obtain consolidated relative distributions, also called a posteriori relative distributions. These consolidated relative distributions allow us to update a set E' of relative distributions, each distribution of which is then consolidated. The table in Figure 4 shows variables that may be involved in the consolidation of one of the N2 = 9 partial distributions. For example, the consolidation of the relative distribution P2 / 1 can be based on the product of the partial distributions p^ and p^-

[0081] According to a particular embodiment, steps 202 to 204 are executed iteratively. At each iteration, the a priori relative distributions are combined, according to steps 202 to 204, with the consolidated relative distributions of the set E' received from the previous iteration to obtain new estimates of the relative distributions, herein referred to as a posteriori. These new estimates are then used as a basis for calculating the new partial relative distributions of the current iteration, themselves used to determine the information to be provided at the following iteration by updating the set E'.

[0082] In other words, the relative distributions of an iteration R are obtained by relating relative distributions E' determined during an iteration R -1 with a priori relative distributions of the set E determined initially. During the first iteration, the matrix E' is initialized from distributions determined for the set E.

[0083] Figure 5 illustrates an example of a trellis (or computational graph) governing the execution of two iterations of the method for the consolidation of a relative distribution P i / o, always for a set of 3 random variables, such as a CCSK-CP-OFDM frame of size 3 (an equivalent graph, not shown here, would be executed in parallel for the consolidation of the other relative distributions). The bold arrows represent the a priori information, i.e. the information available about a variable at the initialization of the method. The a posteriori value of a relative distribution can therefore be defined as the product of the information from the previous iteration by the a priori information.

[0084] If the result of this calculation is likely to be used again due to the execution of an additional iteration of the method, it is important to guard against certain “self-influence loops”. For this, according to a particular embodiment, a masking step is implemented to avoid such influence loops. The masking may consist of not taking into account certain partial relative distributions at the step in the recombination step 203, or simply of not calculating these distributions during step 202.

[0085] In Figure 4, the combination distributions likely to create a self-influence loop have been highlighted by a speckled or striped background pattern. For example, the partial relative distributions pW and ptV are likely to create a self-influence loop when they are used to obtain a relative distribution P2 / 0 because such a recombination involves the intrinsic information of this variable, that is to say, from the observation of the variable that one wishes to update, or from the observation of the symmetrical variable.

[0086] The cells whose pattern is speckled in Figure 4 are distributions partial relative distributions constructed on the basis of intrinsic information, i.e. directly derived from the observation of the variable that we wish to modify. The cells whose grid is striped are partial distributions constructed on the basis of the symmetrical relative distribution. Indeed, the matrix P determined in step 201 being "symmetrical", the symmetrical value can be considered as the intrinsic value with respect to subsequent updates, and it may be desirable not to take it into account either.

[0087] [Fig.6] shows the calculation graph of [Fig.5] on which the distributions likely to produce a self-influence loop have been highlighted by dotted arrows. The masking operation can thus consist of not taking into account the calculation branches comprising dotted arrows in the recombination step 203. Only the extrinsic information is then retained, i.e. the information that can be drawn about a variable by simply observing the other variables linked to it, and the use of intrinsic information is avoided, i.e. the information that can be drawn about a variable by simply observing the latter.

[0088] Thus, in a particular embodiment, it is proposed not to take into account the partial relative distributions constructed on the basis of the intrinsic information: for example, the consolidated value of Po / i must not depend on the partial relative distribution p( L — p @ p which contains the intrinsic value P o / }. 1 \jf -*• 1 / *

[0089] In a particular embodiment, it is proposed not to take into account the partial distributions constructed on the basis of the symmetrical relative distribution: for example, the consolidated value of Fq / i must not depend on the partial relative distribution p($) _ p^ @ p^ which contains the symmetrical value P i / o. Indeed, the z / y[^] ~ “ symmetric can be considered as the intrinsic value, up to noise, with respect to subsequent updates.

[0090] According to a particular embodiment, the method comprises at least one step of normalizing the probabilities H may be necessary to normalize the distributions obtained in the different steps of the method so that their sums are equal to 1 in accordance with their probabilistic natures.

[0091] Furthermore, although the method is described here with reference to discrete laws, the present proposal can be generalized to continuous laws, for example by approximating the continuous law by a discretized law, or by modeling the distributions by their parameters (for example the mean and the variance as parameters of a normal law). It is then appropriate to express the relationship between the parameters of the relative distributions and the partial relative distributions (for the matrix P considered to be “symmetrical” (P k mod ^1) - the value calculation of partial relative distributions), as well as the impact of the product reduction of the partial distributions on the parameters of the relative distribution thus obtained (for the calculation of the product reduction).

[0092] The method finally comprises a demodulation step 205 during which at least part of the relative a posteriori distributions of the set E' are used. For example, the demodulation can be based on the offsets between CCSK sequences relative to a sequence of the frame taken as reference (in other words, using a particular row of the matrix updated after one or more iterations of the method). Knowledge of the relative offsets between sequences of the same frame allows offset demodulation, even when the offset of a CCSK sequence relative to the root sequence is unknown.

[0093] When the relative offsets between each of the sequences transmitted in the frame can be determined, for example when all the symbols in the frame are subject to the same channel, it is possible to perform an offset demodulation using the integrity check data of the frame to determine the offset to be applied to find the transmitted binary words. To do this, the received CCSK sequences can be demodulated by considering a zero offset of a sequence in the frame taken as a reference. The relative offsets being known, the binary words of the frame are obtained. The integrity check, for example a CRC, is then used to determine whether the decoded frame is correct. If this is not the case, i.e. if the integrity check fails, a new demodulation attempt is made after applying an offset to the CCSK sequences of the frame. This is done until the integrity check shows correct demodulation.

[0094] [Fig.7] represents a simplified architecture of a device 700 adapted to implement the demodulation method according to a particular embodiment.

[0095] The device 700 comprises a data processing module comprising a storage space 701, for example a memory (MEM), a processing unit 702, equipped for example with a microprocessor (PROC), and controlled by a computer program (PGR) 703 whose instructions are configured to implement a particular embodiment of the demodulation method described in relation to FIG. 2, and in particular the steps of determining a set E of probability distributions, called relative distributions, a relative distribution P^j being representative of the possible values ​​of the difference between the index values ​​associated with symbols CPi and Cpj from a plurality of pairs (CPi; CPj) formed from the symbols making up the frame, with 4 je [QN - 1] as well as their respective probabilities, said difference being calculated in a finite group of size K;for each relative distribution P^ of the set E, determination of at least one distribution; partial relative by relating a first relative distribution determined from symbols CPi and CPm with a second relative distribution determined from a symbol Cpj and the symbol CPm, with i, j ï m, and m F [0; N - IJ and determining a consolidated relative distribution P^j, by recombination of the determined partial relative distributions, and demodulating the frame from at least part of the consolidated relative distributions.

[0096] At initialization, the code instructions of the computer program 703 are for example loaded into the memory 701 before being executed by the processor of the processing unit 702. The microprocessor of the processing unit 702 implements, according to the instructions of the computer program 703, the steps of the demodulation method described above with reference to [Fig.2].

[0097] For this, in addition to the memory 701 and the processor 702, the device comprises communication means 704, for example an OFDM transducer adapted to receive signals on a plurality of orthogonal carriers. In a particular embodiment, the communication means 704 are configured by computer program instructions to allow the reception of at least one time-frequency frame in which each of the N time steps contains a complete CCSK sequence of size K distributed over all the frequency sub-carriers.

[0098] The device 700 also comprises an OFDM demodulator 705 adapted to generate N vectors of K values ​​from the frame received by the communication means 704 on K subcarriers, the K values ​​corresponding to the K complex values ​​composing a CCSK sequence of K elements coding for an information word of the frame.

[0099] The device 700 also comprises a module 706 for determining a matrix comprising a set E of probability distributions, called relative distributions, each element Pip of the matrix being a vector of K probabilities representative of the possible values ​​of the difference between the index values ​​associated with the CPi and CPj of a plurality of pairs of sequences (CPi; CPj) formed from the sequences composing a frame received by the communication means 704, with i G [0; N - 1] and je [ 0; N - 1], as well as the respective probabilities of these differences, a difference being calculated in a finite group of size K, or modulo the cardinality of the symbol dictionary by considering the index set {0,1,.. .,K-1}.The module 706 is for example implemented by computer program instructions configured to calculate a difference between two sequences of the frame by a cross-correlation and / or convolution operation, determine a probability that this difference corresponds to a particular offset, and store the . result of the difference between the elements of a pair (CPi; Cpj) of sequences at coordinates (i', j) in the matrix.

[0100] The device further comprises a module 707 for calculating partial relative distributions. The module is implemented by computer program instructions configured to relate a first relative distribution determined from symbols CPi and CPm with a second relative distribution determined from a symbol CPj and the symbol CPm, with i,j tn, and m G [0; N - 1], The relating is carried out by a cross-correlation and / or convolution operation. The module 707 carries out the relating from the data of the matrix determined by the module 706. In other words, the module 707 extracts the different pairs of relative distributions from the matrix to calculate partial relative distributions by a cross-correlation and / or convolution operation.

[0101] The device comprises a recombination module 708 adapted to consolidate relative distributions P^j of the matrix from partial relative distributions p^ determined by the module 707. For this, the module 708 comprises program instructions configured to produce a product of partial relative distributions, m G [ 01N - 1 ], for each distribution P^j.

[0102] In a particular embodiment, the module 708 is further configured to apply a filtering of the partial relative distributions before the product reduction, in particular with a view to a new iteration of the method. Such a filtering step is for example implemented by program instructions configured not to take into account the partial relative distributions constructed on the basis of the intrinsic information, that is to say on the basis of the relative distribution for which the partial relative distribution is calculated, nor of its mirror relative distributions. For example, the module is configured so that the distribution ^o / i is not consolidated from a partial distribution of p^) _ p^ @ p^, because such a partial distribution depends on Fo / i, and therefore on the intrinsic value ^o / i.

[0103] The device 700 comprises a module 709 adapted to multiply term-by-term the distributions determined by the module 708 with the relative a priori distributions, determined by the module 706.

[0104] The device 700 finally comprises a demodulation module 710 adapted to carry out a demodulation of the symbols transmitted in the frame from at least a part of the relative distributions consolidated by the module 709. For example, the module 710 is implemented by program instructions adapted to obtain the offset of a predetermined sequence of the frame, for example of a pilot sequence whose position and offset are known by convention, and to demodulate the other sequences of the frame using the consolidated relative distributions, and in particular based on the consolidated relative offsets between the known sequence and the other sequences of the frame, for example from the consolidated offsets in one or more particular rows of the matrix updated by the modules 707 to 709.

[0105] In a particular embodiment, the device is integrated into a communication terminal, a connected object, a vehicle, a gateway, an access point or even a base station.

Claims

Claims

1. Method for demodulating a signal transmitted on a transmission channel on which a frame comprising N information words is received, a word Pi with i G [0; N - 1] being modulated by a particular symbol CPi from a dictionary of K symbols, the method comprising the following steps: - Determination of a set E of probability distributions, called relative distributions, a relative distribution Pip being representative of the possible values ​​of the difference between symbols CPi and Cpj from a plurality of pairs (CPi; Cp^ formed from the symbols composing the frame, with i, je [(>, N - 1] as well as their respective probabilities, - For each relative distribution P^j of the set E;- Determination of at least one partial relative distribution by relating a first relative distribution determined from symbols CPi and CPm with a second relative distribution determined from a symbol CPj and the symbol CPm, with i, j * m, and me [0; N -11 - Determination of a consolidated relative distribution P(yj), by recombination of the determined partial relative distributions, - Demodulation of the frame from at least one part of the consolidated relative distributions.;

2. A method according to claim 1 wherein a CPi symbol is a sequence of complex values ​​obtained by applying a particular cyclic shift to a root sequence of size K, a CPi sequence being transmitted on K subcarriers of an OFDM symbol.

3. A method according to any preceding claim wherein a relative distribution P^j is a vector of size K in which each element is a probability that a difference between the symbols CPi and CPj of the frame corresponds to the value k.

4. Method according to any one of the preceding claims such that it comprises at least one step of normalizing the pro- skills *

5. Method according to any one of the preceding claims in which the step of determining a partial relative distribution p^ comprises at least one cross-correlation operation between a relative distribution Pum and a relative distribution P jim, or between a relative distribution Pmfi and a relative distribution Pm]j with i, jm and me [0; N -1],

6. Method according to any one of the preceding claims in which the step of determining a partial relative distribution comprises at least one convolution operation between a relative distribution Pi / m and a relative distribution Pm / j, with i, j ± m and me [0, N -1],

7. Method according to any one of the preceding claims in which the recombination step does not take into account the partial relative distributions p^ determined from a relative distribution P^ and / or determined from a relative distribution Pj / j.

8. Method according to any one of the preceding claims in which the steps of determining a partial relative distribution and of recombination are repeated, the method further comprising at each iteration, a step of recombination with the relative distributions of the set E initially determined.

9. Device for demodulating a signal transmitted on a transmission channel on which a frame comprising N information words is received, a word P, with i E [0; N - 1] being modulated by a particular symbol CPj from a dictionary of K symbols, the device comprising a processor coupled to a memory in which instructions are recorded adapted to implement the following steps, when they are executed by the processor: - Determination of a set £ of probability distributions, called relative distributions, a relative distribution Pijj being representative of the possible values ​​of the difference between symbols CPi and Cpj from a plurality of pairs (Cpj ; Cpj) formed from the symbols composing the frame, with i, je [0, N - 1], as well as their respective probabilities, - For each relative distribution Pijj of the set E; - Determination of at least one partial relative distribution by relating a first relative distribution determined from symbols CPi and CPm with a second relative distribution determined from a symbol Cp, and from the symbol CPm, with i, j * m, and me [0; N - IJ - Determination of a consolidated relative distribution P-^j, by recombination of the determined partial relative distributions, - Demodulation of the frame from at least one part of the set E of consolidated relative distributions.

10. A communication terminal comprising a demodulation device according to claim 9.

11. Computer program comprising instructions adapted to the implementation of the steps of a demodulation method according to any one of claims 1 to 8, when the program is executed by a processor.

12. A computer-readable information medium on which is recorded a computer program comprising instructions for executing the steps of a demodulation method according to any one of claims 1 to 8.

Images