Method of eliminating false echoes of a signal and corresponding rake receiver

[0001] This application is based upon and claims priority from prior European Patent Application No. 04 290 003.5, filed Jan. 5, 2004, the entire disclosure of which is herein incorporated by reference.

[0002] The present invention relates to signal processing, and more specifically to the processing of incident signals within a signal receiver.

[0003] There is known a receiver which combines several components of multi-path signals that are mutually delayed by different time delays before reaching the receiver. Such a receiver is for example present in code division multiple access (CDMA) wireless communication systems and is currently designated by the name of Rake-type receiver or “Rake” receiver.

[0004] In a wireless communication system, a base station communicates with a plurality of remote terminals, such as cellular mobile telephones. Frequency division multiple access (FDMA) and time division multiple access (TDMA) are the conventional multiple access systems for delivering simultaneous services to a certain number of terminals. The basic idea underlying the FDMA and TDMA systems is dividing the available resource into several frequencies or into several time slots, respectively, such that several terminals can operate simultaneously without causing interference.

[0005] Telephones operating according to the GSM standard belong to the FDMA and TDMA systems in the sense that transmission and reception take place at different frequencies and also in different time slots.

[0006] Unlike these systems using frequency division or time division, CDMA systems enable multiple users to share a common frequency and a common time channel by using a coded modulation. Amongst the CDMA systems are the CDMA 2000 system, the WCDMA system (wide band CDMA), and the IS-95 standard.

[0007] In CDMA systems, a scrambling code is associated with each base station and is used to distinguish one base station from another. In addition, an orthogonal code, known as the OVSF code, is allocated to each remote terminal (such as a cellular mobile telephone). All the OVSF codes are mutually orthogonal which distinguishes one channel from another.

[0008] Before transmitting a signal over the transmission channel to a remote terminal, the signal has been scrambled and spread by the base station using the scrambling code of the base station and the OVSF code of the channel.

[0009] In CDMA systems, those that use a distinct frequency for transmission and reception (CDMA-FDD system) can be distinguished from those which use a common frequency for transmission and reception, but distinct time domains for transmission and reception (CDMA-TDD system).

[0010] The present invention applies advantageously to communication systems of the CDMA type, and is particularly suited to systems of the WCDMA type with terrestrial radio access (UTRA FDD/TDD).

[0011] The incident signal received by a mobile telephone for example comprises different versions delayed in time from the signal initially transmitted, versions or echoes which are the result of the multi-path transmission characteristics of the transmission environment between a base station and the telephone, with each path introducing a different delay.

[0012] The “Rake” receiver in a cellular mobile telephone operating in a CDMA communication system is used to carry out temporal alignment, descrambling, compression, channel correction and combination of the delayed versions of the initial signals in order to deliver the information streams (symbols) contained in the initial signals.

[0013] A “Rake” receiver comprises several fingers, with each finger being intended to demodulate a given path received at a given instant.

[0014] Furthermore, the receiver comprises a channel estimation unit the purpose of which is to identify the various echoes through their delay and their mean energy as well as a mechanism for selecting echoes with a view to their respective assignment to the fingers of the Rake receiver.

[0015] Generally, the number of echoes detected is greater than the processing capacity of the receiver (number of fingers) for hardware complexity limitation reasons.

[0016] Hence, at the present time use is made of a selection mechanism capable of selecting from among the echoes detected the N best (N being the number of fingers of the receiver).

[0017] Generally, the selection of the echoes is based on the power profile measured by the receiver. More precisely, the receiver estimates the impulse response of the channel, based on the pilot signal transmitted by the base station. From this impulse response, it forms its energy, referred to as the “power profile”, and lists this estimation in a non-coherent manner over a given measurement duration.

[0018] The echoes which constitute the radio propagation profile between the base station and the mobile receiver are subsequently estimated with the help of this power profile.

[0019] Generally, the fingers are associated with the N maxima of the power profile, while rejecting any other maximum which might be close to a maximum retained by a duration of less than the duration of a chip.

[0020] It is recalled in this regard that in CDMA systems in particular, the symbols are transmitted within successive frames, with each frame being subdivided into a certain number of time slots. Each time slot conveys a certain number of symbols, with each symbol consisting of a predetermined number of chips.

[0021] This method of selecting echoes operates well in the case where the radio propagation profile exhibits echoes actually separated by more than a duration of a chip. However, in the case where close echoes are present, there is an irremediable loss due to the exclusion of such echoes.

[0022] As a consequence, to increase the performance of the receiver, it is necessary to search for the maxima in the power profile without applying the rejection principle described above.

[0023] However, there is the problem of selecting false echoes, that is to say the echoes which, on account of oversampling, are nothing other than replicas of the shaping filter and therefore afford no additional information.

[0024] If a false echo such as this is actually selected and assigned to a finger of the receiver, it then gives rise to needless energy consumption and worse still, due to the limitation of the hardware complexity of the receiver, the exclusion of the use of other true echoes that are less powerful than the false echo selected, but which afford more information.

[0025] This ultimately gives rise to a lowering of the performance of the receiver.

[0026] In view of these drawbacks, it is an object of the present invention to overcome the above-mentioned drawbacks and to improve the performance of a receiver of the “Rake” type for dense radio propagation profiles, such as for new-generation mobile telephones intended for receiving high-speed data packets (HSDPA).

[0027] In the case of HSDPA systems, it is possible to encounter extremely close echoes, that is to say they are temporally spaced apart by a duration less than the duration of a chip. Another object of the present invention is to introduce a rejection criterion so as to detect these false echoes.

[0028] Yet another object of the present invention is to provide a receiver of the “Rake” type that includes circuitry for detecting echoes of the received signal.

[0029] A further object of the present invention is to provide a wireless communication system component, such as a cellular mobile telephone, that incorporates such a receiver.

[0030] One embodiment of the present invention provides a method of processing a signal emanating from a multi-path transmission medium. According to the method, echoes of the signal that is received are detected, and a rejection processing is performed. The rejection processing includes estimating cross-correlation between temporally neighboring echoes, and if the cross-correlation satisfies a test, rejecting the one of the two neighboring echoes that exhibits lower energy.

[0031] Another embodiment of the present invention provides a Rake receiver that includes an input for receiving a signal emanating from a multi-path transmission medium, a detector, an estimator, and a rejector. The detector detects echoes of the signal that is received, and the estimator estimates cross-correlation between temporally neighboring echoes. If the cross-correlation satisfies a test, the rejector rejects the one of the two neighboring echoes that exhibits the lower energy.

[0032] Other objects, features, and advantages of the present invention will become apparent from the following detailed description. It should be understood, however, that the detailed description and specific examples, while indicating preferred embodiments of the present invention, are given by way of illustration only and various modifications may naturally be performed without deviating from the present invention.

[0033] FIG. 1 illustrates a cellular mobile telephone incorporating a “Rake” receiver in accordance with a preferred embodiment of the present invention,

[0034] FIGS. 2 to 4 illustrate the functionality and internal architecture of a “Rake” receiver according to one embodiment the present invention,

[0035] FIG. 5 illustrates in greater detail the processing circuitry of a Rake receiver according to a first embodiment of the present invention,

[0036] FIGS. 6 and 7 illustrate a method according to one embodiment of the present invention,

[0037] FIG. 8 illustrates the processing circuitry of a Rake receiver according to another embodiment of the present invention,

[0038] FIG. 9 illustrates a method according to another embodiment of the present invention, and

[0039] FIG. 10 illustrates a method according to yet another embodiment of the present invention.

[0040] Preferred embodiments of the present invention will be described in detail hereinbelow with reference to the attached drawings.

[0041] One preferred embodiment of the present invention provides a method of processing a signal emanating from a multi-path transmission medium. According to the method, there are detected echoes of the signal received. Additionally, in a rejection processing, the cross-correlation between temporally neighboring echoes is estimated, and if this cross-correlation satisfies a predetermined test, there is rejected that of the two neighboring echoes exhibiting the lower energy.

[0042] In one embodiment, when the signal received conveys symbols that each consist of a predetermined number of chips, two echoes are declared to be neighboring when they are spaced apart temporally by a duration less than the duration of a chip.

[0043] In some embodiments, when the signal received conveys symbols within successive frames that are each subdivided into a certain number of time slots, the estimation of the cross-correlation is performed over a chosen number of time slots, for example on the order of a hundred for a CDMA system.

[0044] Preferably, the estimation of the cross-correlation between two neighboring echoes comprises the calculation of a normalized cross-correlation coefficient obtained with the help of the complex amplitudes and the energies of the two echoes. For example, the predetermined test can comprise the comparison of the normalized cross-correlation coefficient with a predetermined threshold. This threshold is chosen in such a way that the probability of rejecting the false fingers is as low as possible. In one exemplary embodiment, when taking the modulus of the correlation between two potential candidates, the absolute value of the difference between the normalized cross-correlation coefficient and 1 is compared with 0.3. If this absolute value is less than 0.3, then that of the two neighboring echoes exhibiting the lower energy is rejected.

[0045] Generally, the number of echoes detected is greater than the number of fingers of a Rake receiver. In some embodiments, when such is the case, the number of residual echoes after rejection corresponds to the number of fingers of the Rake receiver, and these residual echoes are allocated to the fingers of the Rake receiver.

[0046] In other embodiment, while the number of echoes detected of the signal is equal to L0, which is greater than the number N of fingers of the Rake receiver, the number of residual echoes arising from the rejection processing is equal to L, and L is always greater than N. In such embodiments, N echoes which will be allocated to the fingers of the receiver are selected from among the L residual echoes by any conventional process, for example using a criterion based on the signal-to-noise ratio. Such a criterion provides a good solution as long as the minimum separation between the various residual echoes of the radio propagation profile remains relatively sizeable. However, to increase the performance of the receiver, it is preferred to take into account the residual neighboring echoes which are therefore true echoes, but which are temporally correlated.

[0047] In one exemplary embodiment of the present invention in which the number of echoes detected of the signal received is equal to L0, which is greater than the number N (number of fingers of the Rake receiver), and in which the number of residual echoes arising from the rejection processing is equal to L, which is greater than N, at least some of the L0 residual echoes are echoes temporally correlated according to a predetermined correlation criterion, and the selection step comprises determining with the help of the initial complex amplitudes associated with at least the temporally correlated echoes weighted individual energies corresponding respectively to temporally decorrelated virtual echoes, and selecting from among the L echoes N echoes as a function of a predetermined power criterion taking into account at least the weighted individual energies.

[0048] Stated otherwise, the new representation with the aid of the weighted individual energies provides an equivalent representation but with independent fadings and a new distribution of power over these echoes. The fact that the new virtual echoes are independent makes it possible to apply a criterion based on powers but now using modified powers. Also, it can be shown mathematically that this new representation affects neither the noise nor the energy of the whole of the process represented by the L echoes.

[0049] According to one embodiment of the present invention, when the signal received conveys symbols that each consist of a predetermined number of chips, two echoes are declared to be temporally correlated when they are spaced apart temporally by a duration less than or equal to the duration of a chip.

[0050] Several variant implementations of the present invention may be envisaged. According to a first variant, weighted individual energies may be determined with the help of the initial complex amplitudes of the whole set of the L echoes. Thus, for example, in the selection step all possible combinations of N echoes from among the L residual echoes are defined, and for each of the combinations, a set of N weighted individual energies is calculated with the help of the initial complex amplitudes of the N echoes of the combination, as is the sum of these N weighted individual energies, and the combination of N echoes whose said associated sum is the largest is selected.

[0051] The calculation of the set of N weighted individual energies which is associated with a combination of N echoes, comprises, for example, the determination of the covariance matrix of the initial complex amplitudes of the N echoes, and the determination of the N eigenvalues of the covariance matrix.

[0052] The implementation of such a variant when the number N is relatively large, for example greater than four, becomes complex by reason of the large size of the covariance matrix that has to be first estimated on the basis of measurement on the echoes and subsequently to be diagonalized. This complexity is relatively large for standard CDMA receivers.

[0053] Thus, provision is made for another variant implementation of the present invention, which is much more suitable for more concrete propagation profiles. According to this variant, the echoes can be grouped together into a family of dense echoes (that is to say temporally correlated) and a family of sparse echoes (that is to say distant).

[0054] Stated otherwise, according to this variant of the present invention, assuming that the L residual echoes comprise L1 temporally correlated echoes and L-L1 possible remaining “far” echoes, with L1 being greater than or equal to 2 and less than or equal to L, the selection step comprises a first selection according to a first selection criterion of L2 echoes from among the L1 echoes, with L2 being less than or equal to L1.

[0055] For example, in the case where L1 is an odd number, at least one of the L1 temporally correlated echoes will be eliminated, in such a way as to obtain an even number L2. If on the other hand L1 is already even, then it will be possible to choose L2=L1 and to perform no first selection.

[0056] Provision is moreover made for a determination with the help of the initial complex amplitudes of the L2 echoes of the weighted individual energies corresponding respectively to temporally decorrelated virtual echoes, as well as an assignment of these weighted individual energies to the L2 echoes selected.

[0057] Next, the N echoes are selected from among the L2 echoes and the L-L1 far echoes as a function of a predetermined power criterion taking into account the L2 weighted individual energies and the L-L1 initial individual energies of the L-L1 possible remaining echoes.

[0058] When the number L2 is even, the L2 echoes are advantageously processed in pairs. Thus, for example, for each pair of echoes a pair of weighted individual energies is calculated with the help of the initial complex amplitudes of the pair of echoes, and the N echoes having the N largest energy values from among the L2 weighted individual energies and the L-L1 initial individual energies of the possible L-L1 far echoes.

[0059] The calculation of a pair of weighted individual energies which is associated with a pair of echoes, comprises, for example, the determination of the covariance matrix of the initial complex amplitudes of the echoes, and the determination of the two eigenvalues of the covariance matrix.

[0060] Another preferred embodiment of the present invention provides a receiver of the “Rake” type that includes an input for receiving a signal emanating from a multi-path transmission medium, and means for detecting echoes of the signal received. The receiver further includes processing means comprising estimation means for performing an estimation of the cross-correlation between temporally neighboring echoes, and rejection means for, if the cross-correlation satisfies a predetermined test, rejecting that of the two neighboring echoes exhibiting the lower energy.

[0061] According to one embodiment of the present invention in which the signal received conveys symbols within successive frames that each are subdivided into a certain number of time slots, the estimation means performs the estimation of the cross-correlation over a chosen number of time slots, for example on the order of a hundred.

[0062] In some embodiments, the estimation means comprises means for calculating a normalized cross-correlation coefficient obtained with the help of the complex amplitudes and the energies of the two echoes. Preferably, the rejection means comprises comparison means for comparing the normalized cross-correlation coefficient with a predetermined threshold.

[0063] In one embodiment, when the number of residual echoes after rejection corresponds to the number of fingers of the Rake receiver, the processing means comprises a unit for allocating these residual echoes to the fingers of the Rake receiver.

[0064] When the detection means detect a number L0 of echoes of the signal received, with L0 being greater than N, and when the rejection means deliver L residual echoes, with L being likewise greater than N, at least some of the L echoes being echoes temporally correlated according to a predetermined correlation criterion, the processing means comprises calculation means for determining with the help of the initial complex amplitudes associated with at least the temporally correlated echoes weighted individual energies corresponding respectively to temporally decorrelated virtual echoes, and selection means for selecting N echoes from among the L echoes as a function of a predetermined power criterion taking into account at least the weighted individual energies.

[0065] According to one embodiment of the present invention, the determination means is able to determine weighted individual energies with the help of the initial complex amplitudes of the whole set of the L echoes.

[0066] According to one embodiment of the present invention, the calculation means is able to define all the possible combinations of N echoes from among the L residual echoes, and for each of the combinations, to calculate a set of N weighted individual energies with the help of the initial complex amplitudes of the N echoes of the combination, and also the sum of these N weighted individual energies, and the selection means selects the combination of N echoes whose said associated sum is the largest.

[0067] According to one embodiment of the present invention, the calculation means is able, for the calculation of the set of N weighted individual energies which is associated with a combination of N echoes, to determine the covariance matrix of the initial complex amplitudes of the N echoes, and to determine the N eigenvalues of the covariance matrix.

[0068] According to another variant of the present invention, the L residual echoes comprise L1 temporally correlated echoes, and L-L1 possible remaining far echoes, with L1 being greater than or equal to 2 and less than or equal to L. The processing means comprises auxiliary selection means for performing a first selection according to a first selection criterion of L2 echoes from among the L1 echoes, with L2 being less than or equal to L1.

[0069] The calculation means is able to determine, with the help of the initial complex amplitudes of the L2 echoes, weighted individual energies corresponding to temporally decorrelated virtual echoes, and to assign these weighted individual energies to the L2 selected echoes and selection means are able to select the N echoes from among the L2 echoes and the L-L1 far echoes, as a function of a predetermined power criterion taking into account the weighted individual energies and the L-L1 individual energies of the L-L1 remaining echoes.

[0070] In the case where L1 is an odd number the auxiliary selection means is able to eliminate at least one of the L1 temporally correlated echoes, in such a way as to obtain an even number L2.

[0071] According to one embodiment of the present invention, the calculation means processes the L2 echoes pairwise. For each pair of echoes, the calculation means calculates a pair of weighted individual energies with the help of the initial complex amplitudes of the pair of echoes, and the selection means selects the N echoes having the N largest energy values from among the L2 weighted individual energies and the L-L1 initial individual energies of the possible L-L1 far echoes.

[0072] According to one embodiment of the present invention, for the calculation of a pair of weighted individual energies which is associated with a pair of echoes, the calculation means determines the covariance matrix of the initial complex amplitudes of the echoes, and determines the 2 eigenvalues of the covariance matrix.

[0073] Another embodiment of the present invention provides a component of a wireless communication system that incorporates a receiver as defined hereinabove. In one exemplary embodiment, this component is a cellular mobile telephone.

[0074] Exemplary embodiments of the present invention will now be described in detail with reference to the figures.

[0075] FIG. 1 shows a cellular mobile telephone incorporating a “Rake” receiver in accordance with a preferred embodiment of the present invention. The cellular mobile telephone TP comprises an analog radio frequency stage ERF connected to an antenna ANT to receive an input signal ISG.

[0076] The ERF stage comprises a low noise amplifier LNA and two processing channels comprising mixers, filters and conventional amplifiers. The two mixers receive from a phase locked loop PLL two signals mutually presenting a phase difference of 90°. After frequency transposition in the mixers, the two processing channels respectively define two streams I (direct stream) and Q (quadrature phase stream) according to the well known denomination.

[0077] After digital conversion in analog/digital converters, the two streams I and Q are delivered to a reception processing stage ETNR.

[0078] The reception processing stage ETNR comprises a “Rake receiver” RR, followed by a demodulator MP which demodulates the constellation delivered by the Rake receiver RR. The demodulator MP is followed by a channel decoder CD.

[0079] Because of possible reflections of the initially transmitted signal on obstacles situated between the base station and the mobile telephone, the transmission environment is in fact a multi-path transmission environment MPC having several different transmission paths (three exemplary transmission paths P1, P2, and P3 are shown in FIG. 1). In consequence, the signal ISG which is received by the mobile telephone comprises different versions delayed in time from the signal initially transmitted, versions or echoes which are the result of the characteristics of multi-path transmission of the transmission environment. And, each path introduces a different delay. Naturally, the received signal ISG could also result from the transmission of initial signals transmitted by different base stations BS1 and BS2.

[0080] FIG. 2 illustrates the basic functionalities of the “Rake” receiver according to one embodiment the present invention. The “Rake” receiver RR is formed of several fingers (here N fingers) FG1-FGN. Each finger is designed to demodulate a given path received at a given moment. Baseband demodulation essentially consists in a descrambling and a despreading. The despreading is in reality a correlation and so requires an integration over the symbol period. The receiver then combines the information received on each finger in a combination circuit MCMB, by summing them after having corrected the phase and amplitude distortions of each path (unit CHU described below). Naturally, the multiple fingers represented in FIG. 2 may be formed of the same physical finger, reconfigured N times to produce the N functional fingers.

[0081] The receiver also comprises a channel estimation unit CHU capable of supplying a transmission channel estimate, with the aim of correcting the channel distortions.

[0082] It is necessary to allocate N paths (echoes) to the digits of the receiver. This is done by a control unit RMU, after selection of N echoes.

[0083] Finally, since the timing relating to the paths may vary over time, a tracking unit monitors the timing of the paths and updates the fingers with that information.

[0084] As shown in FIG. 3, each finger FGi comprises a code generator CG capable of producing the scrambling codes and the OVSF codes, a sub-sampler, in this instance a sub-sampler by four, disposed at the input of the finger, and a channel estimation unit CHU.

[0085] Furthermore, the finger FGi comprises several physical transmission channel demodulators, in this case four demodulators DEM1-DEM5. Each channel demodulator performs the functions of descrambling, despreading and integration, as well as the previously mentioned channel correction functions. The channel estimation unit CHU is shared by all the physical transmission channels.

[0086] FIG. 4 partially illustrates the internal architecture of a “Rake” receiver RR according to this embodiment. As shown, there are between the signal input ES and the output OP which delivers the information relating to the different physical transmission channels (in this case five physical channels), the fingers of the receiver, in this case six fingers FG1-FG6, connected at the output to combination circuit MCMB.

[0087] FIG. 5 illustrates in greater detail the processing circuitry of a Rake receiver according to a first embodiment of the present invention. As shown, the channel estimation unit CHU has detected for example L0 echoes each characterized by its delay t, its complex amplitude (amplitude and phase) s and its “initial” individual energy, which is equal to the product of s and of its complex conjugate (s.s*).

[0088] These L0 echoes have been detected in a conventional manner, subsequent to an estimation of the impulse response of the channel, using for example the method of local maxima.

[0089] A processor MTS, for example embodied in software form, then determines from among the L0 echoes detected, possible false echoes that are nothing other than replicas of the shaping filter and which therefore do not afford any additional information.

[0090] In this regard, an estimator MST performs a processing for estimating the cross-correlation between temporally neighboring echoes. Moreover, if this cross-correlation satisfies a predetermined test, a rejector MRJ rejects that of the two neighboring echoes exhibiting the lower energy.

[0091] In CDMA applications in particular, the information transmitted between a base station and a cellular mobile telephone is conveyed within successive frames FRi, as shown in FIG. 6. Each of these frames is divided into a certain number of time slots SLi (16 slots per frame for example).

[0092] Moreover, each symbol conveyed consists of a predetermined number of chips. By way of indication, in a CDMA system, the duration of a chip may be equal to 260.417 nanoseconds.

[0093] Although it is possible to adopt several criteria to define what is meant by “two neighboring echoes”, a particularly simple way is by using the duration of a chip. Thus, two echoes will be declared to be neighboring when they are temporally spaced apart by a duration less than the duration of a chip.

[0094] Let us now consider two neighboring echoes F1 and F2 which have been detected at the output of the channel estimation unit CHU, once the channel estimation has stabilized.

[0095] During slot n (step 700 of FIG. 7), an elementary cross-correlation coefficient r(n) equal to the modulus of the product of the complex amplitude of one of the echoes times the complex conjugate of the amplitude of the other echo is calculated for the two echoes F1 and F2.

[0096] Next, a smoothing of the cross-correlation from one slot to another is performed (step 701). In some embodiments, this smoothing consists of a simple average or an exponential smoothing. For example, if an exponential smoothing is used, then a smoothed cross-correlation coefficient R(n) equal to αR(n)+(1-α).r(n) is defined, where α is the exponential smoothing coefficient equal for example to 0.98.

[0097] After a predetermined duration, that is say in the present case after a certain number NIT of slots (step 702), a cross-correlation test is carried out (step 703). By way of indication, in a UMTS application, the number NIT is on the order of a hundred.

[0098] The cross-correlation test comprises the comparison of a normalized cross-correlation coefficient with a predetermined threshold. This normalized cross-correlation coefficient is obtained by dividing the coefficient R(n) by the square root of the product of the smoothed energy associated with the echo F1 and of the smoothed energy associated with the echo F2.

[0099] More precisely, the smoothed energy E(n,1) associated with the echo F1 is equal for example to the average of the energies of the echo F1 over the whole set of the NIT slots considered. The same holds for the smoothed energy E(n, 2) associated with the echo F2.

[0100] The absolute value of the difference between this normalized coefficient and 1 is then compared with the threshold d, which is for example equal to 0.3.

[0101] If the cross-correlation test is satisfied, the weakest of the echoes, that is to say the one which exhibits the lowest smoothed energy, is rejected from the list of candidate echoes (step 705).

[0102] If, on the other hand, the cross-correlation test is not satisfied, then this signifies that the two neighboring echoes are actually two true echoes. In this case, both of them are retained (step 704).

[0103] In the example just described, to calculate the normalized cross-correlation coefficient, the modulus is taken between two potential candidates and it is smoothed before being normalized. However, it would be entirely possible to smooth the complex values before taking the modulus thereof or else take the modulus squared thereof, smooth the energy and normalize the energy and not the square root. Stated otherwise, all these formulations are equivalent and make it possible to define a normalized cross-correlation coefficient. Also, according to the formulation chosen, the threshold will be different. Thus, in the case where the energy rather than the modulus is used, the threshold value would for example be equal to 0.1 instead of 0.3.

[0104] Moreover, in the example just described and illustrated in FIG. 5 in particular, it was assumed that after rejection processing, there are still N residual echoes. Since this number of residual echoes is equal to the number of fingers of the Rake receiver, the unit RMU will then allocate these N residual echoes to the N fingers of the Rake receiver.

[0105] That said, in other cases, as illustrated for example in FIG. 8, the number of residual echoes L may. remain greater than the number of fingers N of the Rake receiver.

[0106] In this case, the processor MTS then selects from among these L residual echoes the N echoes which will subsequently be assigned to the fingers of the Rake receiver by the unit RMU.

[0107] Various software blocks are then provided, namely an auxiliary selector MSA, a calculator MC and a selector MS, the functions of which are described in greater detail hereinbelow.

[0108] Among these L residual echoes, some of them, possibly all, may still be echoes that are close together or neighboring, that is to say temporally correlated echoes.

[0109] Here again, a particularly simple but non-limiting way of defining what is meant by “two temporally correlated echoes” is by using the duration of a chip. Two echoes will therefore be declared in this embodiment to be temporally correlated when they are temporally spaced apart by a duration less than the duration of a chip.

[0110] The fadings which affect the L echoes that are generally considered to be complex Gaussian are not independent and it is possible to associate their covariance matrix K with them. This matrix of dimension L×L is represented by its elements A(i,j)=E[s(i)s(j)*], where E designates the “mathematical expectation” operator, and * designates the complex conjugate.

[0111] When the echoes are relatively separated, that is to say not temporally correlated, with for example a minimum separation greater than the duration of a chip, then this matrix is practically diagonal and its diagonal elements consist of the initial individual average energies of the echoes. Also, a representation is obtained solely with the aid of these initial individual energies.

[0112] However, for dense echoes, the covariance matrix can no longer be regarded as a diagonal matrix. Hence, the present invention provides for a new representation of the L echoes, this new representation, equivalent to the previous one, implementing temporally decorrelated virtual echoes associated with independent fadings, as well as a new distribution of power over these echoes.

[0113] The fact that the new echoes (virtual echoes) are independent then makes it possible to apply a power criterion to the weighted individual energies of these new echoes rather than to the initial individual energies.

[0114] More precisely, according to a first variant of the method according to the present invention illustrated in FIG. 9, the whole set of L echoes is taken into account, be they temporally correlated or otherwise, to calculate weighted individual energies corresponding to temporally decorrelated virtual echoes.

[0115] That said, in practice, simply calculating the eigenvalues causes the link between the echo and the eigenvalue associated therewith to be lost and hence it then becomes difficult to decide which echoes to preserve.

[0116] The variant implementation illustrated in FIG. 9 affords a solution to this problem. In this regard, all the possible combinations of N echoes out of L are defined (step 61) SO as to obtain C_L^N combinations CMB_j.

[0117] Next, for each of these combinations CMBJ, the associated covariance matrix K_j is calculated (step 63).

[0118] The N eigenvalues λ_j associated with the N echoes of the combination considered are subsequently determined in a conventional manner (step 64). These eigenvalues constitute the weighted individual energies.

[0119] Next, the sum of these weighted individual energies is calculated so as to obtain a sum S_j associated with the combination considered of N echoes (step 65).

[0120] When all the combinations have been processed, the maximum sum is determined (step 67) and the N echoes of the combination associated with this maximum sum will then be selected.

[0121] Operationally, in this embodiment the calculator MC incorporated into the processor MTS which performs steps 63 to 65, whereas step 67 is performed by the selector MS.

[0122] That said, the implementation of such a variant becomes relatively complicated when the number of fingers is relatively large, for example greater than four. Such is the case for standard CDMA receivers.

[0123] Hence, the present invention proposes another variant implementation which makes it possible to approximate the complete calculation for the most concrete cases of propagation profiles. Such a variant implementation is illustrated in FIG. 10.

[0124] In a general manner, the echoes are grouped into a family of dense echoes (temporally correlated) and (possible) remaining echoes, that is to say distant echoes. Next, the procedure for calculating the weighted individual energies is applied to the family of dense echoes and the remainder of the echoes will be kept intact.

[0125] More precisely, if L echoes (6 in FIG. 10) and three fingers (N=3) are considered, it is seen that the L echoes comprise five dense echoes ECH1-ECH5 formed with two families ECH1-ECH3 and ECH4-ECH5, as well as a distant echo ECH6.

[0126] The auxiliary selector MSA of the processor MTS will then advantageously eliminate a dense echo from among the five echoes ECH1-ECH5. Also, more precisely, since pairs of echoes will be worked on subsequently, the auxiliary selector MSA will delete an echo from among the family of the first three echoes ECH1-ECH3. This deletion may be performed according to a power criterion. It will for example be possible to eliminate the echo having the lowest initial individual energy.

[0127] After the preselection (step 70), it is then assumed that L2 temporally correlated echoes are obtained, formed of two pairs of echoes ECH1-ECH2 on the one hand and ECH4-ECH5 on the other hand, as well as a remaining echo ECH6.

[0128] The calculation means then determines for the pair of echoes ECH1-ECH2 the corresponding covariance matrix which is now a 2×2 matrix. The calculation of the eigenvalues (step 72) then turns out to be particularly simple and provides the two eigenvalues λ1 and λ2 which correspond to the weighted individual energies of the new virtual echoes. Also, since the covariance matrix is a 2×2 matrix, the link between the echo and its eigenvalue is not lost.

[0129] The calculator performs the same operations (step 73 and step 74) for the pair of echoes ECH4 and ECH5 so as to obtain two eigenvalues λ4 and λ5. Moreover, the value λ6 corresponds to the initial individual energy of the echo ECH6.

[0130] Next, since N is here equal to 3, the selector will choose three eigenvalues from among the five eigenvalues λ1, λ2, λ4, λ5 and λ6.

[0131] Also, here, the selection criterion consists simply in choosing the three largest eigenvalues and consequently in selecting the three corresponding echoes, that in the present case are assumed to be the echoes ECH1, ECH4 and ECH6.

[0132] While there has been illustrated and described what are presently considered to be the preferred embodiments of the present invention, it will be understood by those skilled in the art that various other modifications may be made, and equivalents may be substituted, without departing from the true scope of the present invention. Additionally, many modifications may be made to adapt a particular situation to the teachings of the present invention without departing from the central inventive concept described herein. Furthermore, an embodiment of the present invention may not include all of the features described above. Therefore, it is intended that the present invention not be limited to the particular embodiments disclosed, but that the invention include all embodiments falling within the scope of the appended claims.