CN1564499A - Balancing method based on self-correslation match under CDMA low expanding frequency, and its transceiver - Google Patents

Abstract

Basic components of the device include pre filtering sequencer and multisequence low spread spectrum unit at sending end, as well as autocorrelation estimator, crosscorrelation estimator, linear equalizer, and modules for differentiating users, pre filtering and Rayleigh slow fading channel at receiving end. Operating steps are: superposing pre filtering sequence at sending end to make sending signal possess autocorrelation characteristic, training coefficient of equalizer at receiving end; interference between users and codes is eliminated through minimizing crosscorrelation of output signals and restricting width of crosscorrelation; finally, realizing differentiating users and pre filtering for elimination in cascade. Computer simulation and test result indicate that the method and device improve data transmission performance, are suitable to CDMA system in low spread spectrum.

Description

Equalization methods and transceiver under the code division multiple address low band-spreading based on the auto-correlation coupling

Technical field

The invention belongs to the technical field that realizes channel equalization and Multiuser Detection in code division multiple access (CDMA) cellular communication system, be specifically related to equalization methods and transceiver under a kind of code division multiple address low band-spreading based on the auto-correlation coupling.

Background technology

The cdma cellular communication system is the important technology of 3G (Third Generation) Moblie, and it has characteristics such as frequency planning is simple, and power system capacity is big, and ability of anti-multipath is strong, and transmitting power is little and has received increasingly extensive concern.There is the multipath fading phenomenon in the mobile communication system, can causes serious multipath to disturb, produce adjacent-symbol and disturb (ISI), cause channel distortions, data can't accurately be transmitted.When spreading ratio is big, because the interference signal energy is less, can ignore, but along with the raising of code check, the reducing of spreading ratio, this interference just must have been considered.Traditional method has pilot tone (pilot) signal of the information of knowing by reception, by estimation to multipath signal amplitude and phase place, handle multipath fading, RAKE receiver (tapped delay line receiver) just, it generally is used to utilize the diversity (Multi-path Diversity) of multipath effect.Spreading code between the RAKE receiver hypothesis different user is completely orthogonal.But when the bandwidth after sending waveform process spread spectrum surpassed the bandwidth of channel itself, the orthogonality of frequency expansion sequence will be destroyed.The problem of multi-user interference (MAI) will occur, and can impact traditional RAKE receiver performance, therefore needs to use channel equalization technique.Traditional method can be estimated channel parameter by using midamble code, obtain parametric equalizer, but midamble code has also taken bandwidth.And the method for blind Channel equilibrium is only utilized the various statistical properties that send signal, estimates channel parameter, because it need not the midamble code bandwidth and to the satisfying of some characteristic in the WLAN (wireless local area network), has obtained extensive studies.

When using the method for blind equalization, all suppose the situation of user input signal usually as white, utilize then and comprise that the method for lowest mean square and Subspace Decomposition etc. calculates channel parameter, but these methods have channel exponent number is estimated is responsive, the performance when low signal-to-noise ratio that has can worsen, they all require the channel condition of irreducible (irreducible) simultaneously, this all makes their application obtain restriction, and the channel equalization technique that therefore can effectively eliminate MAI and ISI is a key of supporting the spread spectrum system of high-speed data service.

Summary of the invention

The objective of the invention is to propose equalization methods and transceiver under a kind of code division multiple address low band-spreading based on the auto-correlation coupling, so that can effectively eliminate MAI and ISI, and need not midamble code, and performance does not worsen when low signal-to-noise ratio.

The equalization methods based on the auto-correlation coupling under the code division multiple address low band-spreading of the present invention's proposition is at transmitting terminal each road to be sent signal stack pre-filtering sequence earlier, makes the transmission signal have autocorrelation performance; By slow fading rayleigh channel,,, eliminate inter-user interference and intersymbol interference again by the cross-correlation and the autocorrelative width of restriction that minimize output signal to satisfy the ZF condition at receiving terminal training linear equalizer coefficients; Last cascade realizes that the user distinguishes and remove the pre-filtering module.

Among the present invention, the pre-filtering sequence of transmitting terminal stack, can by by the white noise that produces at random through having finite impulse response (FIR) filter of setting different zero points and periodically expand after directly be superimposed upon the user and realize on the sequence of spread spectrum, thereby reach the purpose of multiuser multiplexing on the auto-correlation territory.The training of receiving terminal equalizer is to reach by the auto-correlation width that minimizes about cross-correlation between equalizer output and the output of restriction equalizer; Particularly, the training of linear equalizer can be adopted steepest descent method, minimizes the target function about cross-correlation between equalizer output signal and autocorrelation sequence, and the equalizer that obtains satisfies the ZF condition under muting situation; The user that cascade realizes distinguishes and goes pre-filtering is by differentiating the autocorrelation sequence characteristic of different user output, removing corresponding pre-filtering sequence again and finish after carrier synchronization.

Transceiver based on the auto-correlation coupling is made up of transmitting terminal device and receiving end device two parts, wherein, the transmitting terminal device comprises pre-filtering sequencer (B1), low frequency multiplier (B2) composition of multisequencing, and receiving end device comprises auto-correlation estimator (A1), cross-correlation estimator (A2), linear equalizer (A3), distinguishes the user and removes pre-filtering module (A4).The input that multisequencing is hanged down frequency multiplier (B1) connects user's baseband sampling signal, is connected in series to slow fading rayleigh channel module (C) and linear equalizer after the output sequence signal of output end signal and pre-filtering sequencer is superimposed.Channel output enters receiving end device and handles, at first enter linear equalizer (A3), the output of equalizer enters auto-correlation estimator (A1), cross-correlation estimator (A2) more respectively, and the user distinguishes and goes pre-filtering module (A4), and the estimated result of auto-correlation, cross-correlation estimator will be adjusted parametric equalizer as the input of equalizer simultaneously.Data after the user distinguishes can enter the viterbi algorithm module of separating chnnel coding.

Equalization methods and transceiver based on the auto-correlation coupling, need not training sequence, the multiplexing utilance that improves frequency range of the user of utilization on the auto-correlation territory, have the performance of good anti-white noise simultaneously, and use many antennas (MultipleElement Array) technology further to improve channel utilization.In order to eliminate ISI and the MAI that multipath effect causes, system has utilized auto-correlation coupling (Autocorrelation Matching) principle, and the autocorrelation performance of coupling input signal and output signal carries out the discriminating of signal and the equilibrium of channel.

Equalization methods and transceiver under the code division multiple address low band-spreading proposed by the invention based on the auto-correlation coupling, be applicable under the situation that spreading ratio is lower in the cdma system, the linear equalizer of ZF condition is satisfied in utilization, can effectively eliminate multipath and disturb and multi-user interference, have receptivity well.

Major advantage of the present invention is, can disturb by anti-additive noise, and equalizer satisfies the ZF condition, so anti-near-far interference of energy, less to original CDMA delivery plan change simultaneously, and because the blind balance method that is based on second-order statistic that uses, so do not need the synchronous of midamble code and signal.By realizing the auto-correlation matching principle, our method does not need to use all users' information, does not promptly need to use Multiuser Detection yet.Similar with RAKE receiver, the algorithm of simplified receiver, so the present invention so greatly can effectively satisfy the requirement of high-speed data service.

Description of drawings

Fig. 1 is the general diagram based on the transceiver of auto-correlation coupling under the code division multiple address low band-spreading.Wherein the transmitting terminal device comprises pre-filtering sequencer (B1), and multisequencing is hanged down frequency multiplier (B2).Receiving end device comprises auto-correlation estimator (A1), cross-correlation estimator (A2), and linear equalizer (A3), the user distinguishes and removes pre-filtering module (A4), slow fading rayleigh channel module (C1)

Fig. 2 is a multi-antenna channel model schematic diagram.

Fig. 3 is to use the comparison of linear sowing square method.

Embodiment

Below to the invention in each composition described respectively, thereby further provide the specific embodiment of the present invention.

1, pre-filtering sequencer (B1)

The prefilter sequencer is to have unique autocorrelation performance for the sequence that each user is sent.User's sequence of low spread spectrum output can be considered to satisfy the albefaction condition, so the autocorrelation performance of final transmission signal only depends on the pre-filtering sequence that we will superpose, can be expressed as

s _i(t)＝c _i(t)+v _i(t)， $E\{s_i\left(t\right)\×{\stackrel{\‾}{s}}_i(t-\mathrm{\τ})\}=E\{v_i\left(t\right)\×{\stackrel{\‾}{v}}_i(t-\mathrm{\τ})\},$ [formula 1a]

Need to satisfy signal { v this moment to the pre-filtering sequence of each user's stack _i(l) i=1,2 ..., N} is uncorrelated mutually, and their auto-correlation function r _Vi(τ) satisfy linear independent condition, also promptly following this group collection of functions is linear independent: { r _Vi(k+j _i) | i=1,2 ..., N; j _i=0, ± 1 ... ± L _HK 〉=SF}, wherein SF is a spreading gain, L _HIt is the exponent number of channel.Use following a kind of method:, by i user's prefilter be with the white noise that produces at random for certain given SF:

$F_i\left(z\right)=1+z^{-[\mathrm{SF}+L_G+(2L_G+1)(i-1)]},i=\mathrm{1,2},\·\·\·,N,$ [formula 1b]

L wherein _G=L _W+ L _H, L _WExponent number for equalizer.Can prove that when the input of these filters is incoherent white noise signal, their auto-correlation function will satisfy linear independent condition.

2, the low frequency multiplier (B2) of multisequencing

Use the Wash sign indicating number to hang down spread spectrum among the CDMA, each user's signal becomes behind the spread spectrum:

$s_i\left(t\right)=\underset{i=l}{\overset{P}{\mathrm{\Σ}}}\underset{l=-\∞}{\overset{+\∞}{\mathrm{\Σ}}}{s}_{\mathrm{ij}}\left(l\right)\×c_j(t-l\·\mathrm{SF})+v_i\left(t\right)$

[formula 2]

Suppose i user's s emission signal s _i(t) form by P information sequence.I j information sequence that transmits is defined as s _IjAnd be assumed to be white noise sequence (l).Each information sequence is by orthogonal code (Wash code) c _j(t) band spectrum modulation.SF represents spreading gain.Can see, because the orthogonality of Wash sign indicating number, if at receiving terminal s _i(t) can be restored, by common CDMA method, all P information sequence can both be restored.

3, slow fading rayleigh channel (C1)

Use as shown in Figure 2 model parameter to come the emulation slow fading rayleigh channel, the channel from arbitrary transmitting antenna i to arbitrary reception antenna i can be with a finite impulse response filter H _Ij(z) characterize, the filter maximum order is made as L _H, the coefficient of filter satisfies rayleigh distributed.Institute's noise superimposed is a white Gaussian noise.As suppose that mobile channel is a frequency selectivity slow fading channel model, can think that then in a channel estimating interval, channel parameter is a constant.

4, auto-correlation estimator (A1)

Can know by the auto-correlation matching principle, when input signal satisfies linearity and moves permanence condition, if the autocorrelation sequence value of equalizer output is identical with the autocorrelation sequence that certain user sends sequence, promptly

$r_y\left(k\right)=r_{v_i}\left(k\right)$ For k 〉=SF, [formula 4a]

So, the ZF condition just is met.Use the auto-correlation matching principle that strengthens, making the cross-correlation between equalizer output signal is zero, and the width of the autocorrelation sequence of restriction output is 2 * L _s, L _sDuring for the exponent number of prefilter, equalizer just can satisfy the ZF condition like this.Formula is counted in the calculating of auto-correlation estimator

Rx _Ij(τ, L)=[Rx _Ij(L+ τ) Rx _Ij(L+l+ τ) ... Rs _Ij(+L+ τ)] ^T, [formula 4b] wherein ${\mathrm{Rx}}_{\mathrm{ij}}\left(\mathrm{\τ}\right)=E\{x_i\left(t\right)\×{\stackrel{\‾}{x}}_i(t-\mathrm{\τ})\}=\frac{1}{N-\mathrm{\τ}}\underset{t=\mathrm{\τ}}{\overset{N}{\mathrm{\Σ}}}{x}_i\left(t\right)\×{\stackrel{\‾}{x}}_i(t-\mathrm{\τ}),$ x _i(t) be the t signal on i reception antenna constantly, can there be the calculating stepping type while

${\mathrm{Rx}}_{\mathrm{ij}}(\mathrm{\τ},N)=E\{x_i\left(t\right)\×{\stackrel{\‾}{x}}_i(t-\mathrm{\τ})\}=\frac{N-1-\mathrm{\τ}}{N-\mathrm{\τ}}\×{\mathrm{Rx}}_{\mathrm{ij}}(\mathrm{\τ},N-1)+\frac{1}{N-\mathrm{\τ}}\×x_i\left(N\right)\×{\stackrel{\‾}{x}}_i(N-\mathrm{\τ}),$

[formula 4c]

5, cross-correlation estimator (A2)

In order to make equalizer satisfy the ZF condition, the cross-correlation that equalizer output signal is asked is zero, and this needs the cross-correlation between estimated output signal.The computing formula of cross-correlation estimator is

Rx _Ij(τ, L)=[Rx _Ij(L+ τ) Rx _Ij(L+l+ τ) ... Rs _Ij(+L+ τ)] ^T, [formula 5a] wherein ${\mathrm{Rx}}_{\mathrm{ij}}\left(\mathrm{\τ}\right)=E\{x_i\left(t\right)\×{\stackrel{\‾}{x}}_j(t-\mathrm{\τ})\}=\frac{1}{N-\mathrm{\τ}}\underset{i=\mathrm{\τ}}{\overset{N}{\mathrm{\Σ}}}{x}_i\left(t\right){\stackrel{\‾}{x}}_j(t-\mathrm{\τ}),$ x _i(t) be the t signal on i reception antenna constantly, can there be the calculating stepping type while

${\mathrm{Rx}}_{\mathrm{ij}}(\mathrm{\τ},N)=E\{x_i\left(t\right)\×{\stackrel{\‾}{x}}_i(t-\mathrm{\τ})\}=\frac{N-1-\mathrm{\τ}}{N-\mathrm{\τ}}\×{\mathrm{Rx}}_{\mathrm{ij}}(\mathrm{\τ},N-1)+\frac{1}{N-\mathrm{\τ}}\×x_i\left(N\right)\×{\stackrel{\‾}{x}}_i(N-\mathrm{\τ})$

[formula 5b]

Therefore remain acceptable for the amount of calculation that increases of calculating cross-correlation.

6, linear equalizer (A3)

Equalizer need satisfy the ZF condition, and just the transfer function of equal value for the M of equivalence input M output system need satisfy: G (z)=diag (f ₁(z), f ₂(z) ..., f _M(z)), if f _i(z) be a multinomial, then the multiple access that this moment, equalizer was eliminated between the user disturbs (MAI), but multipath disturbs (ISI) not eliminated fully yet.Ideal situation is to make f _i(z)=z ^-l, l ∈ N, the output to each user just only exists one to postpone item, and the whole system transfer function is exactly transparent like this, has also satisfied the ZF condition.For reaching this purpose, we minimize the target function about the auto-correlation width:

$J_1\left(w\right)=\underset{\mathrm{\τ}=0}{\overset{\mathrm{Ls}}{\mathrm{\Σ}}}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\underset{j=1,j\≠i}{\overset{N}{\mathrm{\Σ}}}\left|\right|w_i\×{\mathrm{Rx}}_{\mathrm{ij}}\left(\mathrm{\τ}\right)\×w_j^T\left|\right|,$ [formula 6a]

With target function about cross-correlation

$J_2\left(w\right)=\underset{\mathrm{\τ}=L_s+1}{\overset{L_s+L_w}{\mathrm{\Σ}}}\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\left|\right|w_i\×{\mathrm{Rx}}_{\mathrm{ij}}\left(\mathrm{\τ}\right)\×w_i^T\left|\right|,$ [formula 6b]

And interpolation restrictive condition ‖ w _i‖=1 minimizes above target function, just can obtain the training formula about equalizer coefficients

$w_i(t+1)=w_i\left(t\right)-{\mathrm{\μ}}_1\frac{\∂J_1\left(w_i\right)}{\∂w_i}-{\mathrm{\μ}}_2\frac{\∂J_2\left(w_i\right)}{{\∂w}_i},$ [formula 6c]

Wherein $\frac{\∂J_1\left(w_i\right)}{\∂w_i}=\underset{r=0}{\overset{L_s}{\mathrm{\Σ}}}{w}_i\×\{R_X\left(\mathrm{\τ}\right)\×w_j\×R_X^T\left(\mathrm{\τ}\right)\}+\underset{\mathrm{\τ}=0}{\overset{L_s}{\mathrm{\Σ}}}{w}_i\×\{R_X^T\left(\mathrm{\τ}\right)\×w_j\×R_X\left(\mathrm{\τ}\right)\}$ [formula 6d]

$\frac{{\∂J}_2\left(w_i\right)}{{\∂w}_i}=\underset{r=L_s+1}{\overset{L_s+L_w}{\mathrm{\Σ}}}[(w_i\×R_X\left(\mathrm{\τ}\right)\×w_i^T)\×w_i\×R_X^T\left(\mathrm{\τ}\right)]+\underset{\mathrm{\τ}=L_s+1}{\overset{L_s+L_w}{\mathrm{\Σ}}}[(w_i\×R_X\left(\mathrm{\τ}\right)\×w_i^T)\×w_i\×R_X\left(\mathrm{\τ}\right)]$ [formula 6e]

For noisy situation, because white noise process equalizer, its autocorrelative width (number of nonzero term) is L to the maximum _w(being the exponent number of equalizer).So we only need minimize τ 〉=L _wThe value of place's auto-correlation function, the method for gained be the influence of the anti-white noise of energy just.For choosing of the exponent number of equalizer, use following formula to determine

$L_w>\frac{{\mathrm{NL}}_h}{M-N}-1$ [formula 6f]

7, the user distinguishes and pre-filtering removal (A4)

According to the auto-correlation matching principle, the equalizer of ZF condition can be eliminated inter-user interference well and multipath disturbs, the uncertainty that also exists the user can exchange at output end position although satisfy.At this moment just need to distinguish the user, our method is to judge the position of the corresponding output of user, amount of calculation that need not be extra by the autocorrelation sequence value that calculating estimates.Removal for the pre-filtering sequence, a kind of simple method is when the pre-filtering sequence energy that superposes is very little, directly remake and once quantize just can remove the pre-filtering sequence, another kind of better also more complicated method is the corresponding user who has judged, deduct corresponding pre-filtering sequence, but this need use fixing white noise sequence to carry out pre-filtering at transmitting terminal, also needs simultaneously to guarantee synchronously.

The method according to this invention can be applied to comprise 3GPPWCDMA and 3GPP2 cdma2000 in any low spreading factor CDMA mobile communication system.Fig. 1 has provided the device detailed structure schematic diagram that this method is applied to the 3GPP2 cdma2000.Pre-filtering sequence generation module (B1) produces each required user's pre-filtering sequence of [formula 1].Multisequencing is hanged down the low spread spectrum computing that spread spectrum module (B2) is finished [formula 2] indication.The channel that slow fading rayleigh channel module (C1) simulation has multipath slow fading characteristic.Auto-correlation estimator (A1) uses the value of the required autocorrelation sequence of [formula 4b] [formula 4] estimation balancing device training.Cross-correlation estimator (A2) uses the value of the required cross-correlation sequence of [formula 5a] [formula 5b] estimation balancing device training.Linear equalizer (A3) trains the equalizer that satisfies the ZF condition according to [formula 6c] [formula 6d] [formula 6e].The user distinguishes and pre-filtering removal module (A4) quantizes to differentiate the user again and recover user's sequence.

It is 5 parts that concrete signal processing is divided into, and is described in detail as follows respectively:

1, at transmitting terminal, with the user's sequence behind each user's to be sent low spread spectrum of signal process multisequencing low spread spectrum module (B2) back generation

2,, the white noise sequence that produces is at random obtained the pre-filtering sequence through pre-filtering sequencer (B1) at transmitting terminal.

3, deliver to each antenna transmission after stack user's sequence and the pre-filtering sequence, just enter Rayleigh slow fading module (C1)

4, at receiving terminal, signal enters the auto-correlation estimator after by baseband filtering and A/D conversion simultaneously, cross-correlation estimator and linear equalizer, the calculating auto-correlation of auto-correlation estimator and cross-correlation estimator iteration and cross-correlation sequential value come the training of iteration again and upgrade the coefficient of equalizer, with the slow variation of adaptive channel parameter.

5, the output of equalizer will enter user's differentiation and pre-filtering removal module, and the subscriber signal of identification and removal pre-filtering can be used as the input of the viterbi algorithm of separating chnnel coding subsequently.

Concrete simulated conditions is as follows:

In emulation, consider the situation of two users (N=2) and four reception antennas (M=4) to CDMA.Spreading factor is made as SF=8.The exponent number of linear equalizer is 1.According to [formula 1b], prefilter can be designed as so

F ₁(z)＝1-z ^-10，F ₂(z)＝1-z ^-15

Ratio is defined as the mean value of the signal to noise ratio on all antennas:

$\mathrm{SNR}=\frac{1}{M}\underset{i=1}{\overset{1}{\mathrm{\Σ}}}(\mathrm{SNR}{)}_t$

M=4 wherein.Modulator approach adopts BPSK.We have done performance to the equalization methods based on the auto-correlation matching principle relatively with traditional least-square methods under fading environment.Transmit 20 Block altogether.We weigh the performance of system with the BER curve, and the BER curve negotiating is 2,000, and the Monte-Carlo computing on 000 obtains.

As can see from Figure 3, the performance of system has a bigger raising, and this point seems particularly evident under the situation of high s/n ratio.But our system does not need to know all users' information when receiving simultaneously, this means that traditional Multiuser Detection is not used.And because blind method does not need to use midamble code, its availability of frequency spectrum is than non-blind method height.

Claims (12)

1, the equalization methods based on the auto-correlation coupling under a kind of code division multiple address low band-spreading is characterized in that at transmitting terminal each road being sent signal stack pre-filtering sequence earlier, makes the transmission signal have autocorrelation performance; By slow fading rayleigh channel,,, eliminate inter-user interference and intersymbol interference again by the cross-correlation and the autocorrelative width of restriction that minimize output signal to satisfy the ZF condition at receiving terminal training linear equalizer coefficients; Last cascade realizes that the user distinguishes and remove the pre-filtering module.

2, the equalization methods under the code division multiple address low band-spreading according to claim 1 based on the auto-correlation coupling, the pre-filtering sequence that it is characterized in that transmitting terminal stack, by by the white noise that produces at random through having the finite impulse response filter of setting different zero points and periodically expand after directly be superimposed upon the user and realize on the sequence of spread spectrum.

3, the equalization methods under the code division multiple address low band-spreading according to claim 1 and 2 based on the auto-correlation coupling, steepest descent method is used in the training that it is characterized in that linear equalizer, minimize the target function about cross-correlation between equalizer output signal and autocorrelation sequence, the equalizer that obtains satisfies the ZF condition under muting situation.

4, the equalization methods under the code division multiple address low band-spreading according to claim 1 and 2 based on the auto-correlation coupling, it is characterized in that it is the autocorrelation sequence characteristic of exporting by the resolution different user that the output symbol sequence of linear equalizer is carried out user's differentiation, what the corresponding pre-filtering sequence of removal realized after carrier synchronization again.

5, the transceiver under a kind of code division multiple address low band-spreading based on the auto-correlation coupling, form by transmitting terminal device and receiving end device, it is characterized in that the transmitting terminal device comprises pre-filtering sequencer (B1), the low frequency multiplier (B2) of multisequencing, the receiving terminal device comprises that auto-correlation estimator (A1), cross-correlation estimator (A2), linear equalizer (A3), user distinguish and remove pre-filtering module (A4) and slow fading rayleigh channel (C1); The input of the low frequency multiplier of the multisequencing of transmitting terminal device links to each other with user's baseband sampling signal output part, the output sequence signal of output end signal and pre-filtering sequencer superimposed back polyphone slow fading rayleigh channel and linear equalizer, the output of equalizer enters auto-correlation estimator, cross-correlation estimator respectively, and the user distinguishes and remove the pre-filtering module, and the output of auto-correlation estimator, cross-correlation estimator simultaneously is also as the input of equalizer; Data after the user distinguishes are as the input of the viterbi algorithm of separating chnnel coding.

6, transceiver according to claim 5 is characterized in that described pre-filtering sequencer (B1) is:

$F_1\left(z\right)=1+z^{-[\mathrm{SF}+L_G+({2L}_G+1)(t-1)]},i=\mathrm{1,2},\·\·\·,N,$ [formula 1b]

L wherein _G=L _W+ L _H, L _WBe the exponent number of equalizer, L _HBe channel exponent number, SF is a spreading gain.

7, transceiver according to claim 5 is characterized in that described multisequencing hangs down spread spectrum module (B2) and implement following computing:

$S_i\left(t\right)=\underset{j=1}{\overset{P}{\mathrm{\Σ}}}\underset{l=-\∞}{\overset{+\∞}{\mathrm{\Σ}}}{s}_{\mathrm{ij}}\left(l\right)\×c_j(t-l\·\mathrm{SF})+v_i\left(t\right)$ [formula 2]

Wherein, s _i(t) be transmitting of i user, form s by P information sequence _Ij(l) be i j information sequence that transmits, and be white noise sequence, c _j(t) be orthogonal code, SF is a spreading gain, V _i(t) be the pre-filtering sequence signal.

8, transceiver according to claim 5 is characterized in that described slow fading rayleigh channel module (C1), and the channel from arbitrary transmitting antenna i to arbitrary reception antenna j adopts finite impulse response filter H _Ij(Z) characterize, its maximum order is L _H, filter coefficient satisfies rayleigh distributed.

9, transceiver according to claim 5 is characterized in that described auto-correlation estimator (A1) adopts the value of the required autocorrelation sequence of following formula formula estimation balancing device training:

Rx _Ii(τ, L)=[Rx _Ii(L+ τ) Rx _Ii(L+1+ τ) ... Rs _Ii(+L+ τ)] ^T, [formula 4b]

Wherein ${\mathrm{Rx}}_{\mathrm{ij}}\left(\mathrm{\τ}\right)=\stackrel{.}{E}\{x_i\left(t\right)\×{\stackrel{\‾}{x}}_i(t-\mathrm{\τ})\}=\frac{1}{N-\mathrm{\τ}}\underset{t=\mathrm{\τ}}{\overset{N}{\mathrm{\Σ}}}{x}_i\left(t\right)\×{\stackrel{\‾}{x}}_i(t-\mathrm{\τ}),$ x _i(t) be the t signal on i reception antenna constantly, the calculating stepping type is arranged simultaneously

$R_{\mathrm{xii}}(\mathrm{\τ},N)=E\{x_i\left(t\right)\×{\stackrel{\‾}{x}}_i(t-\mathrm{\τ})\}=\frac{N-1-\mathrm{\τ}}{N-\mathrm{\τ}}\×{\mathrm{Rx}}_{\mathrm{ii}}(\mathrm{\τ},N-1)+\frac{1}{N-\mathrm{\τ}}\×x_i\left(N\right)\×{\stackrel{\‾}{x}}_i(N-\mathrm{\τ}),$ [formula 4c]

10, transceiver according to claim 5 is characterized in that described cross-correlation estimator (A2) uses following formula to estimate the value of the cross-correlation sequence that training is required:

Rx _Ij(τ, L)=[Rx _Ij(L+ τ) Rx _Ij(L+1+ τ) ... Rs _Ij(+L+ τ)] ^T, [formula 5a]

Wherein ${\mathrm{Rx}}_{\mathrm{ij}}\left(\mathrm{\τ}\right)=E\{x_i\left(t\right)\×{\stackrel{\‾}{x}}_j(t-\mathrm{\τ})\}=\frac{1}{N-\mathrm{\τ}}\underset{t=\mathrm{\τ}}{\overset{N}{\mathrm{\Σ}}}{x}_i\left(t\right)\×{\stackrel{\‾}{x}}_j(t-\mathrm{\τ}),$ x _i(t) be the t signal on i reception antenna constantly, can there be the calculating stepping type while

${\mathrm{Rx}}_{\mathrm{ij}}(\mathrm{\τ},N)=E\{x_i\left(t\right)\×{\stackrel{\‾}{x}}_j(t-\mathrm{\τ})\}=\frac{N-1-\mathrm{\τ}}{N-\mathrm{\τ}}\×{\mathrm{Rx}}_{\mathrm{ij}}(\mathrm{\τ},N-1)+\frac{1}{N-\mathrm{\τ}}\×x_i\left(N\right)\×{\stackrel{\‾}{x}}_j(N-\mathrm{\τ})$ [formula 5b].

11, transceiver according to claim 5 is characterized in that described linear equalizer (A3) uses following formula to train the equalizer that satisfies the ZF condition:

$w_i(t+1)=w_i\left(t\right)-{\mathrm{\μ}}_1\frac{{\∂J}_1\left(w_i\right)}{{\∂w}_i}-{\mathrm{\μ}}_2\frac{{\∂J}_2\left(w_i\right)}{{\∂w}_i},$ [formula 6c]

Wherein

[formula 6d]

[formula 6e].

12, transceiver according to claim 5 is characterized in that described user distinguishes and pre-wave-wave is removed module (A4), is the position of answering output by the autocorrelation sequence that calculating estimates with judging the user, distinguishes the user; When the pre-flock sequence energy of stack when very little, directly remake and once quantize to remove the pre-filtering sequence.

Images