CN103475388A - Communication system multi-user detection method based on order parameter reproducing kernel ICA - Google Patents

Abstract

The invention discloses a communication system multi-user detection method based on order parameter reproducing kernel ICA, and relates to the multi-user detection field in mobile communication. The method comprises the following steps: (1) utilizing autocorrelation matrix diagonalization R of a received signal X to find order parameters lambda k of the X, selecting the first k order parameters and writing out a set of orthogonal basis functions of the signal, (2) initializing unmixing vectors by utilizing user spreading codes, (3) initializing an unmixing matrix W to be a K-dimensional unit matrix, (4) calculating a reproducing kernel, using the order parameters obtained from the step (1) as a set of orthogonal bases of the signal and calculating the Hilbert space where the orthogonal bases are located, and (5) calculating Gram matrixes K1,K2..., KK based on signal vector estimated values [b1,..., bM], using the steepest descent method, namely w (i+1)=w (i)-lambda hamiltonian C, to perform iteration on the w and minimizing the contrast functions C(w) of OPRKICA. According to the communication system multi-user detection method based on the order parameter reproducing kernel ICA, the problem that the capacity of system communication is limited due to the facts that multiple users use the same channel at the same time, and multi-site interference and a near-far effect are caused in an actual MIMO-OFDM system is solved.

Description

Communication system multi-user test method based on order parameter reproducing kernel ICA

Technical field

The present invention relates to the Multiuser Detection field in mobile communication, relate in particular to a kind of communication system MIMO-OFDM multi-user test method based on order parameter reproducing kernel ICA.

Background technology

In multi-user MIMO-OFDM system, because a plurality of users use same channel simultaneously, thereby brought multiple access to disturb and near-far interference, therefore the performance of whole system all can descend, badly influence the reception to targeted customer's signal, thereby limited the raising of system communication capacity, attenuating or elimination multiple access disturb and near-far interference is to be badly in need of the major issue solved, Multiuser Detection (Multi-User Detection) technology improves the recognition capability of system to the targeted customer from the angle of input, is the effective ways of head it off.Yet current multiuser detection still exists the problem of poor effect, bad adaptability.

Summary of the invention

The object of the invention is to solve the poor effect of current multiuser detection existence, the problem of bad adaptability, a kind of communication system multi-user test method based on order parameter reproducing kernel ICA is provided.

In order to achieve the above object, the technical solution adopted in the present invention is: it comprises the steps:

Step 1, utilize the autocorrelation matrix of acknowledge(ment) signal X

diagonalization R, the order parameter Λ of searching X _k, choose front k order parameter, and write out one group of orthogonal basis function of signal, Λ _kfor the characteristic value of R, E is mathematic expectaion, r _mmean to receive signal, M is the number that receives signal;

Step 2, utilize user's spreading code to carry out initialization to separating mixed vector

u _kfor k order parameter, wherein s _kit is spreading code;

Step 3, initially dissolving mixed matrix W is K dimension unit matrix, then the 1st row is replaced with to vector

according to b _m=W ^tz _m, calculate signal vector estimated value b _m, z _mexpression will receive the signal signal later according to reconstructing sequence parameter;

Step 4, calculate reproducing kernel, and by the signal one group of orthogonal basis of the resulting order parameter of step 1, calculating Hilbert space, its place is the reproducing kernel of Hilbert space,

K _ithe reproducing kernel that (t, s) is Hilbert space, order parameter place,

φ _k(s) be one group of orthogonal basis function (s, t>=0 are time variable);

Step 5, calculate based on signal vector estimated value [b ₁..., b _m] the Gram matrix be gram matrix K ₁, K ₂..., K _k, by steepest descent method, be that w (i+1)=w (i)-λ ▽ C carries out iteration to w, minimize the contrast function C (w) of OPRKICA, the OPRKICA here is based on order parameter reproducing kernel independent component analysis,

$C\left(w\right)=-\frac{1}{2}\ln \frac{\det C_k}{\det D_k},$ C (w) is contrast function,

$\▿C=\frac{\∂C}{\∂w}-w{\left[\frac{\∂C}{\∂w}\right]}^{T}w,$ C is spreading code length,

In formula: C _kand D _kformed C by kernel function and unit matrix I structure _koff diagonal element be (C _k) _i,j=K _ik _j, i ≠ j, C _kdiagonal entry be

diagonal matrix D _kdiagonal entry be

wherein μ is the regularization factor of reproducing kernel independent component analysis, if convergence, with regard to basis

desired user is estimated, if do not restrain, just do not forwarded to step 3.The present invention has reduced the interference of system counter multiple access and near-far interference and required precision that power is controlled, can utilize more fully the frequency spectrum resource of up link, the message capacity of larger raising system, using known targeted customer's spreading code as the initial value of separating mixed matrix, by the contrast function designed by kernel function, adaptively carry out the mixed matrix of iterative solution.

The ICA here is independent component analysis, and MIMO-OFDM is multi-input multi-output-orthogonal frequency division multiplexing.

Wherein, the described autocorrelation matrix that utilizes signal X of step 1

find the order parameter Λ of X _k, and write out one group of orthogonal basis function of signal, carry out as follows:

Utilize the autocorrelation matrix of signal X

e is mathematic expectaion, r _mmean to receive signal, M is the number that receives signal, diagonalization R, the characteristic value Λ of calculating X _k, U _kaccording to descending front K order parameter, the formation chosen of λ

Λ _k＝diag(λ ₁,…,λ _K),U _k＝[u ₁,…u _K]，

And write out one group of orthogonal basis function of signal,

k is number of users, and Z is the signal later according to reconstructing sequence parameter.

The present invention adopts as above technical scheme, solve in practical MIMO-ofdm system, because using same channel to produce multiple access simultaneously, a plurality of users disturb and near-far interference, badly influence the reception to targeted customer's signal, thereby limited the problem of system communication capacity, improve the recognition capability of system to the targeted customer from the angle of input, the required precision that the reduction system is controlled power for the near-far interference that resists the multiple access interference, the message capacity of raising system, the present invention has better adaptability and better result of use.

The accompanying drawing explanation

Fig. 1 is structural representation block diagram of the present invention;

Fig. 2 be the present invention when different user is counted and the bit error rate performance of MMSE method relatively;

Fig. 3 be the present invention when different near-far interference α and the bit error rate performance that detects of traditional MMSE relatively.

Embodiment

Below in conjunction with accompanying drawing, the invention will be further described.

Communication system multi-user test method based on order parameter reproducing kernel ICA of the present invention, comprise the steps:

Step 1, utilize the autocorrelation matrix of acknowledge(ment) signal X diagonalization R, the order parameter Λ of searching X _k, choose front k order parameter, and write out one group of orthogonal basis function of signal, Λ _kcharacteristic value for R;

Step 2, utilize user's spreading code to carry out initialization to separating mixed vector

u _kfor k order parameter;

Step 3, initially dissolving mixed matrix W is K dimension unit matrix, then the 1st row is replaced with to vector

according to b _m=W ^tz _m, calculate signal vector estimated value b _m;

Step 4, calculate reproducing kernel, and by the signal one group of orthogonal basis of the resulting order parameter of step 1, calculating Hilbert space, its place is the reproducing kernel of Hilbert space,

K _ithe reproducing kernel that (t, s) is Hilbert space, order parameter place,

it is one group of orthogonal basis function;

Step 5, calculate based on signal vector estimated value [b ₁..., b _m] the Gram matrix be gram matrix K ₁, K ₂..., K _k, by steepest descent method, be that w (i+1)=w (i)-λ ▽ C carries out iteration to w, minimize the contrast function C (w) of OPRKICA, the OPRKICA here is based on order parameter reproducing kernel independent component analysis,

$C\left(w\right)=-\frac{1}{2}\ln \frac{\det C_k}{\det D_k},$ C (w) is contrast function,

$\▿C=\frac{\∂C}{\∂w}-w{\left[\frac{\∂C}{\∂w}\right]}^{T}w,$ C is spreading code length,

In formula: C _koff diagonal element be (C _k) _i,j=K _ik _j, i ≠ j, C _kdiagonal entry be

diagonal matrix D _kdiagonal entry be

wherein μ is the regularization factor of reproducing kernel independent component analysis, if convergence, with regard to basis

desired user is estimated, if do not restrain, just do not forwarded to step 3.The present invention has reduced the interference of system counter multiple access and near-far interference and required precision that power is controlled, can utilize more fully the frequency spectrum resource of up link, the message capacity of larger raising system, using known targeted customer's spreading code as the initial value of separating mixed matrix, by the contrast function designed by kernel function, adaptively carry out the mixed matrix of iterative solution.

Wherein, the described autocorrelation matrix that utilizes signal X of step 1

find the order parameter Λ of X _k, and write out one group of orthogonal basis function of signal, carry out as follows:

Utilize the autocorrelation matrix of signal X

diagonalization R, the characteristic value Λ of calculating X _k, U _kaccording to descending front K order parameter, the formation chosen of λ

Λ _k＝diag(λ ₁,…,λ _K),U _k＝[u ₁,…u _K]，

And write out one group of orthogonal basis function of signal,

k is number of users, and Z is the signal later according to reconstructing sequence parameter.

Effect of the present invention can further illustrate by emulation:

Set up on computers the MIMO-OFDM model, the length of spreading code is C=30, multipath number L=5.The selected reproducing kernel of OPRKICA is based on the reproducing kernel that the order parameter generates, and MMSE is that least mean-square error is estimated.

At first, when number of users is respectively K=2,4,6, by the MIMO-OFDM multiuser detection algorithm based on OPRKICA, detected respectively, background noise is white Gaussian noise, signal to noise ratio between 1～20dB, fixing signal to noise ratio to each, detecting length is the sequence of 100 code elements, result is got the mean value of 10 independent emulation, and each user's power is identical, and the testing result based on OPRKICA and traditional MMSE detection method are compared.As shown in Figure 2, result shows simulation result, and when number of users is different, the error rate of multi-user test method of the present invention always is less than traditional MMSE detection method, so multi-user test method successful of the present invention is better than traditional MMSE detection method.

Then, the fixed-line subscriber number makes K=4, changes near-far interference, and method is to make P _dmean targeted customer's power, other 3 user power are identical is all P _i=α P _dα is far and near coefficient, make α=1/3,1,3, carry out the situation of the relative interference user of simulated target user in low coverage, equidistant, long distance, adopt respectively MIMO-OFDM multiuser detection algorithm based on OPRKICA and traditional MMSE detection algorithm to be detected, to verify the performance of this algorithm antagonism near-far interference, background noise is white Gaussian noise, and signal to noise ratio is being between 1～12dB, fixing signal to noise ratio to each, detecting length is the sequence of 100 code elements, and result is got the mean value of 10 independent emulation.Simulation result as shown in Figure 3, result shows, the error rate of multi-user test method of the present invention always is less than the detection method of traditional MMSE, therefore, multi-user test method of the present invention will be got well compared with the bit error rate performance of traditional MMSE detection method antagonism near-far interference at the bit error rate performance of antagonism near-far interference.

Claims (2)

1. the communication system multi-user test method based on order parameter reproducing kernel ICA, is characterized in that, comprises the steps:

Step 1: utilize the autocorrelation matrix that receives signal X

diagonalization R, the order parameter Λ of searching X _k, choose front k order parameter, and write out one group of orthogonal basis function of signal, Λ _kfor the characteristic value of R, E is mathematic expectaion, r _mmean to receive signal, M is the number that receives signal;

Step 2: utilize user's spreading code to carry out initialization to separating mixed vector

u _kfor k order parameter, wherein s _kit is spreading code;

Step 3: initially dissolving mixed matrix W is K dimension unit matrix, then the 1st row is replaced with to vector

according to b _m=W ^tz _m, calculate signal vector estimated value b _m, z _mexpression will receive the signal signal later according to reconstructing sequence parameter;

Step 4: calculate reproducing kernel, by the signal one group of orthogonal basis of the resulting order parameter of step 1, calculating Hilbert space, its place is the reproducing kernel of Hilbert space,

K _ithe reproducing kernel that (t, s) is Hilbert space, order parameter place,

φ _k(s) be one group of orthogonal basis function (s, t>=0 are time variable);

Step 5: calculate based on signal vector estimated value [b ₁..., b _m] the Gram matrix be gram matrix K ₁, K ₂..., K _k, by steepest descent method, be that w (i+1)=w (i)-λ ▽ C carries out iteration to w, minimize the contrast function C (w) of OPRKICA, the OPRKICA here is based on order parameter reproducing kernel independent component analysis,

c (w) is contrast function,

c is spreading code length,

In formula: C _kand D _kformed C by kernel function and unit matrix I structure _koff diagonal element be (C _k) _i,j=K _ik _j, i ≠ j, C _kdiagonal entry be

diagonal matrix D _kdiagonal entry be wherein μ is the regularization factor of reproducing kernel independent component analysis, if convergence, with regard to basis

desired user is estimated, if do not restrain, just do not forwarded to step 3.

2. the communication system multi-user test method based on order parameter reproducing kernel ICA according to claim 1, is characterized in that, the described autocorrelation matrix that utilizes signal X of step 1

find the order parameter Λ of X _k, and write out one group of orthogonal basis function of signal, carry out as follows:

Utilize the autocorrelation matrix of signal X

e is mathematic expectaion, r _mmean to receive signal, M is the number that receives signal, diagonalization R, the characteristic value Λ of calculating X _k, U _kaccording to descending front K order parameter, the formation chosen of λ

Λ _k＝diag(λ ₁,…,λ _K),U _k＝[u ₁,…u _K]，

And write out one group of orthogonal basis function of signal,

k is number of users, and Z is the signal later according to reconstructing sequence parameter.

Images