Reception of a signal modulated according to a multilevel coding technique

0001The field of the invention is that of signal processing and digital communications. '

0002More specifically, the invention relates to a technique for optimizing the decoding of a signal modulated according to a multi-level coding technique, or MLC (for English "Multi-Level Coding").

0003To date, several coded modulation channel coding techniques are known. Indeed, following Ungerboeck's discovery of lattice coded modulations, also called MCT ("<nplcit id="ncit0001" npl-type="s"><text>Channel Coding with Multilevel / phase Signals "," Channel Encoding with Multilevel / Phase Signals ", IEEE Trans., January 1982, 28, No. 1, pp. 55-67</text></nplcit>), block coded modulations, or MCBs, and multi-dimensional trellis coded modulations have been proposed.

0004MCB modulations have been described by Cusack in "<nplcit id="ncit0002" npl-type="s"><text>Error control codes for QAM signaling ", Electronics Letter, January 1984, 20, pp. 62-63</text></nplcit> and by <nplcit id="ncit0003" npl-type="s"><text>Sayegh in "A class of optimum block codes in signal space" (in French "an optimal block code class in the signal space"), IEEE Trans. COM, October 1986, 34, No. 10, pp. 1043-1045</text></nplcit>.

0005The multidimensional modulations encoded in lattice have in particular been described by <nplcit id="ncit0004" npl-type="s"><text>Wei in "Trellis-coded Modulation with Multidimensional Constellations" (in French "Trellis coded modulation with multidimensional constellations"), IEEE Trans. IT, July 1987, 33, No. 4, pp. 483-501</text></nplcit> and by <nplcit id="ncit0005" npl-type="s"><text>Calderbank and Sloane in "New lattice codes based on lattices and cosets" (in French "New Trellis Codes Based on Trellises and Conjugate Groups"), IEEE Trans. IT, March 1987, 33, No. 2, pp. 177-195</text></nplcit>.

0006Trellis coded modulations of moderate complexity (typically 4 or 8 states) can give a coding gain of 3 to 4 dB. However, in high-capacity radio-relay applications, the implementation of the Viterbi decoder necessary to decode these modulations remains very expensive.

0007A new family of MCB codes has therefore been proposed for these particular applications. The implementation of such codes is simple, but their coding gain is generally limited to 2dB.

0008An attractive coding technique for these high-capacity radio-relay applications has been proposed by <nplcit id="ncit0006" npl-type="s"><text>Imai and Hirakawa in "A new multilevel coding method using error-correction codes" (in French "a new multi-level coding method, using error correcting codes"), IEEE Trans. IT, May 1977, 23, No. 3, pp. 371-377</text></nplcit>. This technique is that of multi-level coding, whose interest lies in the existence of a simple method of decoding, being carried out in stages, and having a good compromise between performance and complexity of implementation.

0009The principle of multi-level coding, as well as the associated step decoding method, is briefly described below.

0010We consider a constellation <i>AT₀</i> at 2<i>^m</i> points, who is therefore able to ensure the transmission of <i>m</i> bits per symbol.

0011Yes <i>m</i> is the number of bits to encode, the constellation <i>AT₀</i> is partitioned in <i>m</i> levels, thus giving <i>2^m</i> subassemblies. The principle of this partition is identical to that defined by Ungerboeck, and serves to maximize the minimal Euclidean distance in the subsets of the partition. If we designate<i>d_i</i> the minimal Euclidean distance in the subsets obtained at <i>i^th</i> partition level, we must check the following inequality: <maths id="math0001" num="(1)"><math display="block"><msub><mi mathvariant="italic">d</mi><mn mathvariant="italic">0</mn></msub><mo mathvariant="italic">&lt;</mo><msub><mi mathvariant="italic">d</mi><mn mathvariant="italic">1</mn></msub><mo mathvariant="italic">&lt;</mo><msub><mi mathvariant="italic">d</mi><mn mathvariant="italic">2</mn></msub><mo mathvariant="italic">&lt;</mo><mo mathvariant="italic">…</mo><mo mathvariant="italic">&lt;</mo><msub><mi mathvariant="italic">d</mi><mi mathvariant="italic">m</mi></msub></math><img file="EP1547289B1_D0001.tif" /></maths>or <i>d₀</i> is the minimum distance in the constellation <i>AT₀</i>.

0012Thus, <i>m</i> bit <i>b₁</i>, <i>b₂</i>, ..., <i>b_m</i>, or <i>b_i</i> is the bit assigned to the <i>i^th</i> partition level, select a subset among the <i>2^m</i>. The<figref idref="f0001">figure 1</figref> gives the schema of this partition in case <i>m</i> = 2. <i>AT₀</i> is partitioned firstly into two subsets <i>B_i , i</i> ∈ <i>{0, 1}</i> and or <i>i</i> = <i>b₁</i>, minimal distance <i>d₁</i>, then into four sub-sets <i>C_i</i>, <i>i</i> ∈ <i>{0, 1,</i> 2, <i>3}</i> and or <i>i</i> = <i>b₁</i> + <i>2b₂</i>, minimal distance <i>d₂</i>. In the case where<i>AT₀</i> is a square constellation of Euclidean distance <i>d₀</i>, <i>d</i>₁ = √2<i>d</i>₀ and <i>d</i>₂ = √2<i>d</i>₁ = 2<i>d</i>₀.

0013This method of assigning the points of the constellation <i>AT₀</i> is intended to classify <i>m</i> bits represented by the emitted symbol, according to their vulnerability to noise. We can indeed see that the bit<i>b₂</i> is less vulnerable than the bit <i>b₁</i>, since it corresponds to a minimum Euclidean distance from <i>d₂</i> > <i>d₁</i>. From relation (1), we can show that if the bits<i>b_k</i>, <i>k</i> ≤ <i>i-1</i> are sufficiently protected to be received correctly, the bit <i>b_i, i</i> ≤ <i>m</i> is better protected against noise than all other bits <i>b_j, j</i> < <i>i</i>. It has therefore been envisaged to code these bits separately with different codes.

0014This is the principle of multi-level coding which consists, after partitioning the constellation <i>AT₀</i> in <i>m</i> levels to use <i>m</i> encoders <i>E_i</i>, <i>i</i> = <i>l</i>, ..., <i>m</i>, to protect these <i>m</i> bits with multiple levels of protection.

0015In other words, the principle of multi-level coding is based on the joint optimization of coding and modulation, to achieve the best transmission performance. Thus, in the context of QAM ("Quadrature_Amplitude Modulation" - for "quadrature amplitude modulation"), greater protection is given to bits which, because of their position in the QAM mapping, are more likely to to be tainted with error. The protection granted to the different bits depends on the coding used.

0016The schema of this coding concept is illustrated by the <figref idref="f0002">figure 2</figref>. The data stream to be transmitted, debit<i>D</i>, is divided by the series-parallel conversion block 21 into <i>m</i> flow trains <i>D_i, i</i> = <i>1</i>, ..., <i>m</i>. The<i>m</i> first trains are coded by <i>m</i> binary codes <i>E_i</i> (<i>not_i</i>, <i>k_i, d_i), i =</i> 1, ..., <i>m</i>, referenced 22, coding rate <i>R_i</i> =-<i>k_i</i> / <i>not_i</i> and minimum Hamming distance <i>d_i</i>. At the input of the modulator 23, the<i>m</i> bitstreams must be synchronous, debit <i>D</i>'/<i>m</i>. We can therefore define an equivalent coding rate<i>R</i> given by : <maths id="math0002"><math display="block"><mi mathvariant="italic">R</mi><mo mathvariant="italic">=</mo><mi mathvariant="italic">D</mi><mo mathvariant="italic">/</mo><mi mathvariant="italic">OF</mi></math><img file="EP1547289B1_D0002.tif" /></maths>

0017If we assume that all n_i are equal, either <i>not_i</i> = <i>n, i = 1</i>, ...,<i>m</i> , and that <i>m</i> codes <i>E_i</i> are codes in block, one can describe this coding by a matrix structure identical to that used for the MCB described in particular by Sayegh in the article cited previously. A code word contains<i>not</i> symbols and can be represented by a matrix to <i>m</i> lines and <i>not</i> columns where the <i>j^th</i> column represents the binary assignment of the <i>j^th</i> block symbol, and the <i>i^th</i> line represents the <i>i^th</i> partition level. Line<i>i, i</i> = <i>1</i>, ..., <i>m</i> is a code word <i>E_i (not_i, k_i</i>, <i>d_i)</i>The minimum Euclidean distance obtained with this coding is given by: <maths id="math0003" num="(3)"><math display="block"><msup><mi>d</mi><mn>2</mn></msup><mo>=</mo><msub><mi>min</mi><mrow><mi>i</mi><mo>=</mo><mn mathvariant="italic">1</mn><mo mathvariant="italic">,</mo><mo mathvariant="italic">…</mo><mo mathvariant="italic">,</mo><mi mathvariant="italic">m</mi><mo mathvariant="italic">+</mo><mn mathvariant="italic">1</mn></mrow></msub><mo>⁢</mo><mfenced><msub><mi>d</mi><mi>i</mi></msub><mo>⁢</mo><msup><msub><mi>d</mi><mrow><mi>i</mi><mo>-</mo><mn>1</mn></mrow></msub><mn>2</mn></msup></mfenced><mo>,</mo><mspace width="3em" /><mi>with</mi><mspace width="1em" /><msub><mi mathvariant="italic">d</mi><mrow><mi mathvariant="italic">m</mi><mo mathvariant="italic">+</mo><mn mathvariant="italic">1</mn></mrow></msub><mo mathvariant="italic">=</mo><mn mathvariant="italic">1.</mn></math><img file="EP1547289B1_D0003.tif" /></maths>

0018Knowing that <i>d_i</i> check relationship (1) above, multi-level coding is optimized if: <maths id="math0004" num="(4)"><math display="block"><msub><mi mathvariant="italic">d</mi><mn mathvariant="italic">1</mn></msub><mo>></mo><msub><mi mathvariant="italic">d</mi><mn mathvariant="italic">2</mn></msub><mo>></mo><mo>…</mo><mo>></mo><msub><mi mathvariant="italic">d</mi><mi mathvariant="italic">m</mi></msub><mn>.</mn></math><img file="EP1547289B1_D0004.tif" /></maths>So we determined that the bit <i>b₁</i> should be the most protected bit and then <i>b₂</i> etc. This matrix description can be generalized in case the codes are arbitrary. If the<i>not_i</i> are not identical, just consider a matrix to <i>m</i> lines and <i>l</i> columns where <i>l</i> is the smallest common multiple of <i>not_i</i>, <i>i = l</i>, ..., <i>m</i>. In the particular case where one of the codes is a convolutional code, the matrix to be considered is semi-infinite.

0019The decoding method conventionally used in association with such multi-level coding is suboptimal step decoding, which has the advantage of being very simple to implement.

0020According to this technique, the decoding method is carried out in stages where each bit is decoded independently by a simple decoder operating on firm decisions, but where the decoder output <i>(I)</i> can make a correction on the bits at the input of the decoder <i>(i</i>+<i>1)</i>. The<figref idref="f0002">figure 3</figref> gives the block diagram of this type of decoder, in the case where <i>m</i> = 2. Given <i>(r₁</i>,<i>r₂</i>,...,<i>r_not)</i> block 31 of <i>not</i> symbols received at the input of the decoder, the decoding operation is performed according to the following successive steps:<ul id="ul0001" list-style="dash" compact="compact"><li>First, the n bits are decoded <i>bⁱ_l, i</i>=<i>1</i>,...,<i>not</i> assigned to the first level of partition <i>(AT₀)</i> : a firm decision 32 in <i>AT₀</i> is performed on all <i>r_i</i>, <i>i</i>=<i>1</i>,...,<i>not</i>. This gives a first estimate of<i>bⁱ₁, i</i>=<i>1</i>,...,<i>not</i> , rated <i>b</i>₁<i>ⁱ</i>, <i>i</i>=<i>l</i>,...,<i>not</i>. Decoding with firm decisions 33 operating on<i>b</i>₁<i>ⁱ</i>, <i>i</i>=<i>1</i>,...,<i>not</i> provides a final estimate noted <i>b</i>₁<i>ⁱ</i> , <i>i</i>=<i>1</i>,...,<i>not</i>.</li><li>or decodes the n bits <i>bⁱ</i>₂, <i>i</i>=<i>1</i>,...,<i>not</i>. assigned to the second partition level (<i>B₀</i> or <i>B_l</i>): depending on the bits <i>b</i>₁<i>ⁱ</i>, <i>i</i>=<i>1</i>,...,<i>not</i>, which are encoded by the same encoder used on transmission, a second decision operation 34 is performed on the symbols <i>r_i i</i>=<i>l</i>,...,<i>not</i> in the subassemblies <i>B_pi</i> with <i>p_l= b</i>₁<i>ⁱ</i> for <i>i</i>=<i>1</i>,....<i>not</i>. Bits<i>b</i>₁<i>ⁱ</i>, <i>i</i>=<i>l</i>,...,<i>not</i> obtained are decoded by the decoder "2" referenced 35 to give a final decision <i>b</i>₁<i>ⁱ</i>, <i>i</i>=<i>1</i>,....<i>not</i>.</li><li>Finally, the remaining uncoded bits are decoded: from the bits <i>b</i>₁<i>ⁱ</i>,<i>b</i>₂<i>ⁱ i</i>=<i>1</i>,...,<i>not</i>, recoded by their associated coder, a third detection 36 is performed. <i>r_b i</i>=<i>1</i>,...,<i>not</i>, in subsets of second level of partition <i>C_i</i>, <i>i</i>=<i>1</i>,...,<i>not</i>. This gives an estimate of<i>m</i>-2 remaining uncoded bits for each of the symbols <i>r_b i</i>=<i>1</i>,...,<i>not</i>.</li></ul>

0021According to the decoding technique associated with the MLC multi-level coding, the first decoding is therefore systematically performed in the subset A₀ of the constellation. The result of this decoding is then used for decoding the next subset B₀.

0022Such a decoding technique is described in particular in the article by L, Papke and K. Fazal entitled "Different Iterative Decoding Algorithms for Combined Concentrated Coding and Multiresolution Modulation," in the field of terrestrial broadcasting of television signals, encoded according to a multi-level coding technique, More precisely, this PAPRE document describes a solution from the multi-level coding and decoding to obtain three streams linked to three different services, the SDVT stream being more robust than the EDTV stream, itself more robust than the HDTV stream. This PAPKE technique aims at protecting the most important flows, by accentuating the robustness of the level with which they are associated. In practice, according to the PAPKE decoding technique, it was first estimated that the bits_i¹ assigned to the constellation at 2^m points and then the bits u_i² assigned to the subsets of the constellation corresponding to u₁¹ etc.

0023However, for such an MLC decoding to be optimal, the coding gain that is necessary to obtain between the different coding levels is 6 dB, which is very difficult to obtain.

0024A disadvantage of this technique of the prior art is that the decoding method conventionally implemented in the context of coding <i>MLC</i> shows poor performance.

0025In particular, such a stepwise suboptimal decoding technique is poorly suited to channels with Gaussian additive noise and Doppler-affected multipath channels.

0026The invention particularly aims to overcome these disadvantages of the prior art.

0027More specifically, an object of the invention is to provide a technique for decoding a modulated signal according to an MLC coding technique, which has improved performances compared with the techniques of the prior art.

0028Another objective of the invention is to implement such a technique, which makes it possible to reduce the bit error rate (or BER) compared to the sub-optimal decoding technique of the previous year.

0029Another object of the invention is to provide such a technique which is simple and inexpensive to implement, and which is adapted to the channels subject to disturbances, and in particular to the channels having Gaussian additive noise and to the multipath channels assigned to them. of Doppler.

0030These objectives, as well as others which will appear later, are achieved by a method of receiving a modulated signal according to claim 1.

0031Thus, the invention is based on a completely new and inventive approach to decoding a modulated signal according to a multi-level coding technique. Indeed, unlike the sub-optimal decoding method used in the prior art, the invention proposes to perform a decoding of different partition levels that takes into account the vulnerability of the latter vis-à-vis the noise. Thus, the most robust level is decoded first, so that the result of decoding this level can be propagated to less robust levels. This results in significantly increased decoding performance compared to suboptimal decoding techniques of the prior art.

0032Advantageously, said predetermined order corresponds to the descending order of the robustness of the coding levels to which said received bits are assigned.

0033Preferably, each of said successive decoding steps takes into account the result of said previous decoding step (s), so as to improve the result of said decoding steps of said bits allocated to the less robust levels.

0034Thus, the result of the decoding of the bits of a given level of robustness is systematically exploited during the decoding of the bits of the directly lower level of robustness, which makes it possible to greatly improve the confidence which one can grant to this second decoding.

0035According to an advantageous variant of the invention, said bits allocated to said most robust level are the most significant bits of said corresponding symbol.

0036This embodiment variant corresponds in particular to the particular mode of implementation selected by the DRM standardization consortium (Digital Radio Mondiale, as presented in document ETSI ES 201 980 V1.2.1 (2002-07)).

0037Preferably, within one of said decoding iterations, each of said successive decoding steps of said received bits is preceded by a corresponding demodulation step.

0038The received bits are first demodulated and then decoded.

0039Advantageously, such a reception method comprises at least two successive decoding iterations, a step of decoding the bits of a given level taking into account, at the time of the^th iteration, where n≥2, of the result of at least some of said decoding steps of said received bits assigned to coding levels less robust than said given level, and implemented in at least one of said previous iterations.

0040Thus, in a particular case comprising three coding levels, the decoding of the bits of the most robust level takes into account, in particular during the second iteration, the result of the decoding of the bits of the two least robust levels obtained during the first iteration.

0041Preferably, such a reception method comprises two successive decoding iterations.

0042Indeed, the inventors have found that the increase in performance resulting from the implementation of a third iteration was low, or at least negligible compared to the increase in the corresponding complexity.

0043Advantageously, at the end of at least some of said iterations, such a reception method implements a step of estimating an emitted symbol, and a step of calculating an extrinsic information taking into account said estimated transmitted symbol, said extrinsic information making it possible to improve the result of said decoding steps of said one or more iterations.

0044Thus, after the first iteration of decoding, extrinsic information is calculated which will be used during the second decoding iteration, to increase its performance.

0045Advantageously, said extrinsic information is of the form: α (S_r-S_e), where α ∈ [0, 1], S_r is said received symbol and S_e is said estimated transmitted symbol.

0046In the particular case where two successive iterations are implemented, the extrinsic information is therefore proportional to the difference of the received symbol and the estimated symbol from the decoded bits of the different levels during the first iteration. This difference is weighted by a characteristic coefficient of the confidence granted to the decoding.

0047In a first advantageous embodiment of the invention, α is substantially equal to 0.25.

0048Such a value of α makes it possible to obtain satisfactory performances during the second iteration of decoding, for most of the transmission channels considered.

0049In a second advantageous embodiment, such a reception method comprises a step of optimizing the value of α as a function of the signal-to-noise ratio.

0050By the intermediary of the coefficient α, one can indeed choose to grant a greater or lesser confidence in the extrinsic information to the estimated transmitted symbol, in order to take more or less account for the following decoding iterations. The optimization of the value of α as a function of the signal-to-noise ratio leads to values ​​of α close to 1 when the signal-to-noise ratio is very good, and to values ​​close to 0 in the opposite case.

0051According to an advantageous characteristic of the invention, such a reception method further comprises a step of determining a signal-to-noise ratio from at least one transmitted reference information, called a pilot, the value of which is known a priori. in reception.

0052It is recalled that in order to estimate the transmission channel in OFDM for example, a conventional technique consists in inserting reference carriers into the carrier stream, at known locations of the receiver. In reception, the values ​​taken by these reference carriers, called pilots, are read, and it is easy to deduce the complex gain of the channel at these reference locations. The complex gain of the channel is then derived over all the points of the transmitted time-frequency network, from the calculated value of the complex gain at the reference locations.

0053Such a mechanism based on pilots can therefore be used, in the context of the invention, to determine the signal-to-noise ratio, and thus to optimize α. It is used in particular by the DVB-T standard ("Digital Video Broadcasting (DVB), Framing Structure, Channel Coding and Modulation for Terrestrial Television (DVB-T)", "Digital Video Broadcasting (DVB), frame structure, coding of channel and modulation for digital terrestrial television (DVB-T), ETS 300 744, March 1997).

0054According to an advantageous embodiment of the invention, such a reception method further comprises, for at least some of said coding levels, an additional de-interleaving step implemented between said demodulation and decoding steps of said received bits.

0055Such an embodiment makes it possible in particular to improve the performance of the reception method vis-à-vis the Doppler affected transmission channels.

0056The invention also relates to a method for decoding a modulated signal according to claim 16.

0057The invention also relates to a device for receiving a modulated signal according to claim 17.

0058The invention also relates to a system for coding / decoding a signal according to claim 18.

0059The invention finally relates to the applications of the reception method described above in at least one of the following areas:<ul id="ul0002" list-style="dash" compact="compact"><li>digital radio transmissions, in particular of DRM ("World Digital Radio") type;</li><li>error correcting codes;</li><li>digital signal processing;</li><li>digital communications;</li><li>the recording / playback of a digital signal.</li></ul>

0060Other characteristics and advantages of the invention will appear more clearly on reading the following description of a preferred embodiment, given as a simple illustrative and nonlimiting example, and the appended drawings, among which:<ul id="ul0003" list-style="dash" compact="compact"><li>the <figref idref="f0001">figure 1</figref>, already described in relation with the prior art, presents an example of partition of a constellation <i>AT₀</i> in <i>m</i> levels, giving <i>2^m</i> subsets, when <i>m</i>=<i>2</i> ;</li><li>the <figref idref="f0002">figure 2</figref>, already described in relation with the prior art, presents a block diagram of a multi-level coder;</li><li>the <figref idref="f0002">figure 3</figref>, also described in relation with the prior art, presents a block diagram of the step decoder implemented in the prior art, in association with the multi-level coder of the <figref idref="f0002">figure 2</figref>, in the case of a three-level coding;</li><li>the <figref idref="f0003">figure 4</figref> presents a comparison of the robustness of different levels of MLC coding according to a Gaussian additive white noise;</li><li>the <figref idref="f0004">figure 5</figref> presents an example of a receiver according to the invention, performing the optimized decoding of a symbol MAQ64 with two iterations and use of the extrinsic information;</li><li>the <figref idref="f0003">figure 6</figref> illustrates the comparative decoding performance of the MLCs according to the sub-optimal decoding method of the state of the art and according to the decoding method of the invention.</li></ul>

0061The general principle of the invention is based on taking into account the noise robustness of the different coding levels of a signal modulated according to a multi-level coding technique MLC, to determine the decoding order of the bits received.

0062We present, in relation to the <figref idref="f0003">figure 4</figref>, the concept of robustness of a coding level, as part of a multi-level MLC coding technique.

0063The robustness of a coding level can be illustrated by the curve of the bit error rate of this level, as a function of the signal-to-noise ratio (S / N): throughout the document, it will be considered that a coding level is all the more robust as the bit error rate associated with it is low.

0064By decoding each level of coding independently, that is, without performing any loopback from one level to the other (in other words, the result of decoding a level is not used during the decoding the next level), it is possible to determine the level of robustness of each level vis-à-vis the noise. In particular, the<figref idref="f0003">figure 4</figref> illustrates the robustness of each MLC coding level with Gaussian additive white noise.

0065Thus, the DRM standardization consortium (Digital Radio Mondiale, as presented in ETSI ES 201 980 V1.2.1 (2002-07)) has retained the multi-level MLC coding for broadcasting a digital signal in the bands. AM (in English "Amplitude Modulated" for "amplitude modulated"), whose frequencies are lower than 30MHz. One of the modes retained by DRM comprises a 64QAM (Quadrature Amplitude Modulation) modulation with an overall coding efficiency of R = 0.6 with R_MSB= 0.8 R_ISB= 0.67 and R_LSB= 0.33, where MSB represents the set of the most significant bits ("Most Significant Bits" in English), LSB represents the set of least significant bits (in English "Least Significant Bits") and ISB represents the set of intermediate bits ("Intermediate Significant Bits").

0066Thus, one point of the 64QAM constellation corresponds to a set of three bits, namely a bit assigned to the MSB level, a bit of the ISB level, and a bit of the LSB level.

0067By decoding the three levels of the QAM, MSB, ISB and LSB modulation, it is noted that the most robust level is that which corresponds to the MSB bits (curve referenced 41), then to the LSB bits (curve referenced 42) and finally to the level intermediate ISB (curve referenced 43), as illustrated by the <figref idref="f0003">figure 4</figref>. Indeed, the curve of TEB 41 associated with the MSB level is the one that decreases the most rapidly as a function of the signal-to-noise ratio (S / N), and the curve of TEB 43 associated with the ISB level is the one which decreases the most slowly in function of the signal to noise ratio (S / N).

0068However, we can analyze again the performances of the ISB and LSB levels, carrying out a re-looping of the most robust level (MSB), that is to say taking into account the result of the decoding of the MSB level during the decoding ISB and LSB levels.

0069It can be seen that the ISB level becomes the second most robust level, in front of the LSB level: thus, the decreasing order of robustness of the coding levels is MSB-ISB-LSB.

0070According to the technique proposed by the invention, the order of the decoding of the optimal MLCs is therefore the descending order of robustness MSB-ISB-LSB.

0071We now present, in relation with the <figref idref="f0004">figure 5</figref>, an embodiment of a receiver according to the invention.

0072The operation of such a receiver is based on four main principles:<ul id="ul0004" list-style="dash" compact="compact"><li>the first principle is based on the demodulation, then the decoding, first of all the most robust level, the result of such a decoding to improve the demodulation, and therefore the decoding, less robust levels. This operation is repeated until the least robust coding level;</li><li>the second principle implemented by a receiver according to the invention is that of an iterative process. Indeed, after demodulation and decoding of all the levels, the operation can be repeated, so as to improve the demodulation of the most robust level using the result of the decoding of the lower levels;</li><li>the third principle of operation is based on the implementation of a test of the relevance of the correction of the demodulated signal as a function of the amplitude of the corrective signal with respect to the signal to be demodulated;</li><li>finally, such a receiver uses extrinsic information, between each iteration, in order to improve the demodulation and therefore the decoding of the received signal.</li></ul>

0073These four principles are presented in more detail in relation to the <figref idref="f0004">figure 5</figref>, which presents a particular embodiment, in the context of a modulation QAM64 (Quadrature Amplitude Amplitude). It will of course be easy for the skilled person to generalize this description to any type of multilevel modulation.

0074In the particular embodiment of the <figref idref="f0004">figure 5</figref>, the most robust level corresponds to the coding level of the Most Significant Bits (MSBs) and the least robust level corresponds to the LSBs (Least Significant Bits). As previously explained in relation to the<figref idref="f0003">figure 4</figref>, the robustness of a level of coding vis-à-vis the noise is inversely proportional to the error rate of this level. In addition, the error rate is a function of the coding efficiency, the power associated with each bit (also called the bit level) and the signal-to-noise ratio (in fact, the errors observed on the signal depend, of course, on the noise who affects it).

0075It is therefore easy to understand that the most robust level is not necessarily the level of the most significant bits. By way of illustration, however, reference is made in the following description to an embodiment of the invention in this particular case.

0076The receiver of the <figref idref="f0004">figure 5</figref> comprises two stages referenced 51 and 52, corresponding to two successive decoding iterations. Indeed, the inventors have found that the improvement of the decoding performance resulting from the implementation of a third decoding iteration was low and, in a preferred embodiment of the invention, only two iterations of the decoding process. are implemented. This gives a good compromise between performance and complexity.

0077Firstly, the first decoding stage referenced 51 is described. This stage is powered by the received symbol MAQ64, also called S_r, which is distributed to the three demodulators referenced 511 to 513, respectively providing demodulations MSB, ISB and LSB. The received symbol S_r is formed of three bits <i>X_RMSB</i>, <i>X_RISB</i>, <i>X_RLSB</i> assigned to the MSB, ISB and LSB levels, which can be expressed as: S_r = <i>X_RMSB</i>+<i>X_RISB</i>+<i>X_RLSB</i>.

0078The first step implemented on receipt of the symbol S_r consists in demodulating the bits affected at the most-robust level to the noise, that is to say in this case the most significant bits (MSB). At the output of the demodulator 511, the demodulated bits are obtained<i>b₃ⁱ</i> , <i>i</i>=<i>1</i>,...,<i>not</i>, which feed the decoder referenced 514. After decoding by the decoder 514, the decoded bits are obtained <i>b₃ⁱ</i> , <i>i</i>=<i>1</i>,...,<i>not</i>.

0079The second step is to encode the decoded bits <i>b₃ⁱ</i> , <i>i</i>=<i>1</i>,...,<i>not</i> with the coder used on transmission, called "encoder 3", referenced 517. The bits thus coded feed the ISB demodulator referenced 512, which takes it into account to demodulate the intermediate-weight bits (ISB) <i>b₂ⁱ</i> , <i>i</i>=<i>1</i>,...,<i>not</i>. The demodulated intermediate weight bits are provided at the input of the decoder referenced 515, which delivers, after decoding, the decoded intermediate weight bits.<i>b₂ⁱ</i> , <i>i</i>=<i>1</i>,...<i>not</i>It should be noted that the level of ISB coding is here the level of robustness to intermediate noise, and that it is therefore demodulated and decoded directly after the MSB level.

0080The decoded intermediate weight bits <i>b</i>₂<i>ⁱ</i> , <i>i</i>=<i>1</i>,...,<i>not</i> are further provided at the input of the encoder referenced 518, which is identical to the encoder used for transmission for the ISB level.

0081By using the recoded bits of the higher robustness levels (MSB and ISB), it is then possible to demodulate the bits of the least robust level, which, in the particular embodiment described in relation to the <figref idref="f0004">figure 5</figref>, corresponds to the level of the LSBs.

0082To do this, the LSB demodulation device referenced 513 is powered by the recoded bits from the coders referenced 517 and 518 of the most robust levels MSB and ISB, and delivers the demodulated bits of low weight. <i>b</i>₁<i>ⁱ</i>, <i>i</i>=<i>1</i>,...,<i>not</i>. After decoding by the decoder referenced 516, the decoded low-order bits are obtained.<i>b</i>₁<i>ⁱ</i>, <i>i</i>=<i>1</i>,...,<i>not</i>.

0083Low decoded bits <i>b</i>₁<i>ⁱ</i> can further feed the coder referenced 519, which is identical to the encoder used in the transmission for the LSB level.

0084After decoding the 3 levels of the MAQ constellation, it is possible to determine (520) an estimate of the transmitted symbol, from the recoded bits delivered by the three coders referenced 517 to 519.

0085Thus, in the particular embodiment described in connection with the <figref idref="f0004">figure 5</figref>, the emitted symbol S_e is of the form S_e = 4b_MSB + 2b_ISB + b_LSB, where b_MSB, b_ISB and B_LSB correspond respectively to the bits of the MSB, ISB and LSB levels.

0086From the estimated transmitted symbol, the Euclidean distance between the emitted symbol S is calculated_e and the received symbol S_r, by weighting this distance by a coefficient α (0 &lt;α &lt;1). An extrinsic information α (S_r-S_e) 521, which can be used in the second stage 52 of the receiver, to improve the decoding of the following iterations.

0087The second decoding stage 52 operates in a manner similar to the first stage referenced 51. It comprises in particular three demodulation devices referenced 521 to 523, and three decoders referenced 524 to 526 associated respectively with the three levels of coding MSB, ISB and LSB.

0088The first step implemented within this stage 52 is the demodulation of the most robust level MSB by the block referenced 521. Such a block 521 is fed, on the one hand, by the recoded bits of the less robust ISB and LSB levels. from coders referenced 518 and 519 of the first decoding stage 51, and secondly by the received symbol S_r subtracted from extrinsic information α (S_r-S_e), ie S_r (1-α) + αS_e.

0089The coefficient α is preferably chosen close to 0.25. In an alternative embodiment, the value of the coefficient α is optimized as a function of the signal-to-noise ratio. In this way, depending on the signal-to-noise ratio, it is possible to choose to give a greater or lesser degree of confidence to the estimate 520 of the transmitted symbol, to take more or less into account during the second decoding iteration, and in particular when decoding the most robust MSB level.

0090Thus, if the signal-to-noise ratio is very good, we will choose α close to 1. In the opposite case, α will be chosen close to 0.

0091Such an optimization of α may in particular be preceded by a step of determining the signal-to-noise ratio, by means of drivers, inserted into the transmitted signal. According to a known technique, the pilots constitute reference information, the value of which is known a priori of the receiver. By comparing this predetermined value of the pilots with the value of the received pilots, the receiver can, by division, estimate the transfer function of the transmission channel, and therefore the signal-to-noise ratio affecting the transmitted signal. This technique also makes it possible to evaluate the robustness of the different levels of coding.

0092After demodulation by the block referenced 521, new demodulated bits are obtained <i>b</i>₃<i>ⁱ</i> , <i>i</i>=<i>1</i>,...,<i>not</i>, improved with respect to the corresponding bits coming from the demodulation block referenced 511, because of the joint consideration of the extrinsic information and the decoding result of the less robust levels LSB and ISB of the first decoding stage 51.

0093These bits <i>b</i>₃<i>ⁱ</i>, <i>i = 1, ..., n</i> demodulated feed the decoder referenced 524, which delivers improved decoded bits <i>b</i>₃<i>ⁱ</i> , <i>i = 1, ..., n.</i> As before, these bits are recoded by the coder, identical to that of the transmission, referenced 527, then feed the ISB level demodulation device referenced 522. This demodulation device 522 is further supplied with input by the difference between the received symbol and extrinsic information, in the form S_r (1-α) + αS_e, and by the recoded bits of the least robust level LSB, from the coder referenced 519.

0094ISB 522 demodulator delivers demodulated intermediate weight bits <i>b</i>₂<i>ⁱ</i> , <i>i = 1, ..., n,</i> which feeds the decoder referenced 525 delivering improved decoded bits <i>b</i>₂<i>ⁱ , i</i>=<i>1</i>,...,<i>not</i>.

0095Again, these bits <i>b</i>₂<i>ⁱ</i> , <i>i</i>=<i>1</i>,...,<i>not</i> are recoded by means of the encoder, identical to that used on transmission, referenced 528, then are supplied as input to the least-robust level demodulation block LSB referenced 523. This demodulation block referenced 523 is further powered by the difference between the received symbol and the extrinsic information, in the form S_r (1-α) + αS_e. It therefore delivers demodulated bits<i>b</i>₁<i>ⁱ</i>, <i>i</i>=<i>1</i>,...,<i>not</i> improved relative to the corresponding bits from the first decoding stage referenced 513, which feed the decoder referenced 526, the latter delivering improved decoded bits <i>b</i>₁<i>ⁱ</i> , <i>i</i>=<i>1</i>,...,<i>not.</i>

0096The implementation of these two successive decoding iterations, combined with the use of extrinsic information, makes it possible to obtain improved performance results compared with the techniques of the prior art, and in particular with respect to the decoding method. suboptimal steps conventionally implemented for the decoding of modulated signals according to a multi-level coding technique.

0097These performances are illustrated by the curves of the <figref idref="f0003">figure 6</figref>, respectively presenting the bit error rates obtained, as a function of the signal-to-noise ratio S / N, for the decoding method of the invention on the one hand, and for the method of sub-optimal decoding of the state of the technique on the other hand.

0098Thus, it can be seen that the curve referenced 61 of the bit error rate obtained according to the invention decreases much more rapidly, as a function of the S / N ratio, than the BER curve referenced 62 obtained according to the stepwise decoding method of FIG. prior art.

0099As part of the previously mentioned DRM consortium, it has been established that a BER bit rate of 10^-4 was the operating threshold of the system. We note, next to the<figref idref="f0003">figure 6</figref>, that a gain of about 2 dB is obtained, for this threshold of 10^-4between the decoding technique of the prior art and the decoding method of the invention.

0100One could further improve the performance of the system of the invention vis-à-vis Doppler affected transmission channels in particular, by adding an interleaver, on each coding level, the transmission. The receiver of the<figref idref="f0004">figure 5</figref> then include means deinterlacing, which would be implemented for each level, after demodulation and before decoding.

0101Note that any type of code can be used in the context of the invention, including turbo codes. In particular, it is possible to envisage implementing a turbo-code for each of the coding levels.