The content of invention
Purpose of the present invention, it is exactly defective at prior art, provide a kind of and can more effectively suppress and reduce DS-CDMA system multi-user interference signals treatment technology and device, promote the scientific and technological progress of 3-G (Generation Three mobile communication system), with the demand that adapts to the continuous increase of cdma system user and communication quality is improved constantly.
For realizing this purpose, the present invention adopts following technical scheme: a kind of compacting and minimizing DS-CDMA system multi-user interference signals processing unit, comprise matched filter array MFB, Viterbi judgement decoder Dv and DS-CDMA base stations control terminal, it is characterized in that: between described matched filter array MFB and Viterbi judgement decoder Dv, embedded " double mode Gauss-Seidel iteration " signal processing module MP who is used to suppress and reduce multi-user interference.
Described signal processing module MP comprises one " Gauss-Seidel iteration data computation basic module Mb ", shunt controller, delay control unit DL, Sign (x) decision device, signal data updating block
The output of the first via of described Mb is joined successively with the feedback data input of controller, DL and Mb along separate routes, constitutes " soft decision Gauss-Sai Deer iterative computation is handled loop S ", the second tunnel output successively with shunt controller, Sign (x) decision device,
And the feedback data input cascade of Mb, constitute " hard decision Gauss-Sai Deer iterative computation is handled loop H ", Third Road output successively with Dv, and the cascade of DS-CDMA base stations control terminal join, finish detection, processing and the judgement decoding task of signal.
Carry out signal processing with described signal processing module MP and comprise following concrete steps:
(1) signal data of earlier MFB being exported is delivered to the initial porch of module S, carrying out " Gauss-Sai Deer iteration " calculates and the convergence differentiation, after this iterative computation is finished, if do not restrain, just after the delay of appropriate time, again the dateout of this iterative computation is delivered to the initial porch of module S, carried out next time " Gauss-Sai Deer iteration " and calculate and the convergence judgement;
(2) if restrain through the soft-decision iteration, just the dateout of Mb is delivered to " hard decision Gauss-Sai Deer iterative computation is handled loop H ", pass through successively Sign (x) decision device with
After doing the renewal of rigid judgement and iterative data, then carry out " basic Gauss-Sai Deer iteration " with Mb again and carry out numerical computations, after iteration proceeds to certain number of times, just Dv and DS-CDMA base stations control terminal are directly delivered in the output of Mb, adjudicate, decoding and subsequent treatment.
The required iterations of described soft judgement can set flexibly according to the requirement of the special error rate of contrast, and the iterations of rigid judgement can be fixed number.
The present invention has realized the purpose of invention, thereby has had following distinguishing feature owing to adopted technique scheme:
(1) compare with prior art, the present invention can increase by 500 the user capacity of system, or under the identical condition of communication quality, the present invention can more effectively suppress or reduce intrasystem interference among multiple users than prior art;
(2) when system load ability K/N=0.5, can be bit error rate BER from 10 of prior art
-1.25Drop to 10
-2.65, illustrate when system user capacity is identical, adopt the present invention that the communication quality of system is improved significantly;
3) signal processing gain is high, " engineering solution " of calculating always exists and complexity Dc lower, the additional delay that is incorporated in the system is shorter, is easy to realization on the engineering.
Obvious, if the technical scheme that 3-G (Generation Three mobile communication system) adopts the present invention to provide will promote the technological progress of DS-CDMA greatly, and produce huge economy and society benefit.
Embodiment
Below in conjunction with accompanying drawing, content of the present invention is described further.
Form schematic diagram as can be seen from the signal processing unit of DS-CDMA system base-station shown in Figure 1, difference between the present invention and the present state-of-the-art technology, be exactly between the matched filter combination array MFB (Matched filter bank) and Viterbi (Viterbi) judgement decoder Dv of prior art, having embedded one is used for compacting and reduces CDMA multiple user signals interference signals processing module MP, the basic task of this module uses " double mode Gauss-Sai Deer iterative technique " that the output of MFB is handled exactly, to suppress or to remove multi-user's cross interference of internal system to greatest extent, and then the user capacity of increase system, the communication quality of raising system.
For the present invention will be described, at first introduce the workflow of DS-CDMA base station, it is like this work: (1) base station receiver receives that at first carrying out radio frequency behind the signal amplifies, and is amplified to A/D conversion behind certain level, then with the output MFB of A/D conversion; (2) send module MP with the output of MFB, and finish signal processing, to suppress, to weaken and even remove the multi-user interference signal of DS-CDMA system by it; (3) send Dv to adjudicate, decode the output of MP, and then finish signal sorting, detect the CDMA signal; (4) output of Dv is sent the control terminal of DS-CDMA base station, and after its information processing, finally realized various types of communications between the user.
Following elder generation provides most crucial content of the present invention in the mode of mathematical derivation, and it also is the creative achievement that the present invention makes.If the signal that the DS-CDMA base station receives can be represented with following mathematic(al) representation:
In the formula, L represents data length, and K represents number of users, and { 0,1} is user's bit signal to d ∈, and w is the energy of signal, and τ is a time delay, n
wBe interchannel noise, and c is a signature waveform; If represent the PN sign indicating number with α, represent signal waveform with p, represent processing gain with N, then can be expressed as c:
As with the bold signs representing matrix, column vector is represented in following setting-out, can be expressed as following matrix form to (1) formula:
y=AW
d+
n (3)
Or
In the formula: A is by 0 and the matrix formed of PN sign indicating number, the diagonal matrix that W is made up of the evolution of signal energy,
dThe matrix of forming by d,
TThe representing matrix transposition, symbol " diag (W) " expression W is a diagonal matrix, thereby can be shown as the described matrix and the list of elements thereof
The received signal matrix
yBehind matching filter array MFB, its output signal matrix is:
y MF=A
T y (9)
With formula (3) substitution (9), can get:
y MF=A
T y=A
TAW
d+A
T n=HW
d+
n′ (10)
In this signal, not only comprising useful signal, also comprising the interference among multiple users signal, also comprising other interchannel noises; In present DS-CDMA system base-station, Viterbi (Viterbi) judgement, decoder, both comprising useful signal according to (10) formula provides exactly, the what is called " mixed signal " that comprising various interference is again adjudicated and is decoded, thereby the error rate (or bit error rate BER) is very high.If we directly do not deliver to Dv to so-called " mixed signal " earlier and adjudicate, decode, but do further processing through a signal processing module earlier, to greatest extent those noise suppressed are fallen, and then send Dv to adjudicate, decode, will reduce the bit error rate of signal greatly, and then increase the capacity of system greatly, improve communication quality.For this reason, we have done many effort, and on the basis of a large amount of l-G simulation tests and contrast, final decision embeds " double mode Gauss-Seidel iteration " signal processing module between MFB and Dv, receive good technical effect, below this has been done deep explanation.
For the convenience of narrating, we do further processing to expression formula (10).If L=3, K=3, N=3, at r constantly, the output signal of matching filter can be expressed as:
By matrix decomposition, and r transmission signals matrix constantly moved to the left side of equal sign, can get:
Wherein: H is a block diagonal matrix, and I is a unit matrix; H
L, r, H shape as shown in Figure 3, and ' up ' and ' low ' represents " go up triangle " and " following triangle " matrix respectively.
(12) formula is exactly that famous Gauss-Sai Deer is iterative.Can prove what it was always restrained, and its calculated value converges on the value [4] that the inverse correlation method calculates.If utilize this formula to carry out only doing rigid judgement (promptly 0,1 judges) after the iterative computation, be exactly the multi-stage iteration method; If (12) formula is at first done soft judgement, after treating that it converges on the inverse correlation method, carry out rigid judgement (i.e. " double mode Gauss-Seidel iteration ") again, just can obtain identical or be better than its effect, but calculate simple more than the latter with " inverse correlation adds multistage iterative method ".
In actual DS-CDMA system, as long as under computer control, give this module MP to finish signal processing the output of MFB, and then finish decoding, the judgement of existing signal by Dv, just can suppress or weaken multi-user's cross interference of DS-CDMA system greatly.
Fig. 8 has provided the simulation results with four kinds of signal processing methods such as " matched filtering method, multi-stage iteration method, inverse correlation method, bimodulus Gauss-Seidel iterations ", the ordinate of curve is bit error rate BER (Bit Error Ratio), abscissa is load capacity K/N, and simulated conditions is SNR=7db; As can be seen: supposition bit error rate BER is 0.01, if signal processing is if only adopt matched filtering MFB, load capacity K/N (establishes the K power system capacity, N is a processing gain), only be 0.15, and after embedding " double mode Gauss-Seidel iteration " signal processing module, load capacity can reach about 0.7, this means: under the prerequisite that guarantees communication quality, can increase by 500 the user capacity of DS-CDMA system; Moreover, all be K/N=0.5 as the load capacity of supposing the whole bag of tricks, when only adopting MFB to carry out signal processing, can get BER=10
-1.25About, adopt " double mode Gauss-Seidel iteration " to carry out after the signal processing, BER just becomes 10
-2.65About; Obvious, adopt technology of the present invention to carry out signal processing, to the multi-user interference of compacting or removal DS-CDMA system, effect is very tangible; In addition, adopt " double mode Gauss-Seidel iteration " signal processing technology, engineering calculation is more or less freely, can prove: square being directly proportional of the computational complexity (Dc) that inversely related adds iterative method and number of users K (is Dc ∝ K
2), and the computational complexity of " bimodulus Gauss-Seidel iteration " only is directly proportional with number of users K, be Dc ∝ KN, the computational complexity that adds iterative method much smaller than inverse correlation, and be in same magnitude with the multi-stage iteration method, this means to be easy on the engineering realize, and the additional delay that is incorporated in the system is shorter.
Fig. 2 has provided and has implemented a concrete instance of the present invention.As can be seen from Figure 2: described signal processing module MP comprises one " Gauss-Seidel iteration data computation basic module Mb ", shunt controller, delay control unit DL, Sign (x) decision device, signal data updating block (transmission signals matrix)
With a signal processing computer (not shown); Mb, shunt controller, DL form so-called " soft decision Gauss-Sai Deer iterative computation processing module S ", and the information flow relation of its iterative computation and soft-decision is: the output of Mb is got back to the feedback data input of Mb again behind shunt controller, DL; Mb, along separate routes controller, Sign (x) decision device,
Form so-called " hard decision Gauss-Sai Deer iterative computation processing module H ", the information flow of its iterative computation and hard decision concerns and is: the output of Mb through shunt controller, Sign (x) decision device,
After, also get back to the feedback data input of Mb; The output of the Third Road of Mb successively with Dv, and the cascade of DS-CDMA base stations control terminal join, under the coordination control of signal processing computer, finally finish detection, processing and the judgement decoding task of signal.It may be noted that: module Mb is specifically designed to and separates system of linear equations, and it does not carry out function soft, hard decision, and decision function is carried out in module S and H.Because when carrying out signal processing with method of the present invention, need to carry out iterative computation and judgement repeatedly with " Gauss-Seidel iteration ", both comprised the soft-decision iterative computation, comprise the hard decision iterative computation again, so the signal processing method this " carrying out iterative computation again; carry out soft, hard decision again " is called " double mode Gauss-Seidel iteration ", it is a creation of the present invention.
MP is performed such signal processing: when the output y of MFB
MFAfter delivering to the signal data inlet of " basic Gauss-Seidel iteration data computation module Mb ", at first carry out " soft-decision iteration numerical computations ", so-called " soft-decision ", promptly only carry out the judgement of " iterative computation " and " whether iteration restrains " with Mb, and without sign function Sign (x) to iteration output carry out 0,1 and adjudicate (being hard decision); After this iterative computation is finished, its data output is added to delayer behind the shunt controller, after reasonable time postpones, send the dateout of this iterative computation the feedback data porch of Mb back to again, proceed the judgement that next iteration calculates and whether restrains When iteration convergence, just the dateout of module S is delivered to the data porch (being the feedback input end of Mb) of " hard decision Gauss-Sai Deer iterative data processing module H ", carry out the hard decision iterative computation, promptly equally also be to call Mb earlier to carry out primary iteration calculating, then the dateout of iteration is added to controller along separate routes, and then with sign function Sign (x) the iteration dateout of Mb is done 0,1 and judge, use
After the iterative data renewal, deliver to the feedback data porch of module Mb again, carrying out next iteration calculates and hard decision ... up to proceeding to certain number of times, just no longer do any rigid determination processing, but these data are directly delivered to the control terminal of Dv and DS-CDMA system base-station, adjudicate successively, subsequent treatment such as decoding.
Fig. 4 has provided the detailed procedure flow process of carrying out information processing with " double mode Gauss-Seidel iteration " (being called for short " G-S method ").Be the convenience of narration, we at first define and are subjected to a certain user that " some other users " disturb system in is " designated user ", and, is referred to as " interference user " all users of other except that " designated user " in the system.Below we describe Fig. 4: this flow chart is a two-layer loop nesting structure, outer cyclic variable i be the reflection system user what, its span is 1 to K (system has K user at most), in cyclic process, wherein each user may become " designated user " and be subjected to the interference of " interference user "; " multi-user interference " that certain " designated user " caused that produced in signal by " interference user " all needs to obtain compacting, and interior loop is calculated multi-user interference and used, the span of variable n also is 1 to K (n ≠ K).Concrete program realizes carrying out in time domain, thinking is to calculate each " interference user " respectively at moment t=r-1, r, the interference Interfer that during r+1 " designated user " is caused, calculate " interference user " interference size total then, upgrade the signal data that send the Mb feedback input end with it then " designated user "
Carry out next iteration and judgement.As can be seen from Figure 4: at first, the outer circulation program is composed initial value to the interference of " interference user ", and promptly Interfer=0 begins interior loop then.In interior loop, at first judge n the time delay difference (τ between " interference user " and " designated user " k
(k)-τ
(n)), if this difference equals 0 (arriving simultaneously), then only need to calculate r constantly n to the interference Int0 of k, because of n is 0 to k in r interference constantly at r-1 and r+1 signal constantly, so on stream, i.e. Int1=0, Int2=0; If (τ
(k)-τ
(n)) greater than 0, then need to calculate user n at r-1 and r signal constantly to user k in r interference constantly, promptly calculate Int1, Int0, at this moment, n is 0 (Int2=0) to user k in r interference constantly at r+1 signal constantly; If (τ
(k)-τ
(n)) less than 0, then need to calculate user n r+1 and r constantly signal to user k in r interference constantly, Int2, Int0, at this moment, n r-1 constantly signal be 0 (being Int1=0) to user k in r interference constantly.The signal of all interference users is added together (∑ Inti) to targeted customer k in r interference contribution constantly, just obtained multi-user interference.After jumping out interior loop, deduct interference value with the output valve of matched filter (MFB)
Promptly be that targeted customer k is at r moment signal Wd
r (k)Estimation, this moment, noise was tentatively suppressed, after several times and above-mentioned similar iteration were adjudicated, multi-user interference just obtained very big compacting.It may be noted that soft number of iterations can be definite flexibly according to the requirement to the wrong rate of bit, and the iterations of rigid judgement can be fixed number.
Now we " interference user " among Fig. 4 at (τ
(k)-τ
(n)) greater than, equal, less than the further refinement of under 0 the different situations " designated user " being disturbed of computational process;
Calculating interference user n at r moment signal to k in the process (see figure 5) of r interference Int0 constantly is: before the circulation, at first compose " 0 " initial value to counter, circulate then, in circulation, at first judge (τ
(k)-τ
(n)) value whether smaller or equal to 0, if must judge that then " i { from 0 to N-1-abs (τ in the PN sign indicating number of user n
(k)-τ
(n)) symbol of value, with (abs (τ in the PN sign indicating number of k
(k)-τ
(n)The symbol of the individual value of)+i) " whether identical.If identical, then counter adds 1, otherwise, subtract 1; If (τ
(k)-τ
(n)) value be greater than 0, then must judge in the PN sign indicating number of targeted customer k i { from 0 to N-1-abs (τ
(k)-τ
(n)) (abs (τ in the symbol of value and the PN sign indicating number of interference user n
(k)-τ
(n)Whether the symbol of the individual value of)+i) is identical.If identical, then counter adds 1, otherwise, subtract 1.After the loop ends, the value of counter is divided by processing gain N and multiply by user n at r signal value constantly, just obtains interference signal.
Calculating interference user n at r-1 moment signal to k in the process (see figure 6) of r interference Int1 constantly is: before the circulation, at first the initial value of counter is composed and placed 0; In circulation, at first judge in the PN sign indicating number of targeted customer k i { from 0 to-(τ
(k)-τ
(n)(N+ (τ in the symbol of)-1} value and the PN sign indicating number of interference user n
(k)-τ
(n)Whether the symbol of the individual value of)+i) is identical, if identical, then register adds 1, otherwise, subtract 1.After the loop ends, the value of counter is divided by processing gain N and multiply by user n at r-1 signal value constantly, just obtains interference signal.
Calculating interference user n at r+1 moment signal to k in the process (see figure 7) of r interference Int2 constantly is: before the circulation, at first the initial value with counter places 0; In circulation, at first judge in the PN sign indicating number of interference user n i { from 0 to (τ
(k)-τ
(n)(N-τ in the symbol of)-1} value and the PN sign indicating number of targeted customer k
(k)+ τ
(n)+ i) whether the symbol of individual value is identical.If identical, then register adds 1, otherwise, subtract 1.After the loop ends, the value of counter is divided by processing gain N and multiply by user n at r+1 signal value constantly, just obtains interference signal.
Signal handler above-mentioned " double mode G-S method " is solidificated in the chip that satisfies the DS-CDMA signal processing requirement (as high speed ROM, E
2ROM etc.), embed between MFB and the Dv by suitable interface then, just can form brand-new, a high performance DS-CDMA base station information treatment system.
[1]R.Lupas?and?S.Verdù,“Linear?multiuser?detectors?forsynchronous?code-division?multiple-access?channels”,IEEETransaction?on?Information?Theory?vol.35,pp.123-136,Jan?1989.
[2]M.K.Varanasi?and?B.Aazhang,“Multistage?detection?inasynchronous?code-division?multiple?access?communications”,IEEE?Trans.Commun.vol.38,pp.509-519,April?1990.
[3]M.K.Varanasi?and?B.Aazhang,“Near-optimum?detection?insynchronous?code-division?multiple-access?communications”,IEEE?Trans.Commun.vol.39,pp.725-736,May?1991.