WO2007003837A2

WO2007003837A2 - Method for the iterative processing of symbol bursts in a receiver

Info

Publication number: WO2007003837A2
Application number: PCT/FR2006/050607
Authority: WO
Inventors: Julie Yuan-Wu
Original assignee: France Telecom
Priority date: 2005-06-28
Filing date: 2006-06-16
Publication date: 2007-01-11
Also published as: FR2887713A1; WO2007003837A3

Abstract

The invention relates to a method for the iterative processing of symbol bursts in a receiver. In a radio communication receiver that performs the iterative processing of received symbol bursts (SLR), a weighting module (MP) pre-stores a sequence (Seq) of weighting coefficients which are each associated with an iteration. The weighting of a series of reconstructed bursts (SLC) in a return module (MR) is controlled in a symbol detector (DS) by the weighting coefficient associated with each iteration relating to a series of received bursts. A module (MA) stops the detection of errors by a controller (CC) for the decoded block relating to the series of received bursts and stores the decoded block when said block contains no errors. Symbol detection is destabilized in a controlled manner by the sequence of weighting coefficients, which rapidly improves the quality of the decoded blocks.

Description

Iterative processing of symbol bursts in a receiver

The present invention relates to an iterative processing of bursts of symbols received by a digital radio receiver included for example in a mobile radio terminal or a base station.

Currently, digital radiocommunication receivers use the iterative processing of propagation channel estimation and symbol detection to provide high quality emission data retrieval despite high complexity. In a known manner, these receivers use linear equalizers per block based on the zero forcing criterion ZF ("Zero Forcing", in English) or on the mean minimum square error criterion MMSE ("Minimum Mean Square Error", in English) to reduce interference due to propagation. At each iteration, a feedback loop improves the accuracy of channel estimation by applying equalized and decoded data at the previous iteration to a symbol detector and a channel estimator. An analysis error occurred especially at the beginning of such an iterative procedure irreparably generates estimates and erroneous analyzes for the following iterations. The data decoded in these receivers do not change after a certain number of iterations, which implies a stabilization of the quality of reproduction of the emission data. Thus, the possibly high number of iterations results in a high consumption of energy in the receivers. The aim of the invention is to improve the performance of the iterative processing relating to channel estimation and the detection of symbols in a digital radio communication receiver, by destabilizing the quality of the data restitution, which leads to a reduction in the number of data. iterations and consequently an increase in processing speed, for an increased quality of the data blocks provided by the receiver.

To achieve this objective, a radio receiver receiving a series of symbol bursts, each bursts including a training sequence, the receiver comprising a detection means for detecting symbols in the received bursts, a decoding means for decoding each series symbol bursts detected in a decoded block including a cyclic redundancy code, return means for iteratively constructing a series of reconstructed bursts based on the decoded block and a prestored learning sequence, and a channel estimator determining iteratively a channel estimate as a function of the series of bursts received, the series of reconstructed bursts and the prestored learning sequence to improve the detection of symbols in the reconstructed bursts as a function of the channel estimate, characterized in what it includes: a way to prememorate a predefined sequence of coefficients each associated with an iteration, one of the weighting coefficients being substantially equal to zero, and means for controlling in the means for detecting symbols a serial salvage weighting reconstructed by the weighting coefficient associated with each iteration relative to a series of received bursts.

The invention controllably destabilizes the detection of symbols by means of the predefined sequence of weighting coefficients which improves the quality of the data blocks provided by the receiver.

In order to effectively reduce the number of iterations for each series of received symbol bursts and thus the receiver processing time, the receiver may further comprise: cyclically redundant code check means for detecting at least one error in each block decoded by the decoding means, means for applying the decoded block to the return means as long as the block contains at least one error and as long as the iterations associated with all the weighting coefficients of the sequence for the series of bursts corresponding to the block decoded are not performed, and means for stopping the control of the decoded block in the control means and storing the decoded block when the block contains no error.

The decrease in the number of iterations advantageously leads to a relatively short processing time for transmitting the decoded blocks to the upper layers, and a decrease in the power consumed by the receiver.

The invention is also applicable to a multi-user receiver which processes in parallel series of bursts of received symbols respectively attached to several users. The receiver then comprises, in parallel for the users respectively, symbol detection means, detected symbol burst series decoding means, cyclic redundancy code checking means for detecting errors in the decoded blocks, means for stopping iterations and storing decoded blocks, return means for iteratively constructing reconstructed salvo sets respectively as a function of the decoded blocks and prestored learning sequences, and applying the reconstructed salvo sets to the channel estimator and the means control. In the multi-user receiver, then each control means no longer detects at least one error in the decoded blocks relating to a series of symbol bursts as soon as a decoded block relating to said series of symbol bursts contains no errors and as long as at least one error is detected in another decoded block, and the means for stopping transmit the stored decoded blocks to upper layers of the receiver when all the blocks contain no errors. According to another characteristic of the invention, in the multi-user receiver, the means for stopping transmit the decoded blocks to the upper layers of the receiver further when the number of iterations for the decoded blocks reaches a predetermined number of iterations.

The invention also relates to a method for iteratively processing bursts of received symbol bursts each including a training sequence in a radio communication receiver. The method comprises a detection symbols in the bursts received, a decoding of each series of bursts of symbols detected in a decoded block including a cyclic redundancy code, an iterative construction of a series of reconstructed bursts as a function of the decoded block and a sequence of bursts. prestored learning, and iterative estimation of a channel estimate based on the received burst series, the reconstructed salvo series, and the prestored learning sequence to improve symbol detection in reconstructed salvos based on of the channel estimate. According to the invention, the method comprises steps of: for all the series of received symbols, prememoring a predefined sequence of weighting coefficients each associated with an iteration, one of the weighting coefficients being substantially equal to zero, and for each iteration, control for the detection of symbols a series weighting of salvos reconstructed by the weighting coefficient associated with said iteration.

Other features and advantages of the present invention will appear more clearly on reading the following description of several preferred embodiments of the invention, given by way of non-limiting examples, with reference to the corresponding appended drawings in which: FIG. 1 is a schematic block diagram of a single-user radiocommunication receiver according to the invention;

FIG. 2 is a schematic block diagram of a multi-user radiocommunication receiver according to the invention; FIG. 3 is a schematic algorithm of a treatment method according to the invention; and

FIGS. 4 and 5 are respectively experimental diagrams of an error rate per block and a bit error rate as a function of a signal-to-noise ratio according to the method of the invention.

In general, the invention described hereinafter relates to a digital radio receiver including a propagation channel estimator for estimating the response of the data propagation channel between a transmitter and the receiver and a symbol detector for detecting symbols in a signal transmitted by the transmitter through the propagation channel and received by the receiver. The receiver is for example included in a mobile radio terminal or in a base station in a radiocommunication network, for example GSM (Global System for Mobile Communications), EDGE (Enhanced Data rates for GSM Evolution) or UMTS (Universal Mobile). Telecommunications System), or a receiver operating in a WLAN (Wireless Local Area Network) or DECT (Digital Enhanced Cordless Telecommunications) wireless LAN.

The receiver thus processes, as an example, a signal for TDMA time division multiple access networks, or CDMA code division multiple access networks such as CDMA broadband networks or synchronous CDMA networks. with time distribution.

In the case of a CDMA network, ISI (Intersymbol Interference) intersymbol interference in a user signal and multiple access interferences MAY (Multiple Access Interference) between signals of multiple users are generated by the propagation in the channel and degrade the quality of the received signal. These impairments are reduced by an estimation of the channel transfer function which is carried out using information known in advance of the receiver, such as a training sequence (for example "midamble" in English) transmitted in a salvo. ("burst" in English) by the issuer. But the additional noise at the reception of the signal deteriorates this estimate and tempers the reduction of the interferences. The receiver then processes in a loop data corresponding to the signal received and noisy through the channel, in order to improve the estimation of the channel. Such a receiver is referred to as an iterative receiver.

In Figure 1 are only shown means included in the iterative receiver RI in connection with the invention. These means are a DS symbol detector and an MD decoding module, and a feedback loop comprising a CC cyclic redundancy code controller, an EC channel estimator, a PM weighting module, a MA stop module, a MR return module and an MX multiplexer. The feedback loop is said to be hard decision since it connects the output of the decoding module MD to the input of the EC channel estimator, bits provided at the output of the decoding being hard information bits.

In a known manner, an amplification, frequency demodulation and filtering stage adapted in the receiver is applied to the descrambled received signal of bursts of complex symbols SLR, also called packets, shaped into a band. base at the entrance of the DS symbol detector. The received signal corresponds to a digital data signal transmitted by the transmitter in which the data signal has undergone block-bit processing in a code generator. After each block, the code generator introduces parity bits constituting a cyclic redundancy code CRC as a function of the block and a cyclic generator polynomial. Then the blocks including their CRC codes undergo such a coding, interleaving, a phase modulation for example four-state to produce complex symbols, a spread spectrum by burst (bursts) for a CDMA coding, an insertion of a sequence premored learning, scrambling and finally filtering, amplification and frequency conversion in an output stage.

An output of the stop module MA provides a received signal composed of BC-corrected blocks to upper layers CS of the receiver. In particular, the useful information in each block is extracted by the upper layers for example to be reproduced in an audio circuit or transcribed by a graphical user interface of the receiver when it is included in a mobile terminal, or transmitted through the fixed part of a radio network when the receiver is included in a network base station.

In the general case of a CDMA multi-user transmission which will be referred to most often in the following description, the receiver RI receives a signal composed of several mixed bursts SLRi to SLRK assigned to K users. A block to be decoded, corresponding to a user associated with a propagation channel relative to the receiver RI, is composed of a series of bursts. Each salvo occupies a predetermined time slot and includes sequences of code elements (chips). In the case of a TDMA receiver, the received signal is divided into periodic time slots allocated to users.

For the sake of clarity, the invention is first described with respect to received bursts relating to a user, in particular for the downlink in a mobile terminal, with reference to FIG. 1. A received block is deduced from several bursts of symbols. SLR each including a training sequence (eg "midamble") SA which is a predetermined sequence of symbols having particular autocorrelation properties. The training sequence SA is previously stored in the receiver RI in order to be correlated to the training sequence of a salvo received SLR in the channel estimator EC which determines a channel estimate h defined by the coefficients of the transfer function of the propagation channel through which the SLR burst was transmitted. The EC channel estimator provides an A channel estimate at the symbol rate so that the DS symbol detector detects each symbol included in each received burst. The symbol detector DS stores a series of bursts of received symbols corresponding to a block to be decoded to process the series of bursts for several iterations via the feedback loop and thus perfect the decoding of a block. The symbol detector DS comprises for example several parallel propagation paths for treating each symbol of the received symbol burst SLR as symbols delayed by the multiple paths of the propagation channel. Each channel delays the received symbol and includes for example an interference canceller intersymbol and a filter adapted to a sequence of spreading code elements and an estimation of the respective multipath. The outputs of the matched filters deliver symbol samples to transverse filters whose coefficients are determined iteratively as a function of the channel estimate h _r and then summed to provide a symbol equalized to the decoding module MD. For example, cross filters are a linear equalizer or a maximum likelihood equalizer.

The received bursts SLR are thus converted into bursts of SLE equalized detected symbols, without a training sequence, which are applied to the decoding module MD. The decoding module includes in cascade a bit symbol converter such as a four-phase phase demodulator matching a complex symbol + j, +1, -1 and -j to a respective pair of consecutive bits (0.0 ), (0,1), (1,0), (1,1), a channel de-interleaver for inhibiting transmission interleaving, a convolution decoder or a turbo-decoder, and a threshold circuit binary decision. The output of the decoding module MD delivers decoded bursts SLD which are grouped into decoded binary blocks BD which normally reconstruct the emitted signal composed of blocks with their cyclic redundancy codes CRC.

Conventionally, the feedback loop connects the output of the decoder module MD to the inputs of the symbol detector DS and the EC channel estimator via the feedback module MR. The feedback loop imposes an iterative operation on the series of bursts in the blocks at the receiver RI by treating at each iteration the bursts received from a block up to what this one is corrected. The MR return module thus reconstructs the bursts of SLC symbols, as in the transmitter, but on the basis of the bits in the decoded BD blocks produced by the decoding module MD. The return module MR comprises, in cascade, reciprocal functional means of those constituting the decoding module MD, such as a convolutional encoder or a turbo-encoder, an interleaver, a bit converter with a complex symbol such as for example a modulator of four-state phase, and a burst formatter for periodically inserting into a predetermined number of symbols the previously stored training sequence SA.

Each decoded block BD delivered by the decoding module MD contributes to the formation of SLC reconstructed bursts which constitute iteration data transmitted to the EC channel estimator and to the DS symbol detector for another iteration. A reconstructed burst SLC although including a training sequence may still include errors. The EC channel estimator then uses the reconstructed burst SLC as an additional source of information for finer estimation and transmits a new estimated h-channel to the DS symbol detector. The symbol detector thus has the new channel estimate, the reconstructed burst SLC and the received burst memory SLR of the initial iteration, for a more accurate detection. Thus, at each iteration, the training sequence SA, which undergoes a burst operation in the return module MR, is processed by the EC channel estimator and the DS symbol detector.

According to the prior art, the so-called iterative procedure described above stops after a number of predetermined iterations when the data in the decoded block processed by the receiver converge and become invariant in order to provide them to the upper layers CS of the receiver and inform the latter of the transmission quality of the physical layer comprising the data of the decoded block. The performances according to said procedure converge in terms of error rate per bit BER ("Bit Error Rate") and BLER block error rate ("Block Error Rate", in English), applied to the decoded data. provided at the last iteration.

The weighting module MP introduced into the receiver RI according to the invention destabilizes, for example substantially periodically, the conventional iterative procedure to avoid convergence to a "local" minimum leading to an infinite iterative procedure. The cyclic redundancy code controller CC controls each decoded block BD at the output of the decoding module MD to detect errors. The stop module MA in conjunction with the DC controller stores information provided by the DC controller to reduce the number of iterations performed by the RI receiver.

For K users, the multi-user CDMA receiver, in particular for the upstream channel in a base station, processes in parallel K blocks having bursts coded with sequences of spreading code elements respectively allocated to the users and including sequences of SAi learning in SAK.

As shown diagrammatically in FIG. 2, the symbol detector DS jointly detects, in parallel, symbols in K mixed bursts received SLRi to SLRK from the users K and the decoding module MD decodes in parallel K sets of bursts of equalized detected symbols SLEi to SLEK provided by the DS symbols detector in K bursts SLDi to SLDK- For a user referenced by the index k, with 1 <k <K, successive decoded bursts SLD ^ without a learning sequence are grouped at each iteration into a decoded binary block BD ^.

The decoded binary block BD ^ is fed back into a respective feedback loop in order to reconstruct respective bursts SLC _k at the output of the feedback module MR as a function of the respective training sequence SA ^ previously stored in the receiver. Thus the receiver has K parallel return loops. In these loops, the DC controller checks K cyclically redundant codes in decoded BDi to BDK binary blocks for errors, the stop module MA progressively stores K information provided by the DC controller, the MR feedback module. reconstructed by means of K chains each comprising a convolutional encoder, for example a turbo-encoder, an interleaver, a phase modulator, for example a four-state modulator, and a burst formatter, K series of bursts of respective symbols SLCi to SLCK to from the decoded blocks BDi to BDK and learning sequences SAi to SAK previously stored in the receiver, the EC channel estimator estimates in parallel K channel estimates _{1 1} to h _κ defined by the transfer functions of the propagation channels respective for the K users, and the weighting module MP applies in parallel the same symbol weighting to K series of mixed bursts received SLRi to SLRK in the DS symbol detector. Filters in the DS symbol detector are respectively adapted to the spreading code element sequences respectively allocated to the users and to the diversity of the multiple paths in the propagation channels on which the channel estimates ^ ₁ to h _κ depend.

With reference to FIG. 3, the processing method according to the invention comprises steps E1 to E8 executed automatically in the iterative receiver R1.

Step E1 predefines a weighting sequence Seq that the weighting module MP premades. The weighting sequence comprises weighting coefficients α, such that 0 <α <1, each associated with an iteration iterating. At least one of the coefficients of the weighting sequence is substantially zero. For example, for a predetermined number of iterations ITER = 14, the weighting sequence Seq = {0C, α, ... α} is composed for example of hard value coefficients (hard) {0; 1; 1; 1; 1; 0; 1; 1; 1; 1; 0; 1; 1; 1; 1}, or soft (soft) {0; 0.7; 0.8; 1; 1; 0.1; 0.8; 0.9; 1; 1; 0.1; 0.8; 0.9; 1; 1}. The Seq weighting sequence is predefined according to the characteristics of the propagation channel for adaptation and optimization provided for the iterations of the method.

The weighting coefficient α takes a value close to the limits of the interval [0,1]. Preferably, as in the two examples of weighting sequence above, the weighting coefficients having a value substantially equal to "0" and the weighting coefficients having a value substantially equal to "1" are distributed substantially regularly in the Seq weighting sequence; in this case, the Seq weighting sequence is said to be regular. As a variant, the weighting sequence Seq is said to be irregular when the weighting coefficients having a value close to "0" and "1" are irregularly distributed in the weighting sequence Seq. The weighting coefficient OC is at a first state when it has a value other than "0". As previously described, the receiver RI receives at least one received burst SLR which is processed at an initial iteration and reconstructed into a burst SLC which is applied to the DS symbol detector and again analyzed and retransmitted as a decoded burst SLD of a block decoded by the decoding module MD to reconstruct another SLC burst in the return module.

The reconstructed symbol bursts SLC _k corresponding to a user k are associated with an error control indicator ICE (BD _k ), for example which is stored at a first logic state "0" in the DC controller and the module of MA stop at the initial iteration.

In step E2, each decoded block BD _k supplied by the decoding module MD is controlled by the cyclic redundancy code controller CC before being applied to the stop module MA in order to detect at least one error in the block . The control of the decoded block is performed by recalculating a cyclic redundancy code on the useful bits of the block and comparing it with the cyclic redundancy code extracted from the block in order to determine the ICE error control indicator (BD _k ) which is at first logic state "0" when the block contains no error and a second logic state "1" when the block contains at least least an error. The DC controller reads the logic state of the ICE error control flag (BD _k ) from the decoded block BD _k to validate the decoded block.

When the ICE error control indicator (BD _k ) of the decoded block BD _k is at the second state "1", the decoded block is invalidated, and if it is not, the decoded block is validated and the process passes from the step E2 in steps E3 and E4. When all the error control indicators ICE (BDi) to ICE (BDK) are in the second state "1", the process goes from step E2 to step E5.

In step E3, the stop module MA stores the decoded block BD _k as corrected decoded block BC _k when the error control indicator ICE (BD _k ) is in the first state "0". As long as at least one block decoded in parallel with the decoded block BD _k and relating to another user leads to an error control indicator at the second state "1", that is to say as long as the K indicators of error control ICE (BDi) to ICE (BDK) are not in the first state "0", the corrected decoded block BC _k stored in the stop module MA is considered valid and is no longer controlled by the controller CC at the following iterations. Furthermore, the corrected decoded block BC _k stored is applied by the stop module MA to the return module MR at the following iterations until the end of the iterative procedure. The corrected decoded block BC _k optionally has an ICE error control indicator (BD _k ) at the first state "0" but is no longer considered by the stop module MA. In this case, the ICE indicator (BD _k ) is stored in correspondence with the corrected block BD _k , indicating that the decoded block BD _k has been validated by the controller CC. Thus, the memorization of the decoded block as and when Iterations ensure correct results that are used until the end of the iterative procedure.

In step E4, the stop module MA reads the number of decoded blocks validated by the DC controller for the K users. For example, the stop module MA reads the K error control indicators ICE (BDi) at

ICE (BDK) stored and checks whether all these indicators are in the first logical state "0". If the previous condition is realized and therefore all the decoded blocks BCi to BCK have been validated by the controller

CC, step E8 of the method is executed, and if not, steps E5 to E7 are executed.

At the end of step E2 when the ICE error control indicators (BD _k ) decoded blocks BD ^ ₁ are in the second state "1", or at the end of step E4 when that least one of the decoded blocks in parallel BDi to BDK presents an error control indicator ICE (BDi) to ICE (BDK) at the second state "1", as previously described, the decoded blocks in parallel BDi to BDK some of which can be stored corrected blocks are transferred in parallel from the stop module MA to the return module MR in step E5. Naturally, the step E5 is executed as long as the ITER Iterations associated with all the weighting coefficients of the sequence for the series of bursts corresponding to the decoded block are not realized. The feedback module MR reconstitutes in respective bursts SLCi to SLCK the transmission structure of the received bursts SLRi to SLRK, in particular by converting the bits into symbols by phase modulation. The MX multiplexer at the output of the return module MR inserts the respective premixed learning sequences SAi to SAK into the reconstructed salvo series SLCi at SLCK to transmit them in parallel to the EC channel estimator and the DS symbol detector.

In step E6, the weighting module MP controls in the symbol detector DS a weighting of the reconstructed burst series SLC1 to SLCK by a weighting coefficient OC according to the weighting sequence Seq. For example, at iteration = 5, the reconstructed bursts are weighted by the coefficient OC. The weighting coefficient indicates the reliability with which the DS symbol detector uses reconstructed bursts to reduce propagation error effects, such as ISI intersymbol interference and MAY multiple access interference. The higher the oc weighting coefficient is close to the hard value "1", the higher the confidence in reconstructed bursts.

In step E7, the DS symbol detector weights the reconstructed bursts SLC1 to SLCK by the weighting coefficient OC, when the weighting coefficient OC for an iteration is in the first state and therefore is different from 0. Also at the initial iteration at the following iterations the EC channel estimator operates relentlessly and reconstructed salvs SLCi to SLCK are applied to it at each subsequent iteration to produce new channel estimates λ ₁ to h _κ at the DS symbol detector.

When OC = 0, the weighting module MP controls the multiplexer MX so that at least the symbols useful in the reconstructed bursts SLCi to SLCK are not transmitted to the symbol detector DS, which constitutes a so-called "release" operation. "hard decision": the DS symbol detector processes the bursts received as it was at the initial iteration SLRi to SLRK and the SAi preloaded learning sequences to SAK instead of reconstructed bursts, but also new channel estimates transmitted by the EC channel estimator. When the weighting coefficient OC has a value between 0 and 1, the "release" operation is called "soft decision". The DS symbol detector then operates as follows. One of the signals received by the symbol detector DS, corresponding to a received burst SLR for a user k at a given instant, is expressed for example in the form: r = s + I + n, where s is the part of the signal itself, I is the sum of the intersymbol interference ISI and the multiple access interference MAI, and n is the Gaussian noise, in the case of a CDMA receiver at the initial iteration. If the symbol detector DS is an interference canceling detector in which the interference in the other propagation paths is subtracted from a symbol processed in a propagation channel, the signal, following the cancellation of interference after a number iter iterations,

,, _^ (iter) (iter) - (iter) - (iter) takes the form: r = r - OC I. I is the sum of the interference to which the weighting coefficient is applied and which is estimated in the DS detector from the SLCi burst series SLCK reconstructed at each iteration. Thus, when OC = 0, the signal after the cancellation, resumes the form of the signal received at the initial iteration, after a number of iterations

(iter) iter, since r = r.

The iterative processing of the symbol detector is typically intended to find a "global minimum" that corresponds to the best performance of the receiver. However, for a given optimization criterion, there is always a "local minimum" that corresponds to certain performance of the receiver. In general, the iterative procedure of the symbol detector converges and stabilizes on local minima. In fact, when a decoded block has a certain error rate, the symbol detector propagates these errors since the analysis of reconstructed bursts in the detector DS is erroneous and the errors are taken into account at each iteration. The iterative processing then converges to a "local minimum" corresponding to said error rate. The "release" operation thus gives more freedom to the convergence of the iterative procedure which avoids the definitive retention of erroneous blocks.

The reconstructed SLCi SLCK bursts analyzed in the symbol detector are then transmitted to the decoding module MD and then to the controller CC after step E7 to execute step E2 again at the next iteration.

In step E8, when all the decoded bit blocks BDi to BDK have been validated by the DC controller in BCi to BCK corrected blocks stored in the stop module MA, all ICE error control indicators (BDi) at Since ICE (BDK) is in the first state "0", the iterative procedure is completed and the stop module MA transmits the corrected blocks BCi to BCK to the upper layers CS of the receiver.

Thus, the end of the iterative procedure is determined by the controller CC and the effective number of iterations is necessarily less than or equal to the predetermined number of iterations ITER. In addition, at the end of step E8, the stop module controls the reinitialization of the receiver R1, in particular of the weighting sequence Seq in the weighting module MP, in order to process a new set of mixed received bursts SLRi to SLRK and thus of decoded parallel blocks BDi to BDK-

With reference to FIGS. 4 and 5, the receiver according to the invention is a time division synchronous ARMC receiver in accordance with the TDD (Time Division Duplex) mode of the UMTS. The iterative procedure uses the OS observation separation principle, described in the article by Yi Yuan-WU et al. The symbol detector uses a filter adapted to the initial iteration (iter = 0) and the filter is adapted to the initial iteration (iter = 0). interference cancellation followed by filtering adapted to the following iterations (iter> 0). The EC channel estimator uses a known pseudo-inversion operation that provides a channel estimate per symbol. The characteristics of the services tested and the receiver considered are as follows:

- uplink (mobile to base station); - 12.2 kbit / s rate;

- propagation environment: channel of the type "ITU Vehicular_A";

relative speed between a mobile terminal and a base station: v = 3 km / h; - assumed channel length: w = 57 pieces of code (chips);

- spreading factor of the ARMC codes: Q = 8.

The curves in Figure 4 represent the block error rate (BLER) as a function of the ratio signal to noise (Eb / NO) with a user number K equal to 1, 4 and 8 respectively from bottom to top, according to iterative procedures with or without a hard decision "release" operation. The small dotted curve (with annotation single_user_data_known) represents the ideal case where the channel estimator receives data without error. For the sake of simplification, the procedural performances designate in the following the performances of the iterative procedure according to the error rate per block applied to the receiver according to the invention.

For K = I, the predetermined number of iterations is ITER = 26 and the "release" operation is performed regularly after 8 iterations. The five measuring points represented by straight crosses (annotated by RL8 26) correspond to the procedural performances with "release" operations. The long dashed curve corresponds to the procedure performance without a "release" operation. For K = 4, the predetermined number of iterations is ITER = 44 and the "release" operation is performed regularly after 8 iterations. The six measurement points represented by stars (annotated by RL8 44) correspond to the procedural performances with "release" operations. The solid line curve corresponds to the procedure performance without "release" operation.

For K = 8, the predetermined number of iterations is ITER = 90 and the "release" operation is performed regularly after 12 iterations. The seven measuring points represented by straight crosses (annotated by RL12 90) correspond to the procedural performances with "release" operations. The alternating dashed curve corresponds to the procedure performance without a "release" operation. The error rate per block for K = I is very close to that for the ideal case of error-free data and the method according to the invention provides a gain of about 0.2 dB in the signal-to-noise ratio with respect to the known procedure of separation of observation. For K = 4, the method according to the invention provides a gain of 0.2 to 0.4 dB. For K = 8, the gain obtained is 0.3 to 0.6 dB.

The curves of FIG. 5 represent the error rate per bit (BER) as a function of the signal-to-noise ratio (Eb / NO), in a manner similar to the curves of FIG. 4. Similarly, for K = I, the method according to the invention provides a gain of about 0.2 dB in the signal-to-noise ratio compared to the known observation separation procedure. For K = 4, the process according to the invention provides a gain of 0.2 to

0.3 dB. For K = 8, the gain obtained is 0.2 to 0.5 dB.

The method according to the invention reduces the iterations and therefore globally the simulation time of the receiver. For example, for K = 8, the "release" operation decreases the simulation time by about 12 times compared to the observation separation procedure.

The invention described herein relates to a receiver and a method for iteratively processing bursts of received symbol bursts so as to weight the processing iterations especially at the level of the symbol detection. According to a preferred implementation, the steps of the method of the invention are determined by the instructions of a computer program incorporated in a receiver, such as the receiver RI. The program includes program instructions which, when said program is loaded and executed in the receiver whose operation is then controlled by the execution of the program, perform the steps of the method according to the invention.

Accordingly, the invention also applies to a computer program, including a computer program on or in an information carrier, adapted to implement the invention. This program can use any programming language, and be in the form of source code, object code, or intermediate code between source code and object code such as in a partially compiled form, or in any other form desirable to implement the method according to the invention.

The information carrier may be any entity or device capable of storing the program. For example, the medium may comprise storage means or recording medium, such as a ROM, for example a CD ROM or a microelectronic circuit ROM or a USB key, or a magnetic recording means, for example a floppy disk or a hard disk.

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 can in particular be downloaded to an Internet type network.

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 carrying out the method according to the invention.

Claims

1 - Radio receiver receiving symbol burst series (SLR), each burst including a training sequence, the receiver comprising a detection means (DS) for detecting symbols in the received bursts, a decoding means (MD) ) to decode each series of detected symbol bursts into a decoded block including a cyclic redundancy code, a return means (MR) for iteratively constructing a series of reconstructed bursts (SLC) according to the decoded block and a sequence of preassisted learning, and a channel estimator (EC) iteratively determining a channel estimate as a function of the series of bursts received, the series of reconstructed bursts and the prestored learning sequence to improve the detection of symbols in reconstructed bursts according to the channel estimate, characterized in that it comprises: means (MP) for prestoring a predefined sequence (Seq) weights

(α) each associated with an iteration (iter), one of the weighting coefficients being substantially equal to zero, and means (MP) for controlling in the means for detecting symbols (DS) a serial weighting of reconstructed bursts (SLC) by the weighting coefficient associated with each iteration relative to a series of received bursts (SLR).

2 - Receiver according to claim 1, further comprising: cyclic redundancy (CC) code checking means for detecting at least one error in each decoded block by the decoding means, means (MA) for applying the decoded block to the return means (MR) as long as the block contains at least one error and as long as the iterations associated with all the weighting coefficients of the sequence for the series of bursts corresponding to the decoded block are not realized, and means (MA) for stopping the control of the decoded block in the means of check and memorize the decoded block when the block contains no errors.

3 - Receiver according to claim 2, characterized in that it comprises in parallel for series of bursts of received symbols (SLRi- SLR _K ) respectively attached to several users, means for detecting symbols (DS), means for decoding detected symbol burst series (MD), cyclic redundancy code (CC) control means for detecting errors in the decoded blocks, means (MA) for stopping iterations and storing decoded blocks, return means (MR) for iteratively constructing reconstructed salvo sets respectively as a function of the decoded blocks and of the prestored learning sequences and applying the reconstructed burst series to the channel estimator (EC) and the control means, that each control means (CC) no longer detects at least one error in the decoded blocks relating to a series of symbol bursts as soon as a decoded block relating to said series of symbol bursts contains no errors and so many at least one error is detected in another decoded block, and in that the means for stopping (MA) transmit the stored decoded blocks (BCi to BC ^) to upper layers (CS) of the receiver when all the blocks do not contain no errors.

4 - Receiver according to claim 3, characterized in that the means for stopping (MA) transmit the memorized decoded blocks (BC] - to BCK) to the upper layers (CS) of the receiver further when the number of iterations for the decoded blocks reaches a predetermined number of iterations.

A method for iteratively processing received signal burst series (SLR), each including a training sequence in a radio communication receiver (RI), comprising a symbol detection in received bursts, a decoding of each series of bursts of received bursts. symbols detected in a decoded block including a cyclic redundancy code, an iterative construction of a reconstructed burst set (SLC) based on the decoded block and a prestored learning sequence, and an iterative estimate of an estimated channel according to the series of received bursts, the series of reconstructed bursts and the prestored learning sequence to improve the detection of symbols in reconstructed bursts based on the channel estimate, characterized in that it comprises steps of: for all the series of symbols received, prememinerizing (El) a predefined sequence (Seq) of weighting coefficients on (α) each associated at an iteration, one of the weighting coefficients being substantially equal to zero, and for each iteration, controlling (E6, E7) for the detection of the symbols a series weighting of reconstructed bursts (SLC) by the weighting associated with said iteration.

β - The method according to claim 5, wherein each weighting coefficient is applied to a sum of interference estimated from the series of reconstructed bursts (SLC ^) at each iteration.

7 - Process according to claim 5, wherein, when the weighting coefficient associated with an iteration is equal to zero, only the prestored learning sequence is considered in place of reconstructed bursts in the detection of symbols.

8 - Process according to any one of claims 5 to 7, further comprising the steps of: detecting (E2) at least one error for each decoded block, applying (E5) the decoded block (BD ^ (SLD _k )) the iterative construction of a series of reconstructed bursts as long as the block contains at least one error and as long as the iterations associated with all the weighting coefficients of the sequence for the series of bursts corresponding to the decoded block are not performed, and stop ( E3) control of the decoded block and memorize the decoded block when the block contains no error. 9 - Process according to any one of claims 5 to 8, wherein the weighting coefficients are between "0" and "1", and the weighting coefficients are distributed substantially regularly in the weighting sequence (Seq).

10 - Computer program for iteratively processing bursts of received symbol bursts

(SLR) each including a training sequence in a radio communication receiver (RI), performing steps of detecting symbols in received bursts, decoding each series of detected symbol bursts into a decoded block including a cyclic redundancy code, iteratively construct a series of reconstructed bursts (SLC) based on the decoded block and a prestored learning sequence, and iteratively determine a channel estimate based on the series of bursts received, the series of reconstructed bursts and the a prestored learning sequence for improving the detection of symbols in reconstructed bursts based on the channel estimate, said program having program instructions which when said program is loaded and executed in the receiver (RI), perform the steps of: for all the series of symbols received, to premade (El) a predefined sequence (Seq) of c weighting factors (α) each associated with an iteration (iter), one of the weighting coefficients being substantially equal to zero, and for each iteration, controlling (Eβ, E7) for the detection of the symbols a series weighting of reconstructed bursts (SLC) by the weighting coefficient associated with said iteration.

11 - Information medium comprising program instructions adapted to the implementation of a method for iteratively processing received signal burst series (SLR) each including a training sequence in a radio communication receiver (RI) according to one of claims 5 to 9, when said program is loaded and executed in the radio receiver.