CN103746728B - The MIMO of a kind of mixed self-adapting receives detection method - Google Patents

Abstract

The invention discloses the MIMO of a kind of mixed self-adapting and receive detection method, detection algorithm every kind alternative adopt training signal training obtain the mapping table of different signal to noise ratio and detection algorithm wrong bitrate in advance, and obtain the complexity value of every kind of detection algorithm, before being received signal detection, estimate the signal to noise ratio obtaining receiving signal, from signal to noise ratio with the mapping table of wrong bitrate obtains every kind of wrong bitrate corresponding to detection algorithm current signal, the combination property of every kind of detection algorithm is calculated further according to this wrong bitrate and complexity value, the detection algorithm choosing combination property minimum detects to received signal.The present invention has considered detection algorithm performance in wrong bitrate and complexity two, the degree of priority of two kinds of performances can be adjusted according to applied environment, and the switching of detection algorithm can be carried out according to the signal to noise ratio self adaptation of current Received Signal, it is ensured that receive the performance of signal detection.

Description

The MIMO of a kind of mixed self-adapting receives detection method

Technical field

The invention belongs to MIMO radio transmission technical field, more specifically say, the MIMO relating to a kind of mixed self-adapting receives detection method.

Background technology

Along with the fast development of wireless Internet multimedia communication, the capacity of wireless communication system and reliability urgently promote, and conventional single antenna transmitting-receiving communication system faces a severe challenge.Multiple-input and multiple-output (Multiple-InputMultiple-Output, MIMO) technology is a kind of multi-antenna technology adopting space time processing, can abundant development space resource, when without increasing frequency spectrum resource and transmitting power, promote capacity and the reliability of communication system exponentially, there is better development prospect.MIMO technology has had been written in many wireless communication standards at present, for instance LTE/LTE-A, IEEE802.11n and HSDPA etc..Fig. 1 is LTEMIMO downlink transmission system schematic.As it is shown in figure 1, emission system is divided into: scrambling, modulation, layer mapping, precoding, resource particle map and OFDM symbol generation module.Reception system is an inverse process of emission system, assume that all users of receiving terminal are through synchronization and timing sampling accurately, then accept system and can be divided into OFDM demodulation, solution resource particle mapping, channel estimating, signal detection, the mapping of solution layer, descrambling module.

In common MIMO communication system, many antennas of transmitter, in identical frequency of identical time, utilize the degree of freedom in space, launch identical signal and are called transmission diversity, launch unlike signal and are called spatial reuse.Owing to the signal of transmission through time-varying multidiameter fading channel, can also have the superposition of white Gaussian noise simultaneously, it is more complicated that this allows for MIMO detection algorithm, can consume more system resource, increases cost.The signal received due to every reception antenna of mimo system is the aliasing that signal launched by many transmitting antennas, and MIMO detection technique is exactly come by the Signal separator received, and recovers the original signal on every transmitting antenna.Traditional signal detection algorithm is divided into the detection algorithm of linear processes.Linear detection algorithm mainly has ZF (ZeroForce, ZF), least mean-square error (MinimumMeanSquareError, MMSE) and interference eliminate (SuccessiveInterferenceCancellation, SIC), these algorithms are all fairly simple in realization, but performance can not reach desirable performance.Non-linear detection algorithm, particularly maximum likelihood algorithm (MaximumLikelihood, ML), reached the optimal performance of detection algorithm, but its algorithm complex be in exponential increase along with the increase of antenna number and order of modulation.

Below several classical MIMO detection algorithms are briefly described:

1) ZF detection

ZF detection belongs to linearity test category, and it is that a filtering matrix W is multiplied by reception signal phasor, and for a certain reception antenna therein, this filtering matrix can by the interference nulling of other antennas, it may be assumed that

W_ZF=H^&=(H^HH)^-1H^H(1)

Wherein W_ZFRepresenting the filtering matrix of ZF detection algorithm, H is channel matrix, and subscript H representing matrix seeks conjugate transpose, and subscript-1 representing matrix is inverted.

ZF detection is using the detection signal that obtains on every transmitting antenna as desired signal, and the signal on other antennas is entirely interference signal, and by its zero setting.The pseudoinverse of channel matrix, while obtaining desired signal, has been multiplied by noise again by ZF detection algorithm, there is the problem amplified by noise.ZF algorithm computing is simple, but easily affected by noise, and effect of noise is very sensitive.

2) MMSE detection

The interference that MMSE detection algorithm had both considered between antenna when design filtering matrix is also contemplated for effect of noise.With least mean-square error for criterion, calculate the mean square error between actual transmission signal and estimated value, it may be assumed that

$\hat{X}=\arg \underset{G}{\min }E{\left|\right|G_{\mathrm{MMSE}}Y-X\left|\right|}^2={(H^{H}H+{\mathrm{\σ}}^2{I}_t)}^{-1}{H}^H---\left(2\right)$

WhereinRepresenting testing result, X represents transmitting signal phasor, and Y is for receiving signal phasor, G_MMSERepresent filtering matrix；Argmin () expression is minimized, namely from launching searching minima signal space；||·||²Represent norm square, E () represent be averaging, σ²Represent the variance of additive white Gaussian noise, I_tRepresenting the unit matrix being sized to t, t is transmitting antenna number.

3) SIC detection

The core concept of SIC detection algorithm detects exactly and eliminates.No matter it is that ZF or MMSE can be used to the method as signal detection.First detection a line signal, the signal detected with linear detection algorithm, as prior information, eliminates its interference to other signals, and such a line a line is until all of signal is all detected.Certainly, error propagation is sixty-four dollar question in SIC detection algorithm, because the error of the signal detected can affect detection signal below always.So determining that detection signal sequence is most important, the strongest component of signal generally all can be detected at first, so can effectively increase signal reliability.Also there is several methods that about detection ordering, wherein compare classical having: based on SINR(SignaltoInterferenceplusNoiseRatio, Signal to Interference plus Noise Ratio), based on row and based on SNR(SignaltoNoiseRatio, signal to noise ratio) sequence detection.

4), ML detection

ML detection algorithm is one optimum in current signal detection algorithm, and it, by the thought of exhaustive search, finds correct raw information.The principle of ML detection is: combine all possible transmitting signal phasor in the planisphere that modulation maps, transmitting signal phasor minimum with receiving signal phasor Euclidean distance after selecting channel conversion from all of transmitting signal combining, it may be assumed that

$\hat{X}=\underset{X\∈{\left\{\mathrm{\χ}\right\}}^t}{\arg \min }{\left|\right|Y-\mathrm{HX}\left|\right|}^2---\left(3\right)$

Wherein,Represent the possible vector space launching signal.It can be seen that the performance of ML detection algorithm is necessarily optimum from formula (3), because this algorithm has traveled through all possible transmitting signal phasor in modulation constellation, therefrom choose probability maximum launch the signal phasor output as detection.But, its computational complexity is in exponential increase along with the increase of antenna amount and order of modulation, in actual application and be of little use, because it can consume substantial amounts of power, the computing capability of test receiver.

In order to enable to reduce further algorithm complex, the maximum likelihood algorithm of improvement, use the principle of binary tree search, that tries one's best reduces hunting zone, under the premise as far as possible reaching maximum likelihood property, reduces algorithm complex as much as possible.In these improvement based in the method for tree search, Sphere Decoding Algorithm (SphereDecoding, SD) is effective is converted to a kind of tree search by the exhaustive search of ML detection algorithm, has deleted unnecessary node searching.Wherein Sphere Decoding Algorithm difference according to its way of search again is divided into: depth-first search, BFS and K-best search.These all of quasi-Maximum Likelihood Detection can show good detection performance when low signal-to-noise ratio, but still considerably beyond linear detection algorithm in algorithm complex and system consumption.

Summary of the invention

It is an object of the invention to overcome the deficiencies in the prior art, the MIMO providing a kind of mixed self-adapting receives detection method, take into account the performance and complexity that receive detection algorithm, make MIMO receive system and select different reception detection methods according to the signal quality being currently received.

For achieving the above object, the MIMO of mixed self-adapting of the present invention receives detection method, comprises the following steps:

S1: arrange alternative detection algorithm, adopts every kind of detection algorithm training signal training to obtain the mapping table of different signal to noise ratio snr and detection algorithm wrong bitrate in advance, and obtains the complexity value C of every kind of detection algorithm_i, wherein i is the sequence number of detection algorithm；

S2: current Received Signal is carried out signal-to-noise ratio (SNR) estimation and obtains SNR, obtains every kind of detection algorithm wrong bitrate P under this SNR according to the mapping table of the SNR being previously obtained in step S1 Yu wrong bitrate_i, calculate the combination property V of every kind of detection algorithm_i, computing formula is: V_i=α * P_i+β*C_i, wherein α, β are the constant parameter more than 0, and alpha+beta=1；

The combination property Vi of S3: the comparison step S2 every kind of detection algorithm obtained, current Received Signal is detected by the detection algorithm selecting Vi value minimum.

Wherein, alternative detection algorithm includes squeeze theorem algorithm, least mean-square error-interference cancellation algorithm, maximum likelihood algorithm.

Wherein, the complexity value of detection algorithm is algorithm execution time.

The MIMO of mixed self-adapting of the present invention receives detection method, detection algorithm every kind alternative adopt training signal training obtain the mapping table of different signal to noise ratio and detection algorithm wrong bitrate in advance, and obtain the complexity value of every kind of detection algorithm, before being received signal detection, estimate the signal to noise ratio obtaining receiving signal, from signal to noise ratio with the mapping table of wrong bitrate obtains every kind of wrong bitrate corresponding to detection algorithm current signal, the combination property of every kind of detection algorithm is calculated further according to this wrong bitrate and complexity value, the detection algorithm choosing combination property minimum detects to received signal.The present invention has considered detection algorithm performance in wrong bitrate and complexity two, the degree of priority of two kinds of performances can be adjusted according to applied environment, and the switching of detection algorithm can be carried out according to the signal to noise ratio self adaptation of current Received Signal, it is ensured that receive the performance of signal detection.

Accompanying drawing explanation

Fig. 1 is LTEMIMO downlink transmission system schematic；

Fig. 2 is a kind of detailed description of the invention flow chart of the MIMO reception detection method of mixed self-adapting of the present invention；

Fig. 3 is the corresponding relation curve chart of detection algorithm SNR and the BER that emulation obtains；

Fig. 4 is the execution time comparison diagram of the detection algorithm that emulation obtains；

Fig. 5 is the combination property comparison diagram of the detection algorithm that emulation obtains.

Detailed description of the invention

Below in conjunction with accompanying drawing, the specific embodiment of the present invention is described, in order to those skilled in the art is more fully understood that the present invention.Requiring particular attention is that, in the following description, when perhaps the detailed description of known function and design can desalinate the main contents of the present invention, these descriptions here will be left in the basket.

Embodiment

The present invention is the combination property by relatively various alternative detection algorithms, namely considers wrong bitrate BER(BitErrorRate) and algorithm complex.In present embodiment, alternative detection algorithm includes ZF ZF detection algorithm, least mean-square error-interference eliminates MMSE-SIC detection algorithm, maximum likelihood ML detection algorithm.Different detection algorithms are different in the BER performance of different signal to noise ratio snr.For deterministic signal, from the angle of signal detection, it is generally adopted the amplitude of transmission signal and the ratio of noise average power, it may be assumed that

$\mathrm{SNR}=101\mathrm{og}\left(\frac{{\left|\right|X\left|\right|}^2}{{\mathrm{\σ}}^2}\right)---\left(4\right)$

The BER performance of every kind of detection algorithm is analyzed as follows:

One, ZF detection algorithm

Reception signal is mainly multiplied by a filtering matrix G by ZF detection algorithm, is filtered out as interference by other aerial signals, and linear operation is G=G (H), and wherein G is t × r linear filter matrix, and t is number of transmission antennas, and r is reception antenna quantity.The paired wrong bitrate of condition of ZF detection algorithm can be expressed as:

$\begin{array}{c}P(X-\hat{X}|H)\\ =P({\left|\right|\mathrm{GY}-\hat{X}\left|\right|}^2<{\left|\right|\mathrm{GY}-X\left|\right|}^1)\\ =P({\left|\right|\mathrm{GHX}-\hat{X}+{\mathrm{GH}}_0\left|\right|}^2<{\left|\right|\mathrm{GHX}-X+{\mathrm{GN}}_0\left|\right|}^2)\\ =P\left(\begin{array}{c}{\left|\right|X-\hat{X}\left|\right|}^2\\ +2<{\mathrm{GH}}_0,X-\hat{X}><0\end{array}\right)+2<(\mathrm{GH}-1)X,X-\hat{X}>\\ =E\left[Q\left(\frac{{\left|\right|X-\hat{X}\left|\right|}^2+2<(\mathrm{GH}-I)X-\hat{X}>}{\sqrt{2N_0{\left|\right|G^H(X-\hat{X})\left|\right|}^2}}\right)\right]\end{array}$

Wherein X represents transmitting signal phasor, and Y represents reception signal phasor,Representing testing result, H represents channel matrix, N₀Variances sigma for additive white Gaussian noise², I represents unit matrix, and Q () represents Q function, is defined asExp () represents exponential function.In ZF detection algorithm, filtering matrix is G=H^&, subscript & represents the pseudo-inverse transformation of matrix.In actual applications, reception antenna number r is typically larger than transmitting antenna number t, such H^HH is just necessarily reversible, it is possible to obtain:

G=H^&=(H^HH)^-1H^H(6)

So can obtain:

Wherein, N represents and launches or receive signal code number.

Formula (7) can show that ZF detection algorithm completely eliminated the spatial interference received between signal.The paired wrong bitrate BER of ZF detection algorithm is:

$P(X\&RightArrow;\hat{X}|H)=E\left[Q\left(\frac{{\left|\right|X-\hat{X}\left|\right|}^2}{\sqrt{{2N}_0\left|\right|H{\left(H^{H}H\right)}^{-1}(X-\hat{X})\left|\right|}}\right)\right]---\left(8\right)$

Two, MMSE-SIC detection algorithm

Most important the seeking to of interference elimination detection algorithm determines detection ordering, and the signal making reliability high first detects, thus being effectively improved the detection performance of entirety.Present embodiment mainly determines detection ordering according to SINR, and signal priority MMSE detection algorithm the highest for SINR detects, and then eliminates the signal detected to other interference receiving signal, is all detected until all of signal.Detection signal is expressed as:

$\hat{X}=X{+G}_{\mathrm{MMSE}}v---\left(9\right)$

Wherein G represents the filtering matrix of MMSE detection algorithm, h_ijRepresent in channel matrix H that i-th transmitting antenna is to the channel between jth root reception antenna, x_jRepresent the reception signal of jth root reception antenna, i^*What represent current detection is i-th^*Row signal.

According to analysis to ZF detection algorithm wrong bitrate BER above, it is possible to the expression formula summing up the paired wrong bitrate BER drawing MMSE-SIC is:

$P(X\&RightArrow;\hat{X})=E\left[Q\left(\frac{{\left|\right|X-\hat{X}\left|\right|}^2+2({(H^{H}H+N_0{I}_t)}^{-1}{H}^{H}H-I_t)X,X-\hat{X}}{\sqrt{{2N}_0\left|\right|H{(H^{H}H+N_0{I}_t)}^{-1}(X-\hat{X})|{|}^2}}\right)\right]---\left(10\right)$

Wherein, I_tRepresent the unit matrix being sized to t.

Three, ML detection algorithm

ML detects the least square of decoding algorithm norm | | Y-HX | | launching signal phasor X corresponding to selecting, it may be assumed that

${\left|\right|Y-\mathrm{HX}\left|\right|}^2=\underset{i=1}{\overset{r}{\mathrm{\&Sigma;}}}\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{|y_{\mathrm{in}}-\underset{j=1}{\overset{t}{\mathrm{\&Sigma;}}}{h}_{\mathrm{ij}}{x}_{\mathrm{jn}}|}^2---\left(11\right)$

Wherein, r, t represent reception antenna number and transmission antenna number respectively, and N represents and sends or receive signal code number, y_inRepresent the n-th receiving symbol of i-th reception antenna, x_jnRepresent jth root transmission antenna n-th sends symbol, h_ijRepresent in channel matrix H that i-th transmitting antenna is to the channel between jth root reception antenna.

For Space Time Coding, the upper limit of Average Error Probabilities is:

$P\left(e\right)\&le;\frac{1}{\left|\mathrm{\&chi;}\right|}\underset{X\&Element;\mathrm{\&chi;}}{\mathrm{\&Sigma;}}\underset{\hat{X}\&Element;\mathrm{\&chi;}\backslash \left\{X\right\}}{\mathrm{\&Sigma;}}P(X\&RightArrow;\hat{X})---\left(12\right)$

Wherein χ represents possible transmission symbol space.

The paired wrong bitrate BER of ML detection algorithm is defined as:

$P(X\&RightArrow;\hat{X})=E\left[Q\left(\frac{\left|\right|H(X-\hat{X})\left|\right|}{\sqrt{{2N}_0}}\right)\right]\&le;E\left[\exp (-{\left|\right|H(X-\hat{X})\left|\right|}^2/4N_0)\right]---\left(13\right)$

As seen from the above analysis, certain corresponding relation is there is between the SNR and the BER of detection algorithm of signal, generally speaking, signal SNR is more big, the BER of detection algorithm is more little, detection performance is more good, but different detection algorithm BER under same SNR has difference, therefore can select different detection algorithms according to this character for the signal of different SNR.In three kinds of alternative detection algorithms that present embodiment is chosen, ZF algorithm belongs to linearity test, shows good BER performance generally in high snr cases；And non-linear detection algorithm, such as ML, good BER performance can be shown when low SNR.And MMSE-SIC BER better performances when medium SNR.

Fig. 2 is a kind of detailed description of the invention flow chart of the MIMO reception detection method of mixed self-adapting of the present invention.Comprise the following steps as in figure 2 it is shown, the MIMO of mixed self-adapting of the present invention receives detection method:

S201: every kind of detection algorithm adopts training signal training obtain the mapping table of different signal to noise ratio snr and detection algorithm wrong bitrate in advance, and obtains the complexity value C of every kind of detection algorithm_i, wherein i is the sequence number of detection algorithm.

In actual applications, the SNR range of expectation can be adopted to determine the SNR value of training signal when adopting training signal training.The mapping table of the signal to noise ratio snr being previously obtained and wrong bitrate can store to search use in reception system.In present embodiment, what the complexity value of detection algorithm adopted is the execution time of algorithm, the time namely spent from input signal to output signal detecting result.Algorithm execution time can adopt the execution time average of the training signal of multiple different SNR.

S202: when channel estimating, current Received Signal is carried out signal-to-noise ratio (SNR) estimation, obtain SNR.

S203: search according to the mapping table of SNR and the BER being previously obtained in step S201 and obtain every kind of detection algorithm BERP under the step S202 SNR obtained_i, calculate the combination property V of every kind of detection algorithm_i, computing formula is: V_i=α * P_i+β*C_i, wherein α, β are the constant parameter more than 0, and alpha+beta=1.

α and β is BER and the relative importance value of two kinds of performances of algorithm complex value.In actual application, according to concrete applied environment, more consideration detection algorithm performances sometimes, i.e. the factor of BER, the factor of the more consideration complexity of some meetings.Such as, to some Military Application, it is necessary to obtaining the decoding of accurate data, any cost can spend, it should selects the non-linear detection algorithm that degree of accuracy is higher, and now BER is preferential, and parameter arranges α > β.But for the communication system that some are civilian, fault freedom is in certain scope, and expense is again top-priority factor, therefore can select the detection algorithm that some complexities are relatively low, now can make α < β.Visible, the present invention can regulate the value of α and β dynamically to adapt to different applied environments.And in actual applications, if BERP_iWith complexity value C_iOrder of magnitude difference relatively big, can be unbalanced on the impact of combination property, now can adopt some mathematical measures, reduce the order of magnitude gap of the two.

The combination property V of S204: the comparison step S2 every kind of detection algorithm obtained_i, select V_iCurrent Received Signal is detected by the minimum detection algorithm of value.

Visible, the present invention according to the difference of the SNR receiving signal, can select suitable detection algorithm, when receiving the SNR change of signal, it is possible to self adaptation carries out the switching of detection algorithm.In actual applications, if channel is relatively stable, it is possible to arrange bigger interval switching time, if channel variation is comparatively rapid, it is possible to arrange less interval switching time, thus ensureing to receive signal detection performance.

For beneficial effects of the present invention is described, adopting the present invention to carry out experiment simulation on LTE link level simulation platform, assume that channel is SCM channel when emulation, signal is through desirable transmission, and has accurate channel estimating.This emulation is with complexity performance priority, in combination property computing formula, and α=0.2, β=0.8.Table 1 is simulation parameter table.

Table 1

Fig. 3 is the corresponding relation curve chart of detection algorithm SNR and the BER that emulation obtains.As it is shown on figure 3, single for BER, the best performance of ML detection algorithm, MMSE-SIC detection algorithm takes second place, and ZF detection algorithm is worst.

Fig. 4 is the execution time comparison diagram of the detection algorithm that emulation obtains.As shown in Figure 4, under different SNR, the execution time of MMSE-SIC detection algorithm and ZF detection algorithm is basically identical, all smaller, close to 0；The execution time of ML detection algorithm is significantly larger than the execution time of other two kinds of detection algorithms.Can be seen that, in this emulation, owing to algorithm execution time is bigger, the numbered magnitude of its numerical value and BER differs greatly, if the impact that direct calculating can cause BER performance is more weak, therefore in ensuing combination property calculates, complexity value does not directly adopt the execution time of algorithm, and the execution time of algorithm is carried out certain process, i.e. complexity value C_i=log₁₀(T_i)/10, wherein T_iRepresent the execution time of detection algorithm.

Fig. 5 is the combination property comparison diagram of the detection algorithm that emulation obtains.As shown in Figure 5, under different SNR, the detection algorithm that in three kinds of detection algorithms, combination property is minimum is had nothing in common with each other, SNR combination property value of ZF detection algorithm when 0 to 9dB is minimum, when 9 to 14dB, the combination property value of ML detection algorithm is minimum, during more than 17dB, the comprehensive comparison of three kinds of detection algorithms is close, and when 14dB to 25dB, the combination property value of MMSE-SIC detection algorithm is minimum.Therefore, it can the detection algorithm that the SNR value according to reception signal selects to be suitable for.

Although above the illustrative detailed description of the invention of the present invention being described; so that those skilled in the art understand the present invention; it is to be understood that; the invention is not restricted to the scope of detailed description of the invention; to those skilled in the art; as long as various changes limit and in the spirit and scope of the present invention determined, these changes are apparent from, and all utilize the innovation and creation of present inventive concept all at the row of protection in appended claim.

Claims (3)

1. the MIMO of a mixed self-adapting receives detection method, it is characterised in that comprise the following steps:

S1: arrange alternative detection algorithm, adopts every kind of detection algorithm training signal training to obtain the mapping table of different signal to noise ratio snr and detection algorithm wrong bitrate in advance, and obtains the complexity value C of every kind of detection algorithm_i, wherein i is the sequence number of detection algorithm；

S2: current Received Signal is carried out signal-to-noise ratio (SNR) estimation and obtains SNR, obtains every kind of detection algorithm wrong bitrate P under the SNR that this estimation obtains according to the mapping table of the SNR being previously obtained in step S1 Yu wrong bitrate_i, calculate the combination property V of every kind of detection algorithm_i, computing formula is: V_i=α * P_i+β*C_i, wherein α, β are the constant parameter more than 0, and alpha+beta=1；

The combination property V of S3: the comparison step S2 every kind of detection algorithm obtained_i, select V_iCurrent Received Signal is detected by the minimum detection algorithm of value.

2. MIMO according to claim 1 receives detection method, it is characterised in that described alternative detection algorithm includes squeeze theorem algorithm, least mean-square error-interference cancellation algorithm, maximum likelihood algorithm.

3. MIMO according to claim 1 receives detection method, it is characterised in that the complexity value of described detection algorithm is algorithm execution time.