CN105262708A - MIMO system soft detection method on the basis of pilot frequency channel estimation - Google Patents

Abstract

The present invention discloses a MIMO system soft detection method on the basis of pilot frequency channel estimation, and the method is mainly used for solving problems of low data transmission rate, highest detection complexity and low detection accuracy in the prior art. The method includes the steps as follows: 1. inputting a data vector and a pilot frequency vector; 2. acquiring a receiving signal; 3. performing the pre-filtering; 4. outputting the bit of soft detection; 5. optimally allocating the data vector transmission power. The MIMO system soft detection method on the basis of pilot frequency channel estimation performs the bit analysis of soft detection for the MIMO (multiple-input multiple-output) detection module by employing a MMSE (minimum mean square error) signal detection method and a log-likelihood ratio method, and the detection performance of the bit of soft detection greatly increases the data transmission rate, reduces the detection complexity, and increases the detection accuracy.

Description

The soft detection method of mimo system based on pilot channel estimation

Technical field

The present invention relates to wireless communication technology field, further relate to (multiple-inputmultiple-outputMIMO) soft detection method of the mimo systems based on pilot channel estimation in multi-antenna technology field.The present invention, when ensureing systems axiol-ogy performance, can complete the detection of the MIMO signal of the soft output mimo systems based on pilot channel estimation with higher accuracy rate.

Background technology

The MIMO technology of mimo systems is the important breakthrough of wireless mobile telecommunication technology field intelligent antenna technology.The growing speech business of the mobile communication that faces the future, data service and Broadband Internet business are to the high request of transmission rate, transmission performance and volume of business, when channel capacity and frequency spectrum resource limited, this technology can increase exponentially capacity and the availability of frequency spectrum of communication system when not increasing system bandwidth, utilizing multiple antennas to suppress channel fading, is the key technology that the third generation mobile communication system must adopt.The height of a communication system communication quality largely depends on that whether input is accurate, and input ensures that MIMO technology is played the key of its advantage accurately.In order to make the superiority of MIMO technology be fully played out, performance measured detection algorithm to the greatest extent should be adopted.But the good detection algorithm of performance is often along with higher signal transacting complexity.Therefore, the balance point that searching one is good between detection perform and complexity is needed.People study the MIMO detection method finding soft output and can realize good detection perform when the lower complexity of mimo system in recent years, and research finds that least mean-square error MMSE detection method can be good at the relation between the detection perform of balance sysmte and complexity.But the MMSE detection method in traditional mimo system is not owing to considering the impact of channel estimation errors on systems axiol-ogy performance, and this limits its application to a certain extent.

L.Jalloul, paper " CommentsonSoftDecisionMetricGenerationforQAMwithChannelE the stimationError " (IEEETransactiononCommunications that the people such as S.Alex deliver at it, 2014,11 (62): 4162-4163) refer to a kind of detection method of single-input single-output (single-inputsingle-outputSISO) system of soft output in.In the method, the existence that take into account channel estimation errors, on the impact of systems axiol-ogy performance, obtains soft output detections matrix when channel estimation errors exists in SISO system.The weak point that the method exists is, the method message transmission rate in system channel capacity and the limited environment of frequency spectrum resource is low, cannot meet the demand of nowadays growing communication service, and the detection error rate of soft output detections matrix is higher than the detection of the bit signal of soft output.

The signal detecting method that a kind of high-order M-QAM modulates mimo system is disclosed in patent " signal detecting method in high order modulation mimo systems " (number of patent application 2012101383990, the publication No. CN102724160A) of Xian Electronics Science and Technology University's application.The method utilizes the tree search method of breadth-first, decomposing with the QR of sequence replaces traditional QR to decompose the trigonometric ratio realized real channel matrix, ordering for same layer survivor path is focused on to be divided into two groups, respectively the survivor path in two groups is expanded to the new route of different number, and sub-optimal path is limited within limits, ensure that the survivor path of every layer of some obtained is optimum in the extensions path of respective layer.The deficiency that this patent application exists is that the detection complexity of the method is very high, when ensureing systems axiol-ogy performance, is a kind of detection algorithm of suboptimum relative to the MMSE detection mode of low complex degree.

LouayM.A.Jalloul, paper " the Soft-OutputMIMODetectorswithChannelEstimationError " (IEEESignalProcessingLetters that the people such as SamP.Alex deliver at it, 2015,7 (22): 993-997) refer to the detection method of signal in a kind of soft output mimo systems of nonrandom channel in.Take into account the impact of channel estimation errors on systems axiol-ogy performance in the method, adopt the estimation model of nonrandom channel, focus on the mode that transmitting terminal signal carries out Delamination Transmission, filtering is passed through again after the signal of receiving terminal is received by layering, obtain the bit log likelihood ratio (log-likelihoodratio, LLR) of soft output detections in nonrandom channel.The weak point that the method exists is, the method is not considered and is not considered mutual interference between receiving end signal and channel randomness, and the detection of the signal of a receiving terminal node can be subject to the interference of other detection signals, channel in practical application simultaneously all has randomness, and thus the method can reduce the detection perform of system.

Summary of the invention

The present invention is directed to above-mentioned the deficiencies in the prior art, propose the soft detection method of mimo system based on pilot channel estimation, realize the accuracy rate of higher input.

For achieving the above object, the present invention includes following steps:

(1) data vector and pilot vector is inputted:

(1a) in multiple-input, multiple-output MIMO detection model, input data signal is the data vector of the 16QAM modulation signal composition adopting Gray code mode to map;

(1b) in multiple-input, multiple-output MIMO detection model, the pilot signal of an input unit matrix arbitrarily;

(1c) utilize power normalization formula, power normalization is carried out to pilot signal, obtains pilot vector;

(2) Received signal strength is obtained:

(2a) in multiple-input, multiple-output MIMO detection model, an arbitrarily setting original channel matrices, the element in original channel matrices obeys average to be 0 variance be 1 normal distribution;

(2b) adopt the channel estimation methods of pilot aided, original channel matrices is estimated, obtain accidental channel matrix;

(2c) in multiple-input, multiple-output MIMO detection model, adopt linear time invariant transmission method, data vector is transmitted, obtains the reception value of data vector;

(3) pre-filtering:

Adopt least mean-square error MMSE signal detecting method, the reception value of data vector is decomposed, obtains pre-filtering weighting matrix;

(4) bit of soft detection is exported:

(4a) row vector of pre-filtering weighting matrix is multiplied with the reception value of data vector, obtains local signal detected value;

(4b) adopt the probability density of Gaussian Profile, pre-filtering weighting matrix is decomposed, obtain the conditional probability density distribution of local signal detected value;

(4c) adopt log-likelihood ratio method, likelihood logarithm is got to the conditional probability density distribution of local signal detected value, obtains the bit of soft detection;

(4d) bit of soft detection is exported;

(5) optimum allocation of data vector through-put power:

(5a) in (0,1) interval, get arbitrarily the distribution ratio of a real number being greater than 0.5 as data vector through-put power;

(5b) figure place of bits different for all two bit values, by bit, is added by the data vector in the bit of softer detection and step (1a), using itself and the value errored bit number as multiple-input, multiple-output MIMO detection model;

(5c) be divided by by total bit number of the data vector in errored bit number and step (1a), the distribution ratio obtaining data vector through-put power is greater than the error rate of the multiple-input, multiple-output MIMO detection model corresponding to 0.5;

(5d) (0,1) the distribution ratio of a real number being less than 0.5 as data vector through-put power is got arbitrarily in interval, repeat step (5b) and step (5c), the distribution ratio obtaining data vector through-put power is less than the error rate of the multiple-input, multiple-output MIMO detection model corresponding to 0.5;

(5e) the distribution ratio of data vector through-put power is set to 0.5, repeat step (5b) and step (5c), the distribution ratio obtaining data vector through-put power equals the error rate of the multiple-input, multiple-output MIMO detection model corresponding to 0.5;

(5f) comparison step (5c), step (5d), the error rate of the multiple-input, multiple-output MIMO detection model in step (5e), obtains the minimum value of the error rate of multiple-input, multiple-output MIMO detection model;

(5g) using the optimum allocation ratio of the distribution ratio of data vector through-put power corresponding for the minimum value of the error rate of multiple-input, multiple-output MIMO detection model in step (5f) as data vector through-put power.

The present invention has the following advantages compared with prior art:

First, because the present invention uses multiple-input, multiple-output MIMO detection model detection of transmitted signals, overcome the shortcoming that message transmission rate is low in system channel capacity and the limited environment of frequency spectrum resource of single-input single-output SISO detection model in prior art, message transmission rate in the present invention is increased substantially.

Second, because the present invention uses least mean-square error MMSE signal detecting method, pre-filtering is carried out to the reception value of data vector, pre-filtering weighting matrix is utilized directly to be multiplied with the reception value of data vector, overcome the shortcoming that in prior art, detection complexity is very high, make detection complexity in the present invention significantly lower, achieve the compromise of detection perform and detection complexity better.

3rd, because the present invention uses log-likelihood ratio method, the distribution of the conditional probability density of local signal detected value is taken the logarithm likelihood ratio, overcome in prior art and do not consider that mutual interference between receiving end signal and channel randomness reduce the shortcoming of Detection accuracy, the Detection accuracy of the bit of soft detection in the present invention is significantly promoted.

Accompanying drawing explanation

Fig. 1 is flow chart of the present invention;

Fig. 2 is the relation analogous diagram of the present invention under the bit condition of the soft detection proposed;

Fig. 3 is the relation analogous diagram of the present invention under the distribution ratio condition of different data vector through-put powers.

Embodiment

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

By reference to the accompanying drawings 1, be described below realizing concrete steps of the present invention:

Step 1, input data vector and pilot vector.

In multiple-input, multiple-output MIMO detection model, input data signal is the data vector of the 16QAM modulation signal composition adopting Gray code mode to map, and each element of data vector forms by 4 bits.

In multiple-input, multiple-output MIMO detection model, the pilot signal of an input unit matrix, utilizes power normalization formula, carries out power normalization, obtain pilot vector to pilot signal arbitrarily.

The formula of power normalization is as follows:

$x_t=\sqrt{\frac{1}{N}}\×I$

Wherein, x _trepresent pilot vector, represent sqrt operation, N represents the number of multiple-input, multiple-output mimo system reception antenna, × representing multiplication operations, I represents the unit matrix of the capable N row of N.

Step 2, obtains Received signal strength.

In multiple-input, multiple-output MIMO detection model, an arbitrarily setting original channel matrices, the element in original channel matrices obeys average to be 0 variance be 1 normal distribution.

Adopt the channel estimation methods of pilot aided, original channel matrices is estimated, obtains accidental channel matrix.

The concrete steps of the channel estimation methods of pilot aided are as follows:

The first step, in multiple-input, multiple-output MIMO detection model, from (0,10) interval, choose arbitrarily the gross power of a real number as multiple-input, multiple-output MIMO detection model, the distribution ratio of a real number as data vector through-put power is chosen arbitrarily from (0,1) interval;

Second step, according to the following formula, the through-put power of calculated data vector:

P _d＝P×r

Wherein, P _drepresent the through-put power of data vector, P represents the gross power of multiple-input, multiple-output MIMO detection model, and r represents the distribution ratio of data vector through-put power;

3rd step, according to the following formula, calculates pilot vector through-put power:

P _t＝P-P _d

Wherein, P _trepresent pilot vector through-put power, P represents the gross power of multiple-input, multiple-output MIMO detection model, P _drepresent the through-put power of data vector;

4th step, according to the following formula, calculates the reception value of pilot vector:

$y_t=\sqrt{P_t}\×H_t\×x_t+v_t$

Wherein, y _trepresent the reception value of pilot vector, P _trepresent pilot transmission power, represent sqrt operation, × represent multiplication operations, H _trepresent original channel matrices, x _trepresent pilot vector, v _trepresent that obeying average is 0, variance is the measurement noises vector of the normal distribution of 1;

5th step, according to the following formula, calculates the estimated value of original channel matrices row vector:

$h_j=\frac{\sqrt{\frac{P_t}{N}}}{\frac{P_t}{N}+1}\×y_t$

Wherein, h _jrepresent the estimated value of original channel matrices row vector, j represents the sequence number of original channel matrices row vector, j=1,2 ..., N, N represent the number of reception antenna in multiple-input, multiple-output mimo system, represent sqrt operation, × represent multiplication operations, P _trepresent the through-put power of pilot vector, y _trepresent the reception value of pilot vector;

6th step, with the estimated value h of original channel matrices row vector _jtransposition and the through-put power P of data vector _dsquare root be multiplied, obtain channel estimate matrix H ₁, j represents the sequence number of the estimated value of original channel matrices row vector, j=1,2 ..., N, N represent the number of reception antenna in multiple-input, multiple-output mimo system;

7th step, with the evaluated error value g of original channel matrices row vector _etransposition and the through-put power P of data vector _dsquare root be multiplied, obtain channel estimation errors matrix H ₂, e represents channel estimation errors matrix H ₂the sequence number of e row vector, e=1,2 ..., N, N represent the number of reception antenna in multiple-input, multiple-output mimo system, the evaluated error value g of channel matrix row vector _eobeying average is 0, and variance is normal distribution, P represents gross power;

8th step, uses channel estimate matrix H ₁with channel estimation errors matrix H ₂be added, obtain the accidental channel matrix H of the capable N row of M, wherein, M represents the number of transmitting antenna in multiple-input, multiple-output mimo system, and N represents the number of reception antenna in multiple-input, multiple-output mimo system.

Adopt linear time invariant transmission method, data vector is transmitted in accidental channel matrix, obtains the reception value of data vector.

The formula of linear time invariant transmission method is as follows:

$y_d=\sqrt{P_d}\×H\×x+v_d$

Wherein, y _drepresent the reception value of data vector, P _drepresent the through-put power of data vector, represent sqrt operation, × representing multiplication operations, H represents accidental channel matrix, and x represents data vector, v _drepresent that obeying average is 0, variance is the interchannel noise vector of the normal distribution of 1.

Step 3, pre-filtering.

Adopt least mean-square error MMSE signal detecting method, the reception value of data vector is decomposed, obtains pre-filtering weighting matrix.

The concrete steps of least mean-square error MMSE signal detecting method are as follows:

The first step, according to the following formula, the consolidation noise of calculated data vector reception value:

$z=\sqrt{P_d}\×g_e\×x_i+\underset{j\≠i}{\Σ}\sqrt{P_d}\×h_j\×x_i+\underset{e\≠i}{\Σ}\sqrt{P_d}\×g_e\×x_i+v_d$

Wherein, z represents the consolidation noise of data vector reception value, P _drepresent the through-put power of data vector, represent sqrt operation, × represent multiplication operations, g _erepresent the evaluated error value of channel matrix row vector, e represents the sequence number of channel estimation errors matrix e row vector, e=1,2 ..., N, N represent the number of reception antenna in multiple-input, multiple-output mimo system, x _irepresent i-th element of data vector, i represents the subscript of data vector element, i=1,2 ..., N, h _jrepresent the estimated value of original channel matrices row vector, j represents the sequence number of the estimated value of original channel matrices row vector, j=1,2 ..., N, Σ represent sum operation, v _drepresent interchannel noise vector;

Second step, according to the following formula, the covariance of the consolidation noise of calculated data vector reception value:

${\mathrm{\σ}}_x=P_d\×\frac{N}{N+P}\×\left|\right|x|{|}^2\×I_N+I_N$

Wherein, σ _xrepresent the covariance of the consolidation noise of data vector reception value, P _drepresent the through-put power of data vector, N represents the number of reception antenna in multiple-input, multiple-output mimo system, and P represents the gross power of multiple-input, multiple-output MIMO detection model, × representing multiplication operations, x represents data vector, || || ²represent and ask two norm operations, I _nrepresent the unit matrix of the capable N row of N;

Step 3, according to the following formula, calculates pre-filtering weighting matrix:

W＝(H ₁ ^T×H ₁+σ _x×I _N) ^-1×H ₁ ^T

Wherein, W represents pre-filtering weighting matrix, H ₁represent channel estimate matrix, () ^trepresent matrix transpose operation, × represent multiplication operations, σ _xrepresent the covariance of the consolidation noise of data vector reception value, I _nrepresent the unit matrix of the capable N row of N, N represents the number of reception antenna in multiple-input, multiple-output mimo system, () ^-1represent inversion operation.

Step 4, exports the bit of soft detection.

The row vector of pre-filtering weighting matrix is multiplied with the reception value of data vector, obtains local signal detected value.

Adopt the probability density of Gaussian Profile, pre-filtering weighting matrix is decomposed, obtain the conditional probability density distribution of local signal detected value.

The concrete steps of the probability density of Gaussian Profile are as follows:

The first step, according to the following formula, the consolidation noise energy of calculated data vector reception value:

${\mathrm{\σ}}_n=P_d\×\left(\right|\left|w_m\right|{|}^2+\underset{j\≠m}{\Σ}\left(\left|\right|w_m\×h_j|{|}^2+|\left|w_m\right|{|}^2\right))+\left|\right|w_m|{|}^2$

Wherein, σ _nrepresent the consolidation noise energy of data vector reception value, n represents the sequence number of the consolidation noise energy of data vector reception value, n=1,2 ..., N, N represent the number of reception antenna in multiple-input, multiple-output mimo system, P _drepresent the through-put power of data vector, || || ²represent and ask two norm operations, × represent multiplication operations, w _mrepresent m row vector of pre-filtering weighting matrix, m represents the sequence number of pre-filtering weighting matrix row vector, m=1,2 ..., N, m=n, Σ represent summation operation, h _jrepresent the estimated value of original channel matrices row vector, j represents the sequence number of the estimated value of original channel matrices row vector, j=1,2 ..., N;

Second step, by the row vector w of pre-filtering weighting matrix _mwith the estimated value h of original channel matrices row vector _jbe multiplied, obtain weight coefficient ρ;

3rd step, according to the following formula, calculates the conditional probability density distribution of local detected value:

$f\left(y_k\right|x_i)=\frac{1}{2\×\mathrm{\π}\×{\mathrm{\σ}}_n}\exp (-\frac{|y_k-\mathrm{\ρ}\×x_i{|}^2}{2\×{\mathrm{\σ}}_n})$

Wherein, f (y _k| x _i) represent the distribution of the conditional probability density of local detected value, y _krepresent local detected value, k represents the sequence number of local detected value element, k=1,2 ..., N, N represent the number of reception antenna in multiple-input, multiple-output mimo system, x _irepresent i-th element of data vector x, i represents the sequence number of data vector element, and i=k, π represent circumference ratio, × representing multiplication operations, exp () expression asks index operation, σ _nrepresent the consolidation noise energy of data vector reception value, n represents the sequence number of the consolidation noise energy of data vector reception value, and n=k, ρ represent weight coefficient, || ²represent the square operation asking modulus value.

Adopt log-likelihood ratio method, likelihood logarithm is got to the conditional probability density distribution of local signal detected value, obtains the bit of soft detection, and export the bit of soft detection.

The concrete steps of log-likelihood ratio method are as follows:

The first step, according to the following formula, calculates the log-likelihood ratio of the conditional probability density distribution of local detected value:

$L\left(b_{l,i}\right)=ln\frac{\underset{x_i\∈S_1}{\Σ}f\left(y_k\right|x_i)}{\underset{x_{\mathrm{\ρ}}\∈S_0}{\Σ}f\left(y_k\right|x_p)}$

Wherein, L (b _l,i) represent the log-likelihood ratio that the conditional probability density of local detected value distributes, b _l,irepresent l bit of data vector i-th element, i represents the sequence number of data vector element, i=1,2 ..., N, N represents the number of reception antenna in multiple-input, multiple-output mimo system, and l represents the sequence number of data vector element bit, l=1,2,, right log operations is taken from 4, ln () expression, ∑ represents sum operation, S ₁represent that in data vector x, l all bit values is the set of the element of 1, x _irepresent i-th element of data vector x, x _i∈ S ₁, ∈ represents and belongs to symbol, S ₀represent that in data vector x, l all bit values is the set of 0 element, x _prepresent p the element of data vector x, p=1,2 ..., N, p ≠ i, x _p∈ S ₀, f (y _k| x _i) represent data vector i-th element x _ilocal detected value conditional probability density distribution, y _krepresent local detected value, k represents the sequence number of local detected value element, k=i, f (y _k| x _p) represent data vector p element x _plocal detected value conditional probability density distribution;

Second step, according to the following formula, simplifies log-likelihood ratio, obtains the bit of soft detection:

$\mathrm{\Λ}\left(b_{l,i}\right)=\frac{1}{2{\mathrm{\σ}}_n}\left(\right|y_k-r_0{|}^2-|y_k-r_1{|}^2)$

Wherein, Λ (b _l,i) represent the bit of soft detection, b _l,irepresent l bit of i-th element of data vector x, i represents the sequence number of data vector element, i=1,2 ... N, N represent the number of reception antenna in multiple-input, multiple-output mimo system, and l represents the sequence number of the bit of data vector i-th element, l=1,2,, 4, × represent multiplication operations, σ _nrepresent the consolidation noise energy of the reception value of data vector, n represents the sequence number of the consolidation noise energy of the reception value of data vector, n=i, || ²represent the square operation asking modulus value, y _krepresent local detected value, k represents the sequence number of local detected value element, k=i, S ₀represent that in data vector x, l all bit values is the set of 0 element, r _f∈ S ₀, r ₀represent and make expression formula | y _k-r _f| ²obtain the r of minimum value _fvalue, ∈ represents and belongs to symbol, S ₁represent that in data vector x, l all bit values is the set of 1 element, r _q∈ S ₁, r ₁represent and make expression formula | y _k-r _q| ²obtain the r of minimum value _qvalue.

Step 5, the optimum allocation of data vector through-put power.

The first step, gets arbitrarily the distribution ratio of a real number being greater than 0.5 as data vector through-put power in (0,1) interval.

Second step, the figure place of bits different for all two bit values, by bit, is added by the bit of softer detection and data vector, using itself and the errored bit number of value as multiple-input, multiple-output MIMO detection model.

3rd step, is divided by total bit number of errored bit number and data vector, and the distribution ratio obtaining data vector through-put power is greater than the error rate of the multiple-input, multiple-output MIMO detection model corresponding to 0.5.

4th step, (0,1) the distribution ratio of a real number being less than 0.5 as data vector through-put power is got arbitrarily in interval, repeat second step and the 3rd step of this step, the distribution ratio obtaining data vector through-put power is less than the error rate of the multiple-input, multiple-output MIMO detection model corresponding to 0.5.

5th step, is set to 0.5 by the distribution ratio of data vector through-put power, repeats second step and the 3rd step of this step, and the distribution ratio obtaining data vector through-put power equals the error rate of the multiple-input, multiple-output MIMO detection model corresponding to 0.5;

6th step, relatively the 3rd step of this step, the 4th step, the error rate of the multiple-input, multiple-output MIMO detection model in the 5th step, obtains the minimum value of the error rate of multiple-input, multiple-output MIMO detection model.

7th step, using the optimum allocation ratio of the distribution ratio of data vector through-put power corresponding for the minimum value of the error rate of the multiple-input, multiple-output MIMO detection model in the 6th step of this step as data vector through-put power.

Be described further below in conjunction with accompanying drawing 2 and accompanying drawing 3 pairs of effects of the present invention.

1. simulated conditions:

Emulation experiment 1 of the present invention and emulation experiment 2 are based on common simulated conditions: simulated environment is MATLABR2014a, the MIMO detection system that mimo systems adopts 22 to receive, the modulation system of data vector is 16QAM, channel type is multiple Gaussian channel, suppose that the channel coefficients of multiple Gaussian channel type remains unchanged within the transmission time of adjacent input signal, the cycle-index of statistics bit error rate is 10000 times.

2. emulate content:

Emulation experiment 1, the bit method of soft detection proposed with the present invention and the bit method of the soft detection of prior art compare, the relation between emulation signal to noise ratio and the error rate.

Emulation experiment 2, the bit method of the soft detection proposed with the present invention is in data vector transmit power allocation ratio r respectively 0.3, and the simulation performance of 0.5,0.7 place's value is made comparisons, relation between emulation signal to noise ratio and the error rate.

3. analysis of simulation result:

Emulation experiment 1, makes comparisons with the bit of the inventive method to the bit of optimal soft detection and traditional soft detection simultaneously.

Accompanying drawing 2 is that the bit of bit to the bit of optimal soft detection and traditional soft detection of the soft detection proposed is made comparisons the analogous diagram obtained simultaneously.Abscissa in Fig. 2 represents signal to noise ratio, and the unit of signal to noise ratio is decibel, and ordinate represents the error rate.Represent the bit performance simulation curve of optimal soft detection with the curve of circle mark in Fig. 2, represent the bit performance simulation curve of the soft detection of proposition with the curve of asterisk mark, represent the bit performance simulation curve of traditional soft detection with the curve of square frame mark.In fig 2, the error rate changes along with the change of signal to noise ratio, and the value of signal to noise ratio is larger, and the value of the error rate is less.The bit simulation curve of soft detection proposed and the bit simulation curve of traditional soft detection are made comparisons, when signal to noise ratio is identical, the error rate that the bit of the soft detection proposed is corresponding is minimum, when the error rate is identical, the signal to noise ratio that the bit of the soft detection of proposition is corresponding has the gain of 4dB nearly.The bit simulation curve of soft detection proposed and the bit simulation curve of optimal soft detection are made comparisons, when signal to noise ratio is identical, the error rate of the bit of the soft detection of proposition is close to the error rate of the bit of optimal soft detection.Therefore, the bit of the soft detection of the present invention's proposition has higher Detection accuracy.

Emulation experiment 2, the bit of the soft detection that the inventive method proposes gets 0.3,0.5 respectively in data vector transmit power allocation ratio r, and detection perform when 0.7 is made comparisons simultaneously.

Accompanying drawing 3 is that the bit of the soft detection proposed compares the analogous diagram obtained when data vector transmit power allocation ratio r gets 0.3,0.5,0.7 respectively.Abscissa in Fig. 3 represents signal to noise ratio, and the unit of signal to noise ratio is decibel, and ordinate represents the error rate.The bit performance simulation curve of the soft detection proposed when representing r=0.3 with the curve of circle mark in Fig. 3, the bit performance simulation curve of the soft detection proposed when representing r=0.5 with the curve of asterisk mark, the bit performance simulation curve of the soft detection proposed when representing r=0.7 with the curve of square frame mark.In fig. 3, the error rate changes along with the change of signal to noise ratio, and the value of signal to noise ratio is larger, and the value of the error rate is less.When signal to noise ratio is identical, minimum compared to error rate during r=0.3 and r=0.7 during data vector transmit power allocation ratio r=0.5.Therefore, the optimum allocation ratio of the data transmission utilization measure of the bit of the soft detection of the inventive method proposition is 0.5.

In sum, the present invention uses least mean-square error MMSE signal detecting method and log-likelihood ratio method to carry out the analysis of the bit of soft detection to multiple-input, multiple-output MIMO detection model, compare with the bit of traditional soft detection with the tree search method of prior art breadth-first, improve the Detection accuracy of the bit of soft detection, reduce detection complexity, improve message transmission rate.

Claims (8)

1., based on the soft detection method of mimo system of pilot channel estimation, comprise the steps:

(1) data vector and pilot vector is inputted:

(1a) in multiple-input, multiple-output MIMO detection model, input data signal is the data vector of the 16QAM modulation signal composition adopting Gray code mode to map;

(1b) in multiple-input, multiple-output MIMO detection model, the pilot signal of an input unit matrix arbitrarily;

(1c) utilize power normalization formula, power normalization is carried out to pilot signal, obtains pilot vector;

(2) Received signal strength is obtained:

(2a) in multiple-input, multiple-output MIMO detection model, an arbitrarily setting original channel matrices, the element in original channel matrices obeys average to be 0 variance be 1 normal distribution;

(2b) adopt the channel estimation methods of pilot aided, original channel matrices is estimated, obtain accidental channel matrix;

(2c) in multiple-input, multiple-output MIMO detection model, adopt linear time invariant transmission method, data vector is transmitted, obtains the reception value of data vector;

(3) pre-filtering:

Adopt least mean-square error MMSE signal detecting method, the reception value of data vector is decomposed, obtains pre-filtering weighting matrix;

(4) bit of soft detection is exported:

(4a) row vector of pre-filtering weighting matrix is multiplied with the reception value of data vector, obtains local signal detected value;

(4b) adopt the probability density of Gaussian Profile, pre-filtering weighting matrix is decomposed, obtain the conditional probability density distribution of local signal detected value;

(4c) adopt log-likelihood ratio method, likelihood logarithm is got to the conditional probability density distribution of local signal detected value, obtains the bit of soft detection;

(4d) bit of soft detection is exported;

(5) optimum allocation of data vector through-put power:

(5a) in (0,1) interval, get arbitrarily the distribution ratio of a real number being greater than 0.5 as data vector through-put power;

(5b) figure place of bits different for all two bit values, by bit, is added by the data vector in the bit of softer detection and step (1a), using itself and the value errored bit number as multiple-input, multiple-output MIMO detection model;

(5c) be divided by by total bit number of the data vector in errored bit number and step (1a), the distribution ratio obtaining data vector through-put power is greater than the error rate of the multiple-input, multiple-output MIMO detection model corresponding to 0.5;

(5d) (0,1) the distribution ratio of a real number being less than 0.5 as data vector through-put power is got arbitrarily in interval, repeat step (5b) and step (5c), the distribution ratio obtaining data vector through-put power is less than the error rate of the multiple-input, multiple-output MIMO detection model corresponding to 0.5;

(5e) the distribution ratio of data vector through-put power is set to 0.5, repeat step (5b) and step (5c), the distribution ratio obtaining data vector through-put power equals the error rate of the multiple-input, multiple-output MIMO detection model corresponding to 0.5;

(5f) comparison step (5c), step (5d), the error rate of the multiple-input, multiple-output MIMO detection model in step (5e), obtains the minimum value of the error rate of multiple-input, multiple-output MIMO detection model;

(5g) using the optimum allocation ratio of the distribution ratio of data vector through-put power corresponding for the minimum value of the error rate of multiple-input, multiple-output MIMO detection model in step (5f) as data vector through-put power.

2. the soft detection method of the mimo system based on pilot channel estimation according to claim 1, it is characterized in that, the data vector of the 16QAM modulation signal composition that the employing Gray code mode described in step (1a) maps refers to, each element x of data vector x _iby 4 bit compositions, i represents the sequence number of i-th element in data vector x, i=1,2 ..., N, N represent the number of multiple-input, multiple-output mimo system reception antenna.

3. the soft detection method of the mimo system based on pilot channel estimation according to claim 1, is characterized in that, the power normalization formula described in step (1c) is as follows:

$x_t=\sqrt{\frac{1}{N}}\×I$

Wherein, x _trepresent pilot vector, represent sqrt operation, N represents the number of multiple-input, multiple-output mimo system reception antenna, × representing multiplication operations, I represents the unit matrix of the capable N row of N.

4. the soft detection method of the mimo system based on pilot channel estimation according to claim 1, is characterized in that, the channel estimation methods of the pilot aided described in step (2b), estimates that the concrete steps of accidental channel matrix are as follows:

The first step, in multiple-input, multiple-output MIMO detection model, from (0,10) interval, choose arbitrarily the gross power of a real number as multiple-input, multiple-output MIMO detection model, the distribution ratio of a real number as data vector through-put power is chosen arbitrarily from (0,1) interval;

Second step, according to the following formula, the through-put power of calculated data vector:

P _d＝P×r

Wherein, P _drepresent the through-put power of data vector, P represents the gross power of multiple-input, multiple-output MIMO detection model, and r represents the distribution ratio of data vector through-put power;

3rd step, according to the following formula, calculates the through-put power of pilot vector:

P _t＝P-P _d

Wherein, P _trepresent the through-put power of pilot vector, P represents the gross power of multiple-input, multiple-output MIMO detection model, P _drepresent the through-put power of data vector;

4th step, according to the following formula, calculates the reception value of pilot vector:

$y_t=\sqrt{P_t}\×H_t\×x_t+v_t$

Wherein, y _trepresent the reception value of pilot vector, P _trepresent the through-put power of pilot vector, represent sqrt operation, × represent multiplication operations, H _trepresent original channel matrices, x _trepresent pilot vector, v _trepresent that obeying average is 0, variance is the measurement noises vector of the normal distribution of 1;

5th step, according to the following formula, calculates the estimated value of original channel matrices row vector:

$h_j=\frac{\sqrt{\frac{P_t}{N}}}{\frac{P_t}{N}+1}\×y_t$

Wherein, h _jrepresent the estimated value of original channel matrices row vector, j represents the sequence number of original channel matrices row vector, j=1,2 ..., N, N represent the number of reception antenna in multiple-input, multiple-output mimo system, represent sqrt operation, × represent multiplication operations, P _trepresent the through-put power of pilot vector, y _trepresent the reception value of pilot vector;

6th step, with the estimated value h of original channel matrices row vector _jtransposition and the through-put power P of data vector _dsquare root be multiplied, obtain channel estimate matrix H ₁, j represents the sequence number of the estimated value of original channel matrices row vector, j=1,2 ..., N, N represent the number of reception antenna in multiple-input, multiple-output mimo system;

7th step, with the evaluated error value g of original channel matrices row vector _etransposition and the through-put power P of data vector _dsquare root be multiplied, obtain channel estimation errors matrix H ₂, e represents channel estimation errors matrix H ₂the sequence number of e row vector, e=1,2 ..., N, N represent the number of reception antenna in multiple-input, multiple-output mimo system, the evaluated error value g of channel matrix row vector _eobeying average is 0, and variance is normal distribution, P represents the gross power of multiple-input, multiple-output MIMO detection model;

8th step, uses channel estimate matrix H ₁with channel estimation errors matrix H ₂be added, obtain the accidental channel matrix H of the capable N row of M, wherein, M represents the number of transmitting antenna in multiple-input, multiple-output mimo system, and N represents the number of reception antenna in multiple-input, multiple-output mimo system.

5. the soft detection method of the mimo system based on pilot channel estimation according to claim 1, is characterized in that, the formula of the reception value of the employing linear time invariant transmission method calculated data vector described in step (2c) is as follows:

$y_d=\sqrt{P_d}\×H\×x+v_d$

Wherein, y _drepresent the reception value of data vector, P _drepresent the through-put power of data vector, represent sqrt operation, × representing multiplication operations, H represents accidental channel matrix, and x represents data vector, v _drepresent that obeying average is 0, variance is the interchannel noise vector of the normal distribution of 1.

6. the soft detection method of the mimo system based on pilot channel estimation according to claim 1, it is characterized in that, employing least mean-square error MMSE signal detecting method described in step (3), the concrete steps calculating pre-filtering weighting matrix are as follows:

The first step, according to the following formula, the consolidation noise of calculated data vector reception value:

$z=\sqrt{P_d}\×g_e\×x_i+\underset{j\≠i}{\Σ}\sqrt{P_d}\×h_j\×x_i+\underset{e\≠i}{\Σ}\sqrt{P_d}\×g_e\×x_i+v_d$

Wherein, z represents the consolidation noise of data vector reception value, P _drepresent the through-put power of data vector, represent sqrt operation, × represent multiplication operations, g _erepresent the evaluated error value of channel matrix row vector, e represents the sequence number of channel estimation errors matrix e row vector, e=1,2 ..., N, N represent the number of reception antenna in multiple-input, multiple-output mimo system, x _irepresent i-th element of data vector, i=1,2 ..., N, h _jrepresent the estimated value of original channel matrices row vector, j represents the sequence number of the estimated value of original channel matrices row vector, j=1,2 ..., N, Σ represent sum operation, v _drepresent interchannel noise vector;

Second step, according to the following formula, the covariance of the consolidation noise of calculated data vector reception value:

${\mathrm{\σ}}_x=P_d\×\frac{N}{N+P}\×\left|\right|x|{|}^2\×I_N+I_N$

Wherein, σ _xrepresent the covariance of the consolidation noise of data vector reception value, P _drepresent the through-put power of data vector, N represents the number of reception antenna in multiple-input, multiple-output mimo system, and P represents the gross power of multiple-input, multiple-output MIMO detection model, × representing multiplication operations, x represents data vector, || || ²represent and ask two norm operations, I _nrepresent the unit matrix of the capable N row of N;

Step 3, according to the following formula, calculates pre-filtering weighting matrix:

W＝(H ₁ ^T×H ₁+σ _x×I _N) ^-1×H ₁ ^T

Wherein, W represents pre-filtering weighting matrix, H ₁represent channel estimate matrix, () ^trepresent matrix transpose operation, × represent multiplication operations, σ _xrepresent the covariance of the consolidation noise of data vector reception value, I _nrepresent the unit matrix of the capable N row of N, N represents the number of reception antenna in multiple-input, multiple-output mimo system, () ^-1represent inversion operation.

7. the soft detection method of the mimo system based on pilot channel estimation according to claim 1, it is characterized in that, the concrete steps that the employing Gaussian Profile probability density described in step (4b) calculates the conditional probability density distribution of local signal detected value are as follows:

The first step, according to the following formula, the consolidation noise energy of calculated data vector reception value:

${\mathrm{\σ}}_n=P_d\×\left(\right|\left|w_m\right|{|}^2+\underset{j\≠m}{\Σ}\left(\left|\right|w_m\×h_j|{|}^2+|\left|w_m\right|{|}^2\right))+\left|\right|w_m|{|}^2$

Wherein, σ _nrepresent the consolidation noise energy of data vector reception value, n represents the sequence number of the consolidation noise energy of data vector reception value, n=1,2 ..., N, N represent the number of reception antenna in multiple-input, multiple-output mimo system, P _drepresent the through-put power of data vector, || || ²represent and ask two norm operations, × represent multiplication operations, w _mrepresent m row vector of pre-filtering weighting matrix, m represents the sequence number of pre-filtering weighting matrix row vector, m=1,2 ..., N, m=n, Σ represent summation operation, h _jrepresent the estimated value of original channel matrices row vector, j represents the sequence number of the estimated value of original channel matrices row vector, j=1,2 ..., N;

Second step, by the row vector w of pre-filtering weighting matrix _mwith the estimated value h of original channel matrices row vector _jbe multiplied, obtain weight coefficient ρ;

3rd step, according to the following formula, calculates the distribution of the conditional probability density of local detected value:

$f\left(y_k\right|x_i)=\frac{1}{2\×\mathrm{\π}\×{\mathrm{\σ}}_n}\exp (-\frac{|y_k-\mathrm{\ρ}\×x_i{|}^2}{2\×{\mathrm{\σ}}_n})$

Wherein, f (y _k| x _i) represent the distribution of the conditional probability density of local detected value, y _krepresent local detected value, k represents the sequence number of local detected value element, k=1,2 ..., N, N represent the number of reception antenna in multiple-input, multiple-output mimo system, x _irepresent i-th element of data vector x, i represents the sequence number of data vector element, and i=k, π represent circumference ratio, × representing multiplication operations, exp () expression asks index operation, σ _nrepresent the consolidation noise energy of data vector reception value, n represents the sequence number of the consolidation noise energy of data vector reception value, and n=k, ρ represent weight coefficient, || ²represent the square operation asking modulus value.

8. the soft detection method of the mimo system based on pilot channel estimation according to claim 1, is characterized in that, the concrete steps that the employing log-likelihood ratio method described in step (4c) calculates the bit of soft detection are as follows:

The first step, according to the following formula, calculates the log-likelihood ratio of the conditional probability density distribution of local detected value:

$L\left(b_{l,i}\right)=ln\frac{\underset{x_i\∈S_1}{\Σ}f\left(y_k\right|x_i)}{\underset{x_{\mathrm{\ρ}}\∈S_0}{\Σ}f\left(y_k\right|x_p)}$

Wherein, L (b _l,i) represent the log-likelihood ratio that the conditional probability density of local detected value distributes, b _l,irepresent l bit of data vector i-th element, i represents the sequence number of data vector element, i=1,2 ..., N, N represents the number of reception antenna in multiple-input, multiple-output mimo system, and l represents the sequence number of data vector element bit, l=1,2,, right log operations is taken from 4, ln () expression, Σ represents sum operation, S ₁represent that in data vector x, l all bit values is the set of the element of 1, x _irepresent i-th element of data vector x, x _i∈ S ₁, ∈ represents and belongs to symbol, S ₀represent that in data vector x, l all bit values is the set of 0 element, x _prepresent p the element of data vector x, p=1,2 ..., N, p ≠ i, x _p∈ S ₀, f (y _k| x _i) represent data vector i-th element x _ilocal detected value conditional probability density distribution, y _krepresent local detected value, k represents the sequence number of local detected value element, k=i, f (y _k| x _p) represent data vector p element x _plocal detected value conditional probability density distribution;

Second step, according to the following formula, simplifies log-likelihood ratio, obtains the bit of soft detection:

$\mathrm{\Λ}\left(b_{l,i}\right)=\frac{1}{2\×{\mathrm{\σ}}_n}\left(\right|y_k-r_0{|}^2-|y_k-r_1{|}^2)$

Wherein, Λ (b _l,i) represent the bit of soft detection, b _l,irepresent l bit of i-th element of data vector x, i represents the sequence number of data vector element, i=1,2 ... N, N represent the number of reception antenna in multiple-input, multiple-output mimo system, and l represents the sequence number of the bit of data vector i-th element, l=1,2,, 4, × represent multiplication operations, σ _nrepresent the consolidation noise energy of the reception value of data vector, n represents the sequence number of the consolidation noise energy of the reception value of data vector, n=i, || ²represent the square operation asking modulus value, y _krepresent local detected value, k represents the sequence number of local detected value element, k=i, S ₀represent that in data vector x, l all bit values is the set of 0 element, r _f∈ S ₀, r ₀represent and make expression formula | y _k-r _f| ²obtain the r of minimum value _fvalue, ∈ represents and belongs to symbol, S ₁represent that in data vector x, l all bit values is the set of 1 element, r _q∈ S ₁, r ₁represent and make expression formula | y _k-r _q| ²obtain the r of minimum value _qvalue.