CN1925345A - Multi-user detection using an adaptive combination of joint detection and successive interference cancellation - Google Patents

Abstract

A time division duplex communication system using code division multiple access transmits a plurality of data signals over a shared spectrum in a time slot. A combined signal is received over the shared spectrum in the time slot. The plurality of data signals are grouped into a plurality of groups. The combined signal is matched filtered based on symbol responses associated with the data signals of one of the groups. Data from each data signal in this one group is jointly detected. An interference signal is constructed based on the detected data, of this one group. The constructed interference signal is subtracted from the combined signal. Data from the other groups is detected by processing the subtracted signal.

Description

Adopt the self adaptation combination of joint-detection and serial interference elimination to carry out Multiuser Detection

The application is that application number is 01809452.X, and the applying date is February 21 calendar year 2001, and denomination of invention is divided an application for the Chinese patent application of ' adopt the self adaptation of joint-detection and serial interference elimination to make up and carry out Multiuser Detection '.

Background of invention

The present invention relates generally to wireless communication system.Particularly, the present invention relates to the joint-detection of multiple user signals in the wireless communication system.

Fig. 1 is the diagram of wireless communication system 10.This communication system 10 has and subscriber equipment (UE) 14 ₁To 14 ₃The base station 12 that communicates ₁To 12 ₅Each base station 12 ₁All have a corresponding work district, in this service area, base station and the interior subscriber equipment 14 in district ₁To 14 ₃Communicate.

In some communication system, for example in the tdd systems (TDD/CDMA) of code division multiple access (CDMA) and employing code division multiple access, multichannel communication uses same frequency spectrum to send, and is generally distinguished by its signal patch code sequence between each communication.For more effectively utilizing frequency spectrum, the TDD/CDMA communication system adopts the repeating frame that is divided into some time slots to communicate.According to the different bandwidth of communication, the communication that sends in this type systematic all has one or more relevant signal patch code and time slots.

Because this type systematic can use same frequency spectrum to send multichannel communication simultaneously, so the receiver in this system must be distinguished each road communication.A kind of method that detects such signal is matched filtering.Matched filtering can detect the communication of using a single code to send, and other communication is handled as disturbing.Therefore for detecting a plurality of codes, must use a plurality of matched filters of respective numbers.Another kind method is serial interference elimination (Successive InterferenceCancellation SIC).This method detects one tunnel communication, deducts the signal base value of this road communication thereafter from received signal, to be used to detect next road communication.

In some application scenario, for improving communication performance, requirement can detect multichannel communication simultaneously.Detect multichannel communication simultaneously and be referred to as joint-detection.Some combined detector adopts the Cholesky decomposition method to carry out least mean-square error (MMSE) and detects and use ZF block equalizers (ZF-BLE).These detector complexity height need take receiver resources widely.

United States Patent (USP) NO.5,933,423 have disclosed a kind of receiver.This receiver is divided into some groups with the signal that receives, and every group comprises at least two signals, and the signal in each group is detected simultaneously.Before detecting the data of other groups, from received signal, subtract and first group of detection signal.

DE19616828 A1 has disclosed a kind of receiver that has a demultiplexer.Signal is organized into groups according to received power.First group of signal carried out joint-detection.The first group of signal that detects subtracts from received signal.Signal residual signal is detected according to subtracting.

Tsatsanis and paper that Xu shows " adaptive blind interference eliminated in the cdma system " (Adaptive Blind Interference Cancellation in CDMA System) have disclosed the method that adopts lowest mean square and least mean-square error instrument cancellation receiver to disturb.

Therefore, need to seek alternative method and carry out Multiuser Detection.

Summary of the invention

A kind of employing code division multiple access mode, the shared frequency spectrum of a time slot of use sends the tdd communication systems of a plurality of data-signals.Shared frequency spectrum by this time slot can receive a composite signal.A plurality of data-signals are divided into a plurality of data-signal cohorts.This composite signal is carried out matched filtering according to the part symbol response relevant with the data-signal of one of cohort.The data of each data-signal detect simultaneously in same group.Interference signal detects the data generation according to the part of this cohort.This interference signal is deducted from composite signal.Subtract signal and can the data of other group be detected by handling this.

Description of drawings

Fig. 1 is a wireless communication system.

Fig. 2 is the simplification transmitter and the receiver that adopt the joint-detection mode

Fig. 3 is the diagram of a communication burst.

Fig. 4 is the flow chart of a joint-detection and the combination of serial interference elimination self adaptation

Fig. 5 is the diagram of a joint-detection and serial interference elimination self adaptation composite set

Fig. 6-12 is the performance comparison diagram of joint-detection and the combination of serial interference elimination self adaptation, complete joint-detection and RAKE receiver.

Embodiment

The joint-detection (JD) and the simplification transmitter 26 of serial interference elimination (SIC) self adaptation combination " SIC-JD " and the diagram of receiver 28 of Fig. 2 for using in the TDD/CDMA communication system.In a canonical system, transmitter 26 is positioned at each UE14 ₁To 14 ₃In, and a plurality of radiating circuits 26 that are used to send multichannel communication are positioned at each base station 12 ₁To 12 ₅In.Base station 12 ₁Require each efficient communication UE14 ₁To 14 ₃All has a radiating circuit 26 at least.SIC-JD receiver 28 can be positioned at base station 12 ₁In, UE 14 ₁To 14 ₃In, or be arranged in both simultaneously.The communication that SIC-JD receiver 28 receives from a plurality of transmitters 26 or radiating circuit 26.

Each transmitter 26 all sends data by a radio channel 30.The data that data generator 32 in the transmitter 26 is generated transfer to receiver 28 by a reference channel.According to the communication bandwidth requirement, reference data is assigned to one or more codes and/or time slot.Modulation is expanded with 34 pairs of reference datas of expanding unit, and uses the training sequence in assigned timeslot or the code, and the reference data after the expansion is become time multiplexing date.The sequence that produces is referred to as a communication burst.This communication burst is modulated to radio frequency by modulator 36.Antenna 38 is by the antenna 40 radiated radio frequency (RF) signals of radio channel 30 to receiver 28.The types of modulation that is used for this type of emission communication can be known any one pattern of person of ordinary skill in the field, for example direct phase-shift keying (DPSK) or quaternary PSK (QPSK).

As shown in Figure 3, a representative communication pulse train 16 has 20, one guard periods 18 of a training sequence and two data pulse trains 22,24.20 pairs of data pulse trains of training sequence 22,24 separate, and 18 pairs of communication burst of guard period are separated, and the pulse train that is intended to different transmitters is sent arrived in the different time.Two data pulse trains 22,24 comprise the data of communication burst, and have identical symbol lengths in the ordinary course of things.This middle part training sequence (midamble) comprises a training sequence.

The antenna 40 of receiver 28 receives various radiofrequency signals.The signal that receives generates a baseband signal after demodulator 42 demodulation.Baseband signal by for example channel estimating apparatus 44 and SIC-JD device 46 in time slot and use assignment to handle for the responding communication pulse train code of corresponding transmitter 26.Channel estimating apparatus 44 uses the training sequence component in the baseband signal that channel information is provided, for example channel impulse response.SIC-JD device 46 utilizes channel information that the emission data of the communication burst that receives are estimated as hard symbol then.

The known extensions code that SIC-JD device 46 utilizes channel information that channel estimating apparatus 44 provided and transmitter 26 to use is estimated the various communication pulse string datas that receive.Describe although herein SIC-JD device 46 and TDD/CDMA communication system are combined, this method also is applicable to other communication system, for example CDMA.

Fig. 4 has carried out graphic extension to a kind of method of carrying out SIC-JD in certain particular time-slot of TDD/CDMA communication system.In this particular time-slot, some communication burst mutual superposition, for example K communication burst.K pulse train can be from K different transmitter.Yet when some transmitter used a plurality of code in this particular time-slot, K pulse train can be from being less than K transmitter.

Two data pulse trains 22,24 in the communication burst 16 all have the emission symbol of predetermined quantity, for example N _sEach symbol all adopts the spreading code signal patch emission of predetermined number, and spreading code is flare factor (SF).In typical TDD communication system, each base station (12 ₁To 12 ₅) communication data in all comprise a relevant scrambling code, this scrambling code can make between each base station difference mutually.Generally speaking, the scrambling code can not influence flare factor.Although still use term " spreading code " and " spreading coefficient " for the system that uses the scrambling code hereinafter, for following situation, " spreading code " will refer to the combination of scrambling code and spreading code.Therefore, data pulse string 22,24 all has N _s* SF signal patch.After through one the channel of W signal patch impulse response being arranged, the length of the pulse train that each receives is SF * N _s+ W-1, this numerical value also can the Nc signal patch represent.K in this K pulse train ^ThThe code of pulse train is with C ^(k)Expression.

Each K ^ThPulse train is received by receiver and 1 expression of available following formula:

r ^(k)=A ^(k) d ^(k), k=1 ... K formula 1

r ^(k)Be this K that receives ^ThThe pulse train base value, A ^(k)Be aggregate channel response (N _c* N _sMatrix).A ^(k)J in the matrix ^ThClassify d as ^(k)J ^ThThe symbol response s of element ^(k)The zero padding pattern.This symbol response s ^(k)Estimation response for this pulse train h ^(k)With spreading code C ^(k)Convolution. d ^(k)Be the unknown data symbol that sends in the pulse train.Each K ^ThThe estimation response of pulse train h ^(k)Length be W signal patch, available following formula 2 expressions:

${\underset{\‾}{h}}^{\left(k\right)}={\mathrm{\γ}}^{\left(k\right)}\·{\underset{\‾}{\overset{~}{h}}}^{\left(k\right)}$ Formula 2

γ wherein ^(k)Gain of expression transmitter and/or path loss;

The distinctive fading channel response of indicating impulse string; And for a pulse train cohort of similar channel, The distinctive channel response of expression cohort.For uplink communication, each pulse train h ^(k), γ ^(k)And Different; For Down-Link Communications, each pulse train

All identical, and γ ^(k)Different.And if in downgoing line, adopt transmission diversity system, the then γ of each pulse train ^(k)With

All different.

Can represent with formula 3 by all K pulse train sum vectors that radio channel receives:

$\underset{\‾}{r}=\underset{i=1}{\overset{k}{\mathrm{\Σ}}}{\underset{\‾}{r}}^{\left(k\right)}+\underset{\‾}{n}$ Formula 3

Wherein nRepresent a zero-average noise vector.

If A with all data pulse strings ^(k)Be merged into matrix A, all unknown data of each pulse train d ^(k)Be merged into matrix d, then formula 1 has promptly become formula 4.

r=A d+ nFormula 4

Each K ^ThThe received power of pulse train by SIC-JD according to the priori of receiver 28, measure from the estimation of pulse train channel or the matched filter banks of the specific training sequence of pulse train.This K pulse train is according to the received power size of its mensuration, by the sequence arrangement of successively decreasing.

The pulse train of power grade identical substantially (for example being in the same threshold) is grouped in together and is arranged in G the group, and 48.This G the watt level of organizing according to each group, by the sequence arrangement of successively decreasing, for example, order is for to begin to G from organizing 1, and this group 1 has the highest received power.Fig. 5 carries out the diagram of SIC-JD for SIC-JD device 46 according to G group.

For the group 1 of highest power received, pulse train symbol response matrix A in only should organizing _g ⁽¹⁾Determined that this matrix only comprises the symbol response of pulse train in the group 1.Organize 1 reception vector simultaneously rWith x _g ⁽¹⁾Expression.Therefore, for group 1, formula 4 becomes formula 5.

x _g ⁽¹⁾=A _g ⁽¹⁾ d _g ⁽¹⁾+ n Formula 5

Wherein d _g ⁽¹⁾For organizing the data in 1 pulse train.Formula 5 has embodied the effect of intersymbol interference (ISI) with multiple access interference (MAI).Therefore, the effect of other group (group 2 is to group G) is not considered.

Receive vector x _g ⁽¹⁾Through organizing 1 matched filter 66 ₁Become the symbol response of pulse train in the group 1 after the filtering, this process is with formula 6,50 expressions.

${\underset{\‾}{y}}_g^{\left(1\right)}=A_g^{{\left(1\right)}^H}{\underset{\‾}{x}}_g^{\left(1\right)}$ Formula 6

Wherein y _g ⁽¹⁾Be the matched filtering result.

The joint-detection device 68 of group 1 ₁Carry out joint-detection to organizing 1, utilize the matched filtering result y _g ⁽¹⁾Make the soft-decision estimation

A kind of method of joint-detection is to calculate least square, ZF result according to formula 7.

${\underset{\‾}{\overset{^}{d}}}_{g.\mathrm{soft}}^{\left(1\right)}={\left(A_g^{{\left(1\right)}^H}{A}_g^{\left(1\right)}\right)}^{-1}{\underset{\‾}{y}}_g^{\left(1\right)}$ Formula 7

A _g ^{(1) H}Be A _g ⁽¹⁾Hermitian form.Another method is to calculate the least mean-square error result according to formula 8.

${\underset{\‾}{\overset{^}{d}}}_{g.\mathrm{soft}}^{\left(1\right)}={(A_g^{{\left(1\right)}^H}{A}_g^{\left(1\right)}+{\mathrm{\σ}}^{2}I)}^{-1}{\underset{\‾}{y}}_g^{\left(1\right)}$ Formula 8

Wherein I is a unit matrix, σ ²Be standard deviation.

This method is only carried out joint-detection to set of pulses string, and its advantage is will be lower than the complexity that all signals are analyzed to the single group of complexity of analyzing.Because A _g ^{(1) H}With A _g ⁽¹⁾Be strip code character toeplitz matrix, so the complexity of formula 7 or 8 solution procedurees has also obtained reduction.In addition, the performance reduction that is caused when utilization Cholesky decomposes also can be ignored.It is quite complicated that a plurality of pulse trains is carried out the Cholesky decomposition, yet when less user's group was carried out the Cholesky decomposition, its complexity can significantly reduce.

Soft-decision-hard decision piece 70 ₁With soft-decision

Reception data as group 1 are converted to hard decision

54.When other low-power group is handled, organize the multiple of 1 pair of low-power group generation and advance to connect interference by a group 1 interference generation piece 72 ₁Estimate 56 according to formula 9.

${\underset{\‾}{\overset{^}{r}}}^{\left(1\right)}=\begin{array}{cc}{A}_g^{\left(1\right)}& {\underset{\‾}{\overset{^}{d}}}_{g.\mathrm{hard}}^{\left(1\right)}\end{array}$ Formula 9

Wherein Give for organizing 1 rThe estimation base value.

For adjacent group 2, receiving vector x _g ⁽¹⁾In cut (for example by subtracter 74 ₁) obtain behind group 1 the estimation base value x _g ⁽²⁾, shown in formula 10,58.

${\underset{\‾}{x}}_g^{\left(2\right)}={\underset{\‾}{x}}_g^{\left(1\right)}-{\underset{\‾}{\overset{^}{r}}}^{\left(1\right)}$ Formula 10

Consequently, can eliminate effectively in the received signal by what group 1 produced and multiplely advance to connect interference.The high power group (promptly organizing 2) of the next one is used x _g ⁽²⁾, and by group 2 matched filters 66 ₂, group 2JD piece 68 ₂, soft-decision-hard decision piece 70 ₂And organize 2 and disturb generation piece 72 ₂, 60 carry out similar processing.The group 2 that is generated is disturbed To from organize 2 interference cancelling signal, cut (for example by subtracter 24 ₂), thereby obtain ${\underset{\‾}{x}}_g^{\left(2\right)}-{\underset{\‾}{\overset{^}{r}}}^{\left(2\right)}={\underset{\‾}{x}}_g^{\left(3\right)},$ 62。Use this program, can handle each group one by one, until last group G.Because group G is last group, therefore do not need to generate interference signal.Thereby group G only need use group G matched filter 66 _G, group GJD piece 68 _GAnd be used to recover hard character soft-decision-hard decision piece 70 _G, 64.

When at UE 14 ₁On when carrying out SIC-JD, may not handle all groups.If UE 14 ₁All pulse trains that receive all are in highest power received group or the higher received power group, and then UE 14 ₁Only need the cohort that comprises its pulse train is handled.Therefore, UE 14 ₁Required processing can further be simplified.UE 14 ₁The simplification cpable of lowering power of handling consumes, thus extending battery life.

Because N _c* KN dimension matrix is by G JD dimension level N _c* n _iN _s(wherein, i=1 to G, n _iBe i ^ThPulse train number in the group) replace, the complexity of SIC-JD is lower than single step JD.The complexity of JD and the pulse train number of wanting joint-detection square to cube being directly proportional.

The advantage of this method has been to realize the balance between computation complexity and the performance.If all pulse trains are placed one single group, then find the solution problem and can be reduced to the JD problem.By all pulse trains are forced to place same group, perhaps use the threshold value of broad, can realize single marshalling.On the other hand, when cohort only comprises a signal or only receives a signal, find the solution and to be reduced to the SIC-LSE problem.And use narrower threshold value, and perhaps so that being forced to place, each pulse train organizes separately by the scale of each group of hard restriction, can realize this condition.By selecting threshold value, can realize the balance between performance and the computation complexity arbitrarily.

Fig. 6 to 12 is under various multichannel fading channel conditions, the error rate (BER) the performance analog result relatively of SIC-JD and complete JD and RAKE class receiver.Selected parameter is 3G UTRA TDD cdma system parameter: SF=61; W=57.The length of each TDD pulse train/time slot is 2560 signal patch or 667 microseconds.These pulse trains have two and respectively are with N _sThe data field of QPSK symbol, a training sequence hurdle and a guard period.Mo Ni range of operation is 1000 time slots each time.Under any circumstance, pulse train is counted K and all is chosen to be 8.Suppose that herein all receivers all know the channel response of each pulse train exactly, thereby can correctly these pulse trains be sorted and organize into groups.Suppose that simultaneously channel response becomes response when being non-on a time slot, and time slot stands the non-correlation channel response one by one.In this simulation, do not use chnnel coding.All K pulse trains of JD algorithm joint-detection.RAKE class receiver is an i ^ThPulse train code matched filter banks $({\underset{\‾}{\overset{^}{d}}}^{\left(i\right)}=A^{{\left(i\right)}^H}{\underset{\‾}{r}}^{\left(i\right)}).$ Maximal ratio combiner (MRC) level lies in these filters, because they and whole symbol response are complementary.

Performance simulation is to use the multipath program file of ITU channel model definition to carry out under the fading channel condition, the ITU channel model comprises Indoor A, Pedestrian A, Vehicular A model, and 3GPP UTRA TDD Working Group 4 Case 1, Case 2 and Case 3 models.In Vehicular A and Case 2 channels, in the 1%-10%BER scope, compare with complete JD, SIC-JD has experienced falling progressively of 1 decibel (dB) the highest.For all other channels, SIC-JD all is in the 0.5dB scope with the aberrations in property of complete JD.Because Vehicular ACase 2 is the worst conditions in all research situations, therefore only provide its performance chart.In all channels of simulation, Vehicular A and Case 2 channels have maximum delay expansion.Vehicular A is 6 branch model, and its relative delay was respectively for 0,310,710,1090,1730 and 2510 nanoseconds, and average power is respectively 0 ,-1 ,-9 ,-10 ,-15 and-20 decibel (dB) relatively.Case 2 is 3 branch model, and its each branch has identical average power and the relative delay was respectively for 0,976 and 1200 nanoseconds.

Fig. 6 and Fig. 7 compare the error rate (BER) and the signal chip level signal to noise ratio (snr) performance of SIC-LSE receiver under two kinds of multichannel fading channel conditions with complete JD and RAKE class receiver.The scale of group is mandatory to be set to 1, all to form the K group in transmitter and receiver.Also showed binary phase shift keying (BPSK) error rate (BER) theoretical value in additivity white Gaussian noise (AWGN) channel among the figure; Awgn channel has been stipulated the lower bound of BER.BER averages in all pulse train scopes.What Fig. 6 represented is the example of a different channels, and the fading channel of supposing each pulse train in this embodiment and being passed through is separate, but all channels all have the same average power that can draw identical average signal-to-noise ratio (SNR).In such cases,

(i=1 to K) has nothing in common with each other, and γ ⁽ⁱ⁾(i=1 to K) is all identical.This kind situation is present in that power control only compensates long-term decline and/or path loss and in the uplink of uncompensation short-period fading.In each time slot, pulse train is all according to corresponding h ⁽ⁱ⁾(i=1 to K) arranges by power.Fig. 7 represents the similar graph under the common signal channel situation.Suppose all pulse trains all (promptly among this figure by same multipath channel And γ i=1 to K) and all identical, ⁽¹⁾(i=1 to K) difference.At δ ⁽¹⁾Selection on, when making, the difference power of a 2dB is arranged between two adjacent pulse trains according to power hierarchy arrangement pulse train.For example, this kind difference power can be present in base station 12 ₁At different UE (14 ₁-14 ₃) pulse train applies in the downgoing line of different transmission gains.Fig. 6 and Fig. 7 show, in the error rate (BER) scope of 1%-10%, compare with JD, and falling progressively that SIC-LSE stands is not more than 1dB.Non-coding BER (original BER) scope that these common just people are concerned about.Since can not optimization process ISI, significantly falling progressively has appearred in RAKE receiver.Because difference power increases between the pulse train, the performance of SIC-LSE is improved, and when difference power be 1-2dB when (depending on different channels), its performance can compare favourably with complete JD.

Fig. 8,9,10 and 11 compares the BER and the SNR performance of SIC-JD receiver under two kinds of multichannel fading channel conditions with complete JD and RAKE class receiver.8 codes respectively are divided into 4 groups, every group of two codes in transmitter and receiver.In all pulse train scopes, obtain BER mean value.What Fig. 8 and Fig. 9 showed is the example of distinct channels, suppose that in this embodiment the fading channel that each burst blocks passes through is separate, but all channels has the same average power that can draw identical average SNR.All pulse trains in same group obtain identical channel response.In the case, (g=1 to G) is different, and the channel response h of each pulse train in should organizing _g ⁽ⁱ⁾(i=1 to n _s) identical.N wherein _sBe g _ThPulse train number in the group.This has represented the many codes situation in the uplink potentially, in this case, and each UE 14 ₁Send 2 codes.SIC-JD receiver 28 will with same UE 14 ₁Relevant a plurality of codes are organized in same group, thereby form 4 groups.Figure 10 and Figure 11 have showed the situation of common signal channel.All burst blocks of supposition are all by same multipath channel, promptly among this figure

(g=1 to n _s) all identical, and γ _g(g=1 to G) difference.Select γ _gThe time, make when arranging burst blocks according to the power rank, the difference power of a 2dB is arranged between two adjacent groups.This has represented the many codes situation in the down link potentially, in the case, and base station 12 ₁Be each UE14 ₁Send 2 codes.Figure 10 is similar with the performance trend of observed SIC-LSE shown in Fig. 9 to Fig. 8 with trend shown in Figure 11.In the 1%-10%BER scope, the performance of SIC-LSE and JD be (being that difference is not more than 1dB) quite, and the working range of this scope non-coding BER that people were concerned about just.When difference power is 1-2dB when (depending on different channels), the performance of SIC-LSE can compare favourably with complete JD.As shown in the figure, its performance improves with the increase of difference power between two pulse trains.

Figure 12 is similar to FIG. 10, and difference is that two burst blocks are only arranged among Figure 12, and each group comprises 4 pulse trains.As shown in figure 12, in the 1%-10%BER scope, the performance of SIC-JD can compare favourably with JD (being that difference is not more than 1dB).

The complexity of SIC-JD is lower than complete JD.The reduction of complexity comes from uses G JD level dimension matrix N _c* n _iN _s(i=1 to G) replaced the dimension matrix N of single step JD _c* KN _sSimultaneously, because JD relates to matrix inversion in the ordinary course of things, and cube being directly proportional of the complexity of inverting and pulse train number, so the overall complexity of multistage JD will be well below the complete JD of single-stage.And the complexity of SIC part and pulse train number are linear relationship, so can obviously not weaken the advantage of SIC-JD aspect complexity.For example, the complexity of G-1 level interference eliminated can be derived as follows.Because A _g ⁽ⁱ⁾Serial row piece be the shifted version of first row piece, and the supposition Element belong to one of 4 QPSK conformations (general layout) point, therefore can calculate 4n _iIndividual possibility vector, this is for calculating product

Be essential.This step requires per second to carry out $4\mathrm{\α}\·(\mathrm{SF}+W-1)\·\frac{\mathrm{Rate}}{{10}^6}\underset{i=1}{\overset{G-1}{\mathrm{\Σ}}}{n}_i$ 1,000,000 real arithmetics (MROPS).Wherein α=4 are for carrying out the real arithmetic number of times that complex multiplication operation or multiplication and accumulation (MAC) computing are carried out; Rate is SIC-JD the number that per second carries out.Owing to calculated above-mentioned 4n _iIndividual vector, x _g ⁽ⁱ⁺¹⁾The calculation requirement per second carry out $\frac{\mathrm{\α}}{2}\·N_s\·(\mathrm{SF}+W-1)\·\frac{\mathrm{Rate}}{{10}^6}\underset{i=1}{\overset{G-1}{\mathrm{\Σ}}}{n}_i$ 1,000,000 real arithmetics.Owing to only carry out the complex addition computing, therefore carrying out a complex operation only need carry out twice real arithmetic, so above-mentioned formula adopts coefficient Thus, the complexity of G-1 level interference eliminated can be by formula 11 expressions.

$z=\mathrm{\&alpha;}(\mathrm{SF}+W-1)\&CenterDot;(4+\frac{N_s}{2})\&CenterDot;\frac{\mathrm{<figure-callout\; id="10"\; label="Rate"\; filenames="A20061013571800191.png,A20061013571800201.png"\; state="\{\{state\}\}">Rate</figure-callout>}}{{10}^6}\underset{i=1}{\overset{G-1}{\mathrm{\&Sigma;}}}{n}_i$ Formula 11

Soft-decision to the complexity of hard decision conversion can be ignored.

There are several known methods can realize the matrix inversion of JD.For its complexity is described, adopted a kind of very effective approximate Cholesky factor algorithm, to compare with being just accord with the Cholesky factor algorithm, the performance loss that this algorithm relates to is insignificant.This algorithm can be used for finding the solution cohort JD.SIC-JD in the 3GPP UTRA TDD system sees Table 1 with the complexity of complete JD.The complexity of table 1 pair various different scales cohorts compares.As can be seen, when K increases or group scale when reducing, SIC-JD advantage with respect to complete JD aspect complexity also increases thereupon.When the group scale was 1, complexity and the K of SIC-LSE were linear, and when K=16, its complexity is 33% of complete JD.Attention: the maximum number of pulse train is 16 in the UTRA TDD system.When employing is just accord with the Cholesky decomposition algorithm, SIC-JD advantage with respect to complete JD aspect complexity will be more remarkable.Because being just accord with the Cholesky decomposition algorithm has stronger dependence to K, therefore when reducing the JD dimension by SIC-JD, its complexity will further reduce.

Table 1

The pulse train sum	The complexity of SIC-JD is represented with the single step JD complexity percentage form of all K pulse trains
						The K group, every group of scale is 1 (SIC-LSE)	The K/2 group, every group of scale is 2	The K/4 group, every group of scale is 4	The K/8 group, every group of scale is 8
8	63％	67％	76％	100％
					16	33％	36％	41％	57％

As shown in table 1, when number of codes and scale become complete self adaptation on by monitoring basis at interval, the complexity of SIC-JD will be lower than complete JD on average.On average, because the difference of packet threshold, all pulse trains that arrive receiver do not have identical power, so the group scale will be less than the pulse train sum that arrives.In addition, if maximum permission group scale is restricted to firmly less than maximum possible pulse train number, also may reduce the peak value complexity.When roughly the same and pulse train number surpassed maximum permission group scale when the power of the pulse train that arrives receiver, this method can cause performance to a certain degree to reduce.For this reason, SIC-JD provides a kind of mechanism, its realizability can and peak value complexity or required peak value processing power between balance.

Claims (8)

1. the subscriber equipment of the tdd communication systems of an employing code division multiple access, wherein said subscriber equipment are used to receive by a time slot shares a plurality of data-signals that frequency spectrum sends, and it is characterized in that this subscriber equipment comprises:

Be used for receiving the device of a composite signal by the shared frequency spectrum of this time slot;

Be used to estimate the device of the received power level of each data-signal;

The power stage that is used for optionally a plurality of data-signals being grouped into the data-signal that is received is in the device of a plurality of groupings in the certain threshold level, and wherein this certain threshold level is to be adjusted to make this subscriber equipment obtain a desirable error rate;

Be used for detecting separately the device of the data of each grouping data-signal; And

Be used for estimating the device of the received power level of each data-signal according to the priori of subscriber equipment.

2. subscriber equipment according to claim 1 further comprises the device that is used for estimating according to the power stage of the training sequence relevant with each data-signal the received power level of each data-signal.

3. subscriber equipment according to claim 1 further comprises and is used for device that the received power level of each data-signal is estimated, and wherein each matched filter all is complementary with a corresponding data signal code.

4. subscriber equipment according to claim 1, wherein the power stage of this certain threshold level is 2 decibels.

5. subscriber equipment according to claim 1, wherein the power stage of this certain threshold level is 1 decibel.

6. subscriber equipment according to claim 1, wherein this error rate is in the 1%-10% scope.

7. subscriber equipment according to claim 1, further comprise be used for all data-signals be forced to branch in a grouping to surmount the device of selectivity grouping step.

8. subscriber equipment according to claim 1 further comprises and is used for each data-signal is forced to divide to go into grouping separately to surmount the device of selectivity grouping step.

Images