CN117278370A

CN117278370A - Adaptive MMSE (minimum mean Square error) equalization method and device based on noise and interference frequency domain distribution characteristics

Info

Publication number: CN117278370A
Application number: CN202311380335.6A
Authority: CN
Inventors: 方锐; 张继栋; 王伟; 周继华; 张帅; 赵涛
Original assignee: Aerospace Xintong Technology Co ltd
Current assignee: Aerospace Xintong Technology Co ltd
Priority date: 2023-10-23
Filing date: 2023-10-23
Publication date: 2023-12-22

Abstract

The invention relates to the technical field of Z, in particular to a self-adaptive MMSE equalization method and device based on noise and interference frequency domain distribution characteristics, wherein the method comprises the following steps: step1, calculating an Rnn matrix of each resource block RB; step2, respectively acquiring NI energy distribution of each resource block RB through an Rnn matrix of the resource block RB; step3, analyzing the difference condition of the NI energy among the resource blocks RB based on the NI energy distribution statistics of the resource blocks RB, and carrying out pull-flush uniform processing on the Rnn matrix according to the difference condition; step4, judging whether the non-diagonal elements of the Rnn matrix can be cleared according to the ratio of the autocorrelation coefficient and the cross correlation coefficient in the Rnn matrix by the Rnn matrix subjected to pull flush uniform treatment, and carrying out self-adaptive clear treatment on the non-diagonal elements of the Rnn matrix according to a judging result; step5, calculating MMSE balance weights for each resource block RB according to the Rnn matrix after the self-adaptive zero clearing processing. The invention can lead the MMSE equalization receiver to exert the best performance from the aspect of the interference noise distribution characteristic in the frequency domain range.

Description

Adaptive MMSE (minimum mean Square error) equalization method and device based on noise and interference frequency domain distribution characteristics

Technical Field

The invention relates to the technical field of Z, in particular to a self-adaptive MMSE equalization method and device based on noise and interference frequency domain distribution characteristics.

Background

A minimum mean square error (MMSE, minimum Mean Square Error) equalization technique, which is a common equalization technique in wireless communication systems, filters a received signal to eliminate interference and noise in the received signal by using a principle of minimizing a mean square error between the received signal and an original signal.

The received signal is: y=h×x+ni, where H is a wireless transmission channel, X is an original transmission signal, N in NI is a reception noise of the receiver, and I is external interference received by the receiver. The transmit signal estimated by the receiver is:wherein W is an equalization weight of the receiver, and the receiver multiplies the received signal by the equalization weight, thereby eliminating noise and interference in the received signal. The equalization weights of the MMSE receiver are: />The original transmission signal is transmitted through a wireless channel, and the receiving side performs MMSE equalization on the received signal to recover the original transmission signal, and the schematic diagram is shown in fig. 1. The Rnn matrix in the MMSE weight is a noise interference correlation matrix between antennas, and is specifically shown in fig. 2.

The main diagonal element of Rnn is the noise interference autocorrelation coefficient of each receiving antenna, and can be regarded as the energy of the noise and the interference of each receiving antenna. The non-main diagonal elements of Rnn are cross-correlation coefficients of noise interference between different receiving antennas, and as the noise is gaussian independent distribution satisfying low correlation, the cross-correlation of the noise between the antennas is close to zero, so the non-main diagonal elements only leave cross-correlation coefficients of interference between different receiving antennas.

In interference limited systems, rnn is not a major diagonal element and the MMSE equalization technique is called interference rejection combining (IRC, interference Rejection Combining). In noise limited systems, rnn non-main diagonal elements can be ignored, and the MMSE equalization technique is called maximum ratio combining (MRC, maximum Ratio Combining).

The conventional IRC and MRC adaptive technology only carries out IRC and MRC adaptive switching based on interference noise ratio, and cannot be from the aspect of interference noise distribution characteristics in a frequency domain range, so that the optimal performance of an MMSE equalization receiver is difficult to develop. Therefore, there is a need for an Rnn matrix adaptive technique suitable for various frequency domain noise interference distribution characteristics, to achieve optimal performance of an MMSE equalizing receiver.

Disclosure of Invention

One of the purposes of the present invention is to provide an adaptive MMSE equalization method based on noise and interference frequency domain distribution characteristics, which can enable an MMSE equalization receiver to perform best from the viewpoint of interference noise distribution characteristics in the frequency domain.

In order to achieve the above object, an adaptive MMSE equalization method based on noise and interference frequency domain distribution characteristics is provided,

the method comprises the following steps:

step1, calculating an Rnn matrix of each resource block RB;

step2, respectively acquiring NI energy distribution of each resource block RB through an Rnn matrix of each resource block RB;

step3, analyzing the difference condition of NI energy among the resource blocks RB based on NI energy distribution statistics of the resource blocks RB, and carrying out pull-flush uniform processing on an Rnn matrix of each resource block RB according to the difference condition;

step4, judging whether the non-diagonal elements of the Rnn matrix can be cleared according to the ratio of the autocorrelation coefficient and the cross correlation coefficient in the Rnn matrix by the Rnn matrix subjected to pull flush uniform treatment, and carrying out self-adaptive clear treatment on the non-diagonal elements of the Rnn matrix according to a judging result;

step5, calculating MMSE balance weight values for each resource block RB according to the Rnn matrix after the self-adaptive zero clearing process

Further, the Rnn matrix is denoted as Rnn _rb The expression is as follows:

wherein antN represents an antenna index, N is a positive integer, RB represents an RB index;

in Step2, the NI energy distribution of each resource block RB is calculated as follows:

for matrix Rnn _rb Summing the main diagonal elements of the resource block RB to obtain the NI energy of the resource block RB, denoted NI _rb The formula is as follows:

further, the Step3 specifically includes the following steps:

step3-1, and counting NI among RBs of each resource block _rb The existing difference, the method includes calculating NI between RBs _rb Standard deviation of (2);

step3-2 based on NI between RBs of each resource block _rb Judging whether the Rnn matrix of each resource block RB needs to be pulled flush or not under the existing difference condition: the judging mode is as follows:

if NI _rb The standard deviation is smaller than the frequency domain NI difference threshold FreNIDIfThr, and the standard deviation is not required to be processed by pulling and flushing;

if NI _rb And if the standard deviation is larger than a frequency domain NI difference threshold FreNIDIFIffThr, carrying out pull flush equalization processing on the Rnn matrix of the resource block RB, wherein the pull flush equalization processing method comprises the step of averaging to obtain a broadband level Rnn matrix.

Further, the broadband level Rnn matrix is:

further, the Step4 specifically includes the following steps:

step4-1, if the Rnn matrix of the resource block RB is pulled flush in the Step Setp3, executing steps 4-3, step4-4 and Step4-5 to perform self-adaptive zero clearing treatment on the broadband Rnn matrix;

step4-2, if the Rnn matrix of the resource block RB is not subjected to pull flush processing in the Step Setp3, executing the steps of Step4-3, step4-4 and Step4-5 to respectively perform self-adaptive zero clearing processing on the Rnn matrix of the resource block RB;

step4-3, calculating a receiving antenna autocorrelation coefficient, and representing energy summation of main diagonal elements of the Rnn matrix;

step4-4, calculating cross correlation coefficients of the receiving antennas, and representing energy summation of non-main diagonal elements of the Rnn matrix;

step4-5, ratio E according to the antenna autocorrelation coefficient and cross correlation coefficient _diag /E _off Judging whether the Rnn off-diagonal line element is cleared;

if E _diag /E _off <DiaEleRatioThr proves that interference exists between the autocorrelation coefficient and the cross correlation coefficient, the off-diagonal elements are not cleared, and the Rnn matrix after processing is as follows:

otherwise, the interference between the autocorrelation coefficient and the cross correlation coefficient is proved to be absent, the off-diagonal line element is cleared, and the Rnn matrix after processing is as follows:

wherein DiaEleRatioThr is expressed as: diagonal element to non-diagonal element ratio threshold.

Further, the Rnn matrix for calculating the MMSE equalization weight in Step5 is generated according to steps 1-4, and specifically includes:

case1-1: when the full bandwidth NI is not different and has no interference, adopting the pulled-up broadband level Rnn and zero clearing of non-diagonal line elements;

case1-2: when the full bandwidth NI is not different and has interference, adopting the leveled broadband level Rnn and the non-diagonal line elements are not cleared;

case2-1: when the full bandwidth NI is different and has no interference, adopting an unordered RB level Rnn and zero clearing of non-diagonal line elements;

case2-2: when the full bandwidth NI is different and has interference, the unordered RB level Rnn is adopted and the off-diagonal elements are not cleared.

Further, the method also comprises the following steps:

step6, merging and simplifying the Case in the Step 5; the simplification is as follows:

case1: when the full bandwidth NI is not different, adopting the leveled broadband grade Rnn;

case2: when the full bandwidth NI is different, the unordered RB level Rnn is adopted and the off-diagonal line elements are self-adaptively cleared.

The second objective of the present invention is to provide an adaptive MMSE equalizing device based on the noise and interference frequency domain distribution characteristics, which uses the adaptive MMSE equalizing method based on the noise and interference frequency domain distribution characteristics.

Principle and advantage:

1. combining the noise and interference distribution characteristics of the frequency domain dimension, calculating an inter-antenna noise interference correlation matrix with optimal MMSE, and improving the performance of signal equalization demodulation;

2. by summarizing four kinds of scenes through the noise and interference distribution characteristics of the frequency domain dimension, different MMSE equalization strategies are adapted under different scenes based on the traditional MMSE equalization algorithm, so that an equalization matrix based on MMSE calculation can better retain original signals, and interference and noise are eliminated as much as possible.

Drawings

Fig. 1 is a schematic diagram of an original transmission signal transmitted through a wireless channel in the prior art, and a receiving side performs MMSE equalization on a received signal to recover the original transmission signal;

fig. 2 is a schematic diagram of noise interference correlation matrix between Rnn matrix and antenna in MMSE weight in the prior art;

fig. 3 is a flowchart of an adaptive MMSE equalization method based on noise and interference frequency domain distribution characteristics according to an embodiment of the present invention;

FIG. 4 is a plot of BER-SNR for the 4 different Rnn matrix correspondence schemes of example 1;

FIG. 5 is a plot of BER versus IOT for the 4 different Rnn matrix correspondence schemes of example 1;

FIG. 6 is a plot of BER-SNR for the 4 different Rnn matrix correspondence schemes of example 2;

FIG. 7 is a plot of BER versus IOT for the 4 different Rnn matrix correspondence schemes of example 2;

fig. 8 is a schematic diagram of performance gain ordering of various equalization algorithms for different NI distribution characteristics.

Detailed Description

The following is a further detailed description of the embodiments:

abbreviation interpretation: MMSE (Minimum Mean Square Error) minimum mean square error, RB (Resource Block) resource blocks, BER (Block Error Ratio) block error rate, SNR (Signal to Noise Ratio) signal to noise ratio, IOT (Interference over Thermal) interference thermal rise.

Examples

An adaptive MMSE equalization method based on noise and interference frequency domain distribution characteristics, LTE and NR belong to a multi-carrier transmission mode, data is jointly transmitted by a plurality of RBs on a frequency domain, a plurality of antennas of a receiver receive the received data of the RBs, the modeling of the receiver with the dimensions of the RBs and the antennas is as follows,

Y _antn,rb ＝H _antn,rb ×X _rb +NI _antn,rb

where antN represents the antenna index, RB represents the RB index;

MMSE equalization weights are calculated for each RB:

the Rnn matrix is determined based on the noise and interference frequency domain distribution characteristics, basically as shown in fig. 3, and the principle steps are as follows:

step1, calculating an Rnn matrix of each resource block RB, wherein the Rnn matrix refers to an RB-level Rnn matrix and is recorded as Rnn _rb The expression is as follows:

the Step3 specifically includes the following steps:

if NI _rb The standard deviation is smaller than the frequency domain NI difference threshold FreNIDIffThr, which means that the NI distribution characteristics are more uniform, and the processing of pulling and flushing is not needed;

if NI _rb Standard deviation ofWhen the frequency domain NI difference threshold FreNIDIFIffThr is larger than the frequency domain NI difference threshold FreNIDIffThr, indicating that the NI distribution characteristics are uneven, carrying out pull flush equalization on an Rnn matrix of a resource block RB, wherein the method for the pull flush equalization comprises the steps of averaging to obtain a broadband level Rnn matrix.

The broadband level Rnn matrix is:

the Step4 specifically includes the following steps:

step4-5, ratio E according to the antenna autocorrelation coefficient and cross correlation coefficient _diag /E _off Judging RWhether the nn off-diagonal element is clear;

wherein, NI _ant1,rb ·NI _ant1,rb Equivalent to Rnn _ant1,ant1

Step5, calculating MMSE balance weights for each resource block RB according to the Rnn matrix after the self-adaptive zero clearing processing;

the Rnn matrix for calculating the MMSE equalization weight in Step5 is generated according to steps 1-4, and specifically includes:

Step6, merging and simplifying the Case in the Step 5; and when the Case1-1 full bandwidth NI is not different and has no interference, the non-diagonal line elements in the Rnn matrix are cross correlation coefficients of noise among antennas, and the independent Gaussian distribution with the average value of 0 is obeyed. If the number of received RBs is large, according to the big number theorem, the average value of the 0 Gaussian distribution of the plurality of average values tends to 0, so that the off-diagonal line elements of the broadband level Rnn are automatically cleared after being pulled and averaged. Therefore, to reduce the amount of computation, case1-1 is combined with Case 1-2;

in addition, case2-1 and Case2-2, because of the difference of full bandwidth NI, all adopt the RB grade Rnn that does not pull out, and according to NI distribution situation, clear the off-diagonal line element adaptively according to step 4. The simplification is as follows:

Generating a 5G uplink transmission block with RB=100\MCS=4, and comparing error rates under different SNR\IOT through radio link simulation to verify the correctness and performance benefits of the flow of the technology. The key simulation parameters of the uplink transmission block are configured as the following table:

simulation parameter configuration table:

simulation parameters	Configuration of
		Subcarrier spacing	15kHz
Carrier center frequency point	2.1G
		MCS	4
RB	100
		Number of transmitting antennas	2
Number of receiving antennas	2
		Radio channel characteristics	TDLC Low

The 5G uplink transmission block respectively performs flow and performance verification in 4 types of scenes including the Case1-1, the Case1-2, the Case2-2 and the Case2-2, and performs performance verification on schemes of 4 different Rnn matrixes under each scene. The 4 different Rnn matrices are:

rnneangirc: the leveled broadband grade Rnn; broadband level IRC-Rnn matrix for short;

rnnnnotavgirc: the un-pulled RB level Rnn and the off-diagonal elements are not cleared; an RB-level IRC-Rnn matrix for short;

rnnnnotavgmrc: the un-pulled RB level Rnn and the off-diagonal element are cleared; an RB-level MRC-Rnn matrix for short;

rnnnnotavgadpt: the un-leveled RB level Rnn and the off-diagonal elements are adaptively cleared. Abbreviated as RB-level Adpt-Rnn matrix.

Example 1 corresponds to Case1 above, provided that: full bandwidth NI is not differential; the process is performed according to the above-described flow steps as follows:

step1, calculating an RB-level Rnn matrix;

step2, acquiring RB-level NI distribution through an RB-level Rnn matrix;

step3, carrying out a pull-in analysis processing on the Rnn matrix based on the RB-level NI distribution characteristic. In the scene of no difference of full bandwidth NI, the NI distribution characteristic is uniform, the NI standard deviation on each RB is smaller than the frequency domain NI difference threshold FreNIDIffThr, and the RB level Rnn matrix is subjected to pull-flush uniform processing to generate a broadband level Rnn matrix;

step4, because the matrix is a broadband Rnn matrix, the simplified flow is carried out, and the off-diagonal element self-adaptive zero clearing flow is not carried out;

step5, the Rnn matrix in the final MMSE equalization weight is rnnAvgIRC.

In the full-bandwidth NI non-differential scenario, the BER of 4 different schemes for generating Rnn matrix under different SNR under this scenario is shown in fig. 4.

Different IOT's are set to increase different interference levels, and the BER of 4 different schemes for generating Rnn matrices under this scenario is shown in fig. 5.

In summary, in a full-bandwidth NI non-differential scene, whether interference exists or not, the BER of the scene of RnnAvgIRC under the same SNR or IOT is the lowest, and the performance of the scene of RnnAvgIRC is proved to be optimal, so that the scene meets the scheme expectation.

Example 2 corresponds to Case2 above, provided that: full bandwidth NI varies; the process is performed according to the above-described flow steps as follows:

step1, calculating an RB-level Rnn matrix;

step2, acquiring RB-level NI distribution through an RB-level Rnn matrix;

step3, carrying out a pull-in analysis processing on the Rnn matrix based on the RB-level NI distribution characteristic. In the scene of difference of full bandwidth NI, the NI distribution characteristics are uneven, the NI standard deviation on each RB is larger than the frequency domain NI difference threshold FreNIDIffThr, the RB grade Rnn matrix is not subjected to pull flush uniform processing, and the RB grade Rnn matrix is continuously used;

step4, because the RB-level Rnn matrix is adopted currently, a non-diagonal element self-adaptive zero clearing process is needed;

step5, the Rnn matrix in the final MMSE equalization weight is rnnnNotAvgAdpt.

And setting the noise of the first 50 RBs of the 5G uplink transmission block unchanged, and lifting the noise of the last 50 RBs by 4dB to generate a scene with different full bandwidth NI and no interference. Because no interference exists, the ratio of the main diagonal energy to the non-main diagonal energy of the RB-level Rnn matrix in Step4 is larger than the threshold value FrenidiffThr, the non-main diagonal element of the Rnn matrix is cleared, and the RnnNotAvgAdpt is close to RnnNotAvgMRC. BER of 4 different schemes for generating the Rnn matrix under different SNR in the scene is shown in figure 6;

the interference of the first 50 RBs of the 5G uplink transmission block is set to be unchanged, and the interference of the last 50 RBs is raised by 4dB to generate a scene with difference of full bandwidth NI and interference. Because of the existence of interference, the ratio of the main diagonal energy to the non-main diagonal energy of the RB-level Rnn matrix in Step4 is larger than the threshold value FreNIDiffThr, RB, and the non-main diagonal element of the RnnNOAvgAdpt is not cleared, so that the RnnNotAvgIRC is close to the RnnNotAvgAdpt. Different IOT's are set to increase different interference levels, and the BER of 4 different schemes for generating Rnn matrices under this scenario is shown in fig. 7.

In summary, in a full bandwidth NI differentiated scenario, rnnnnotavgadpt would tend to the performance of rnnnnotavgmrc when there is no interference, and rnnnnotavgadpt would tend to the performance of rnnnnotavgirc when there is interference. The scene has different SNR or lowest BER of the RnnNotAvgAdpt under the IOT, and the performance of the RnnNotAvgAdpt under the scene is proved to be optimal, so that the scene accords with the scheme expectation.

The flow and performance verification result arrangement of the above two examples is shown in fig. 8:

when NI is uniformly distributed, performance without interference is expressed as: broadband level IRC > RB level MRC > RB level Adpt > RB level IRC;

when NI is uniformly distributed, in the presence of interference, the performance is expressed as: broadband level IRC > RB level adpt=rb level IRC > RB level MRC;

when NI is unevenly distributed, the performance without interference is expressed as: RB level Adpt is more than or equal to RB level MRC > RB level IRC > broadband level IRC;

when NI is unevenly distributed, with interference, the performance is expressed as: RB level Adpt > RB level MRC > broadband level IRC > RB level IRC.

The foregoing is merely an embodiment of the present invention, and general knowledge of specific structures and features well known in schemes is not described in any way herein, so that a person of ordinary skill in the art would know all of the prior art to which the present invention pertains before the application date or priority date, and would be able to learn all of the prior art in this field, and have the ability to apply conventional experimental means before this date, so that a person of ordinary skill in the art could complete and implement this scheme in combination with his own capabilities, given the benefit of this application, and some typical known structures or known methods should not be an obstacle to the implementation of this application by those of ordinary skill in the art. It should be noted that modifications and improvements can be made by those skilled in the art without departing from the structure of the present invention, and these should also be considered as the scope of the present invention, which does not affect the effect of the implementation of the present invention and the utility of the patent. The protection scope of the present application shall be subject to the content of the claims, and the description of the specific embodiments and the like in the specification can be used for explaining the content of the claims.

Claims

1. The adaptive MMSE equalization method based on the noise and interference frequency domain distribution characteristics is characterized by comprising the following steps of:

step1, calculating an Rnn matrix of each resource block RB;

step5, calculating MMSE balance weights for each resource block RB according to the Rnn matrix after the self-adaptive zero clearing processing.

2. The adaptive MMSE equalizing method based on the noise-interference frequency domain distribution characteristic as claimed in claim 1, wherein: the Rnn matrix is denoted as Rnn _rb The expression is as follows:

3. the adaptive MMSE equalizing method based on the noise-interference frequency domain distribution characteristic as claimed in claim 2, wherein: the Step3 specifically includes the following steps:

step3-1, and counting NI among RBs of each resource block _rb The existing difference, the method includes calculating the standard deviation of NIrb between RBs;

4. The adaptive MMSE equalizing method based on noise-interference frequency domain distribution characteristics as claimed in claim 3, wherein: the broadband level Rnn matrix is:

5. the adaptive MMSE equalizing method based on noise-interference frequency domain distribution characteristics as claimed in claim 4, wherein: the Step4 specifically includes the following steps:

if E _diag /E _off <DiaEleRatioThr demonstrates autocorrelation coefficients and correlationsThe numbers have interference, the non-diagonal line elements are not cleared, and the Rnn matrix after processing is as follows:

6. The adaptive MMSE equalizing method based on noise-interference frequency domain distribution characteristics of claim 5, wherein: the Rnn matrix for calculating the MMSE equalization weight in Step5 is generated according to steps 1-4, and specifically includes:

7. The adaptive MMSE equalizing method based on the noise-and-interference frequency-domain distribution characteristic as claimed in claim 6, further comprising the steps of:

8. An adaptive MMSE equalizing device based on noise and interference frequency domain distribution characteristics, characterized in that an adaptive MMSE equalizing method based on noise and interference frequency domain distribution characteristics as defined in any one of claims 1-7 is applied.