CN109150775B - Robust online channel state estimation method for dynamic change of self-adaptive noise environment - Google Patents

Robust online channel state estimation method for dynamic change of self-adaptive noise environment Download PDF

Info

Publication number
CN109150775B
CN109150775B CN201810919313.5A CN201810919313A CN109150775B CN 109150775 B CN109150775 B CN 109150775B CN 201810919313 A CN201810919313 A CN 201810919313A CN 109150775 B CN109150775 B CN 109150775B
Authority
CN
China
Prior art keywords
channel state
channel
model
state information
period
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810919313.5A
Other languages
Chinese (zh)
Other versions
CN109150775A (en
Inventor
徐宗本
薛江
孟德宇
邓芸
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Jiaotong University
Original Assignee
Xian Jiaotong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian Jiaotong University filed Critical Xian Jiaotong University
Priority to CN201810919313.5A priority Critical patent/CN109150775B/en
Publication of CN109150775A publication Critical patent/CN109150775A/en
Application granted granted Critical
Publication of CN109150775B publication Critical patent/CN109150775B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00—Baseband systems
    • H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/0202—Channel estimation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B7/00—Radio transmission systems, i.e. using radiation field
    • H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
    • H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
    • H04B7/0413—MIMO systems
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00—Baseband systems
    • H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/0202—Channel estimation
    • H04L25/024—Channel estimation channel estimation algorithms
    • H04L25/0242—Channel estimation channel estimation algorithms using matrix methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Power Engineering (AREA)
  • Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Monitoring And Testing Of Transmission In General (AREA)
  • Mobile Radio Communication Systems (AREA)
  • Noise Elimination (AREA)
  • Radio Transmission System (AREA)

Abstract

The invention relates to a robustness online channel state information estimation method for dynamic change of a self-adaptive noise environment, which comprises the following steps: constructing an online machine learning model of dynamic noise estimation based on the characteristic of real-time change of a communication noise environment; embedding a channel change regular pattern in the model based on the continuous change characteristic of the channel, constructing a complete statistical model, and obtaining a complete online channel estimation machine learning model according to a maximum posterior estimation method; and storing the environmental noise distribution parameters and the channel state information of the last time period by utilizing the storage equipment of the base station, and combining an online channel estimation model to obtain high-precision channel state information estimation in real time. The method is based on the machine learning principle, realizes the on-line channel state information estimation method which is rapid, high in precision and capable of adapting to the noise environment, and has important significance for reducing communication delay, reducing pilot signal use and improving information transmission rate in practical application.

Description

Robust online channel state estimation method for dynamic change of self-adaptive noise environment
Technical Field
The invention relates to a wireless communication channel estimation method, in particular to a robustness online channel state information estimation method adaptive to dynamic changes of a noise environment.
Background
Wireless communication is one of the most active areas of scientific and technological development, and is also a branch of the rapid development of the field of communication systems. Since the transmitted signal interacts with the environment in a very complex manner, changes in the wireless communication channel state information have some adverse effect on the transmitted signal. One of the main challenges in achieving optimal performance in a wireless communication system is to provide accurate channel state information at the receiving end of the system, i.e., to perform accurate channel state information estimation.
Accurate estimation of the channel state information is the basis for improving the technical accuracy of signal reconstruction, signal source detection, signal number detection and the like. Accurate estimation of channel state information in real time is still a great challenge on the premise of ensuring high precision and high efficiency.
In the field of wireless communication, there are a number of techniques for channel state information estimation. Common techniques include a channel estimation method based on least squares, a channel estimation method based on minimum mean square error, a channel estimation method based on singular value decomposition, and the like.
The channel estimation method based on least square and the channel estimation method based on minimum mean square error assume that the noise distribution of a signal system is Gaussian white noise and is not consistent with the actual complex noise distribution. In addition, the MoG and MoPE methods consider a more refined statistical distribution hypothesis for the communication environment, and use some mixed distributions (such as gaussian mixtures) to fit the environmental noise, thereby obtaining a better channel estimation effect.
Although the prior method has achieved remarkable effect, there is a certain gap from the actual application. In the current society, signal data of wireless communication is rapidly increased every moment, and the channel state information estimation technology is required to have high efficiency on the premise of ensuring high precision; on the other hand, in the face of signal data which is constantly emerging at all times, a real-time online channel state information estimation technology needs to be provided. In addition, in a real communication environment, there is complex noise that dynamically changes in real time continuously, and it is also an important issue to need to be studied urgently to improve the robustness of a channel estimation method to such a real noise environment. At present, an online channel state information estimation method is not available, and multiple requirements of high precision, high efficiency and robustness can be met simultaneously.
Disclosure of Invention
Aiming at the defects of the prior art, the invention aims to provide a robustness online channel state information estimation method capable of adapting to the dynamic change of a noise environment, which can more fully and accurately utilize communication knowledge prior information and noise environment change information thereof to carry out full statistical modeling, thereby achieving the channel state information estimation effect with higher precision and more robustness, ensuring the high efficiency of processing, and effectively adapting to the change of a communication environment and a channel.
In order to achieve the purpose, the invention adopts the technical scheme that:
a robustness online channel state information estimation method for self-adaptive noise environment dynamic change comprises the following specific steps:
1) carrying out parametric distribution modeling based on randomness of communication noise environment change, enabling the channel state information estimation model to be adaptive to dynamic changes of the communication noise environment in different time and different scenes, and further carrying out relevance constraint coding on noise information of the channel state information estimation model embedded into the communication noise environment information of the previous time period to realize adaptive modeling of the dynamic communication noise environment;
2) modeling is carried out based on the continuous change characteristic of the channel, and relevance information coding of the channel state information in the last period of time is embedded into the channel state information estimation model in the step 1), so that self-adaptive modeling of the channel is realized; constructing a complete statistical model by combining the step 1), and establishing a complete online channel state information estimation machine learning model according to a maximum posterior estimation method;
3) and (3) storing the noise environment distribution parameters and the channel state information of the last period of time by using a base station storage device, and obtaining high-precision channel state information estimation by combining the online channel state information estimation machine learning model in the step 2).
The invention establishes the inherent statistical prior based on the wireless communication noise environment and the channel, respectively carries out the pertinence analysis and the coding, and can realize the online channel state information estimation with high speed, high precision and robustness. The accurate and efficient channel state information estimation has important application significance for improving signal reconstruction, signal source detection, signal number detection and the like.
Drawings
FIG. 1 is a flow chart of the present invention.
FIG. 2 is a comparison chart of channel estimation accuracy using a least squares channel estimation method, a hybrid Gaussian based off-line channel estimation method, and a proposed machine learning on-line channel estimation method in a dynamic noise environment and channel state information; it can be seen that under the dynamic noise environment and the channel state information, the channel precision obtained by the method is obviously higher than that of the classical LS algorithm and the off-line algorithm based on the mixed Gaussian, which shows that the method can adaptively learn the dynamic noise environment and the channel state.
FIG. 3 is a graph comparing the time required for channel estimation for a hybrid Gaussian-based off-line channel estimation method and a proposed machine-learned on-line channel estimation method at different sample sizes; it can be seen that the proposed method can significantly reduce the time required for channel estimation, which is an order of magnitude less as the number of training samples increases.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples.
Referring to fig. 1, the method for estimating robust online channel state information of dynamic changes of a self-adaptive noise environment of the present invention includes the following specific steps:
1) carrying out parametric distribution modeling based on randomness of communication noise environment change, enabling the channel state information estimation model to be adaptive to dynamic changes of the communication noise environment in different time and different scenes, and further carrying out relevance constraint coding on noise information of the channel state information estimation model embedded into the communication noise environment information of the previous time period to realize adaptive modeling of the dynamic communication noise environment; the method specifically comprises the following substeps:
a) acquiring a pilot signal and received signal data thereof;
b) according to the communication principle, for a multiple-input multiple-output (MIMO) system with s transmit antennas and r receive antennas, the channel model can be expressed as:
Yt＝HtXt+Et(18)
wherein
Figure GDA0002357290760000041
A matrix of pilot signals transmitted during the t-th period,
Figure GDA0002357290760000042
is a matrix of signals received by the receiving end during the t-th period,
Figure GDA0002357290760000043
for the channel matrix to be estimated for the t-th period,
Figure GDA0002357290760000044
representing the ambient noise matrix, where d represents the t-th period signalThe number of samples of (a); for the convenience of algorithm design, the model is changed from a complex domain to a real domain:
Figure GDA0002357290760000045
wherein
Figure GDA0002357290760000046
Where Re (·), Im (·) represent the real and imaginary parts, respectively. So that the result in the real domain can be transformed into the complex domain. After adopting
Figure GDA0002357290760000047
And (4) showing.
c) Based on the assumption of mixed Gaussian distribution of noise, there are
Figure GDA0002357290760000048
Wherein
Figure GDA0002357290760000049
Is YtThe number of the ith, jth element of (a),
Figure GDA00023572907600000410
is HtThe number of the ith row of (a),
Figure GDA00023572907600000411
is XtThe (c) th column of (a),
Figure GDA0002357290760000051
a hidden variable is represented by a number of hidden variables,
Figure GDA0002357290760000052
the ith, j-th element representing the t-th period does not belong to the kth mixture component in the mixture gaussian distribution,
Figure GDA0002357290760000053
represents the t-th periodBelongs to the kth mixture component in the mixture Gaussian distribution and satisfies
Figure GDA0002357290760000054
Multi represents a distribution of polynomials,
Figure GDA0002357290760000055
is the variance of the k-th mixture component,
Figure GDA0002357290760000056
the K is the mean value of the K mixed components, K represents the number of the mixed components, N (-) represents normal distribution, and T represents matrix transposition;
in order to embed regularization noise information coding of communication noise environment information in the previous period and realize self-adaptive modeling of a real communication dynamic noise environment, conjugate prior form hypothesis is respectively carried out on a noise distribution variable in a model (19):
Figure GDA0002357290760000057
Figure GDA0002357290760000058
here, the
Figure GDA0002357290760000059
Representing a gaussian-inverse Gamma joint distribution, Dir representing a Dirichlet distribution,
Figure GDA00023572907600000510
wherein N ist -1，
Figure GDA00023572907600000511
In order to simplify the intermediate variables of the formula,
Figure GDA00023572907600000512
is the ratio of the kth mixed gaussian component during the t-1 period,
Figure GDA00023572907600000513
the expression of (a) is as follows:
Figure GDA00023572907600000514
where P (-) represents the probability distribution,
Figure GDA00023572907600000515
representing the degree of membership representing the degree to which the ith, j-th element of the t-th period belongs to the kth mixture component in the mixture gaussian distribution,
Figure GDA00023572907600000516
the meaning of (2) is the same as in (20).
2) Modeling is carried out based on the continuous change characteristic of the channel, and relevance information coding of the channel state information in the last period of time is embedded into the channel state information estimation model in the step 1), so that self-adaptive modeling of the channel is realized; constructing a complete statistical model by combining the step 1), and establishing a complete online channel state information estimation machine learning model according to a maximum posterior estimation method; the method specifically comprises the following substeps:
a) according to the gaussian distribution assumption for the channel, there are:
Figure GDA0002357290760000061
where ρ is the channel variation coefficient, A is the hyperparameter,
Figure GDA0002357290760000062
is a semi-positive definite matrix.
b) Combining the step 1) and the maximum posterior estimation method, the following channel state information estimation optimization problem can be obtained:
Figure GDA0002357290760000063
the simplification is as follows:
Figure GDA0002357290760000064
wherein DKL(. I. represents KL divergence, RE(Πt,∑t,Mt) Is a noise regularization term of the form:
Figure GDA0002357290760000065
RH(Ht) Is a channel regularization term of the form:
Figure GDA0002357290760000066
here, the
Figure GDA0002357290760000067
Is the residual E of the t-th periodtThe mixture coefficient of the mixture Gaussian distribution is contained,
Figure GDA0002357290760000068
representing the variance, M, of each mixture component of the mixed Gaussian distributiont＝{μ1,μ2,…,μKT-th time interval mixing Gaussian distribution average value of each mixing component,
Figure GDA0002357290760000071
is to simplify the intermediate variable, Π, defined by the formulat-1，∑t-1，Mt-1Representing the corresponding mixing coefficient, variance vector and mean vector at time t-1, C representing nt,∑t,MtAn independent constant.
3) And (2) storing the noise environment distribution parameters and the channel state information of the previous period by utilizing a storage device of the base station, and obtaining high-precision channel state information estimation by combining an online channel state information estimation machine learning model of the step 2) based on the pilot signal data and the received signal data input in the step 1). The method specifically comprises the following substeps:
a) various parameters were initialized for the first 200 sample data using MoG algorithm 1 as follows:
algorithm 1: MoG initialization algorithm
Inputting: first 200 pilot data and corresponding signal data
And (3) outputting: h0,M0,Π0,∑0
Step 1, initializing II, Sigma, M by random sampling on [0,1], and initializing H by random
Step 2, repeat:
e-step:
Figure GDA0002357290760000072
m-step:
Figure GDA0002357290760000073
Figure GDA0002357290760000074
Figure GDA0002357290760000075
Figure GDA0002357290760000076
Figure GDA0002357290760000081
step 3. unitil conversion
b) Fixed H ═ Ht-1Updating pi using EM algorithmt,∑t,MtUsing the mixing coefficient, variance vector and mean vector information pi of the mixed gaussians in the last time period pre-stored in the base station for the initial iteration data in each time periodt,0,∑t,0,Mt,0The iteration (superscript s denotes the number of iterations) format is as follows:
e-step:
Figure GDA0002357290760000082
1≤i≤r,1≤k≤d (26)
m-step:
Figure GDA0002357290760000083
Figure GDA0002357290760000084
Figure GDA0002357290760000085
wherein:
Figure GDA0002357290760000086
Figure GDA0002357290760000087
Figure GDA0002357290760000088
Figure GDA0002357290760000089
wherein
Figure GDA0002357290760000091
Nt,s，
Figure GDA0002357290760000092
Is to simplify the intermediate variables defined by the formula.
c) And (3) iteration termination conditions:
Figure GDA0002357290760000093
II thereint,s+1,∑t,s+1,Mt,s+1The mixing coefficient, variance vector and mean vector, pi, representing the corresponding s +1 th iterative mixing gaussians at time tt,s,∑t,s,Mt,sAnd (3) a mixing coefficient, a variance vector and a mean vector which represent the corresponding mixing gaussians iterated for the time t s, wherein L (-) is an objective function defined by the formula (25).
d) Renewing pi according to the above processt,∑t,MtThen, update HtThe specific optimization model is as follows:
Figure GDA0002357290760000094
wherein WtAn indication matrix representing a t-th period,
Figure GDA0002357290760000095
⊙ represents a dot product.
The model (27) has the following explicit solution:
Figure GDA0002357290760000096
wherein
Figure GDA0002357290760000097
Represents WtThe number of the ith row of (a),
Figure GDA0002357290760000098
Figure GDA0002357290760000099
represents HtRow i, a, b are hyper-parameters,
Figure GDA00023572907600000910
and
Figure GDA00023572907600000911
the expression of (a) is as follows:
Figure GDA00023572907600000912
Figure GDA00023572907600000913
Figure GDA0002357290760000101
e) updating relevant information in the base station for the next time period channel estimation;
the information saved includes: optimal channel estimation H in current time period*And optimal environment noise parameter recovery measurement parameter pi*,∑*,M*And realizing the optimal channel state information estimation in the next time period.
In the dynamic noise environment and the channel state information, the least square channel estimation method, the off-line channel estimation method based on the mixed gauss and the on-line channel estimation method of the invention are compared in accuracy, and referring to fig. 2, it can be seen that in the dynamic noise environment and the channel state information, the channel accuracy obtained by the invention is obviously higher than that of the classic LS algorithm and that of the off-line algorithm based on the mixed gauss, and the on-line channel estimation method of the invention can adaptively learn the dynamic noise environment and the channel state.
Comparing the time required for channel estimation of the off-line channel estimation method based on the mixed gauss and the on-line channel estimation method of the present invention on different sample sizes, referring to fig. 3, it can be seen that the present invention can significantly reduce the time required for channel estimation, and as the training sample size increases, the time consumption is reduced by orders of magnitude.

Claims (1)

1. A robust online channel state estimation method for adaptive noise environment dynamic change is characterized by comprising the following steps:
1) carrying out parametric distribution modeling based on randomness of communication noise environment change, enabling the channel state information estimation model to be adaptive to communication noise environment dynamic change under different time and different scenes, embedding the noise information of the communication noise environment information of the last time period into the channel state information estimation model, and carrying out relevance constraint coding on the noise information, so as to realize adaptive modeling of a real communication dynamic communication noise environment;
the step 1) comprises the following substeps:
a) acquiring a pilot signal and receiving signal data;
b) according to the wireless communication principle, for a multiple-input multiple-output (MIMO) system with s transmitting antennas and r receiving antennas, a channel model is expressed as:
Yt＝HtXt+Et(1)
wherein
Figure FDA0002357290750000011
A matrix of pilot signals transmitted during the t-th period,
Figure FDA0002357290750000012
is a matrix of signals received by the receiving end during the t-th period,
Figure FDA0002357290750000013
for the channel matrix to be estimated for the t-th period,
Figure FDA0002357290750000014
representing an ambient noise matrix, d representing the number of samples of the signal during the t-th period; for the convenience of algorithm design, the channel model is changed from a complex domain to a real domain:
Figure FDA0002357290750000015
wherein
Figure FDA0002357290750000016
Figure FDA0002357290750000017
Where Re (-) and Im (-) represent the real and imaginary parts, respectively, to transform the results in the real domain to the complex domain, using
Figure FDA0002357290750000018
Represents;
c) the parameterized distribution modeling of the step 1) is to model a noise variable E in the model (2)tThe coding is mixed Gaussian distribution, so that the coding is adaptive to the dynamic change of the communication environment under different time and different scenes, and the corresponding model is as follows:
Figure FDA0002357290750000021
wherein
Figure FDA0002357290750000022
Is YtThe number of the ith, jth element of (a),
Figure FDA0002357290750000023
is HtThe number of the ith row of (a),
Figure FDA0002357290750000024
is XtThe (c) th column of (a),
Figure FDA0002357290750000025
Figure FDA0002357290750000026
a hidden variable is represented by a number of hidden variables,
Figure FDA0002357290750000027
the ith, j-th element representing the t-th period does not belong to the kth mixture component in the mixture gaussian distribution,
Figure FDA0002357290750000028
the ith, j-th element representing the t-th period belongs to the kth mixture component in the mixture Gaussian distribution, andsatisfy the requirement of
Figure FDA0002357290750000029
Multi represents a distribution of polynomials,
Figure FDA00023572907500000210
is the variance of the k-th mixture component,
Figure FDA00023572907500000211
the K is the mean value of the K mixed components, K represents the number of the mixed components, N (-) represents normal distribution, and T represents matrix transposition;
the noise information of the communication noise environment information of the last period of time embedded into the channel state information estimation model is subjected to relevance constraint coding, so that the self-adaptive modeling of the real communication dynamic noise environment is realized, and the process is as follows: respectively carrying out conjugate prior form hypothesis on noise distribution variables in the model (3):
Figure FDA00023572907500000212
Figure FDA00023572907500000213
wherein
Figure FDA00023572907500000214
Representing a gaussian-inverse Gamma joint distribution, Dir representing a Dirichlet distribution,
Figure FDA00023572907500000215
wherein N ist-1，
Figure FDA00023572907500000216
In order to simplify the intermediate variables of the formula,
Figure FDA00023572907500000217
is the kth mixing high during the t-1 periodThe ratio of the components of the mixture to each other,
Figure FDA00023572907500000218
the expression of (a) is as follows:
Figure FDA00023572907500000219
where P (-) represents the probability distribution,
Figure FDA0002357290750000031
representing the degree of membership representing the degree to which the ith, j-th element of the t-th period belongs to the kth mixture component in the mixture gaussian distribution,
Figure FDA0002357290750000032
the meaning of (A) is the same as in formula (3);
2) modeling is carried out based on the continuous change characteristic of the channel, and relevance information coding of the channel state information in the last period of time is embedded into the channel state information estimation model in the step 1), so that self-adaptive modeling of the channel is realized; constructing a complete statistical model by combining the step 1), and establishing a complete online channel estimation machine learning model according to a maximum posterior estimation method;
the step 2) comprises the following substeps:
a) in order to embed the regularized coding of the channel state information of the last period into the model (3) so that it is adaptive to the dynamic change of the channel due to the small scale effect, the corresponding model is modeled as:
Figure FDA0002357290750000033
where ρ is the channel variation coefficient, A is the hyperparameter,
Figure FDA0002357290750000034
is a semi-positive definite matrix;
b) constructing a statistical model based on the step 1):
Figure FDA0002357290750000035
where P (-) represents the probability distribution,
Figure FDA0002357290750000036
is the residual E of the t-th periodtThe mixture coefficient of the mixture Gaussian distribution is contained,
Figure FDA0002357290750000037
representing the variance, M, of the mixture components of the Gaussian mixture distribution in the t-th periodt＝{μ1,μ2,…,μKT-th time interval mixing Gaussian distribution average value of each mixing component,
Figure FDA0002357290750000038
in order to simplify the intermediate variables defined by the formula,
Figure FDA0002357290750000039
as defined by formula (6);
according to the maximum a posteriori estimation principle, the channel state information estimation model transformed by the statistical model can be transformed into the following optimization problem:
Figure FDA00023572907500000310
Figure FDA0002357290750000041
the simplification is as follows:
Figure FDA0002357290750000042
wherein R isE(∏t,∑t,Mt) Is a noise regularization term of the form:
Figure FDA0002357290750000043
DKL(. I. represents KL divergence, C represents Πt,∑t,MtIndependent constants
RH(Ht) Is a channel regularization term of the form:
Figure FDA0002357290750000044
here, the
Figure FDA0002357290750000045
Is the residual E of the t-th periodtThe mixture coefficient of the mixture Gaussian distribution is contained,
Figure FDA0002357290750000046
representing the variance, M, of each mixture component of the mixed Gaussian distributiont＝{μ1,μ2,…,μKT-th time interval mixing Gaussian distribution average value of each mixing component,
Figure FDA0002357290750000047
Figure FDA0002357290750000048
is to simplify the intermediate variable, Π, defined by the formulat-1，∑t-1，Mt-1Representing the corresponding mixing coefficient, variance vector and mean vector at time t-1, C representing nt,∑t,MtAn independent constant;
3) storing the noise environment distribution parameters and the channel state information in the last period of time by using a storage device of a base station, combining the online channel estimation machine learning model in the step 2), and updating the parameters pi in the online channel state information estimation model by adopting an EM (effective electromagnetic) algorithmt,∑t,MtObtaining the high-precision channel state information estimation specifically comprises the following substeps:
a)giving out the membership degree of the step E in the EM algorithm
Figure FDA0002357290750000051
The update formula of (2):
Figure FDA0002357290750000052
the superscript s in the formula represents the s-th iteration;
b) and (3) giving an iteration format and a termination condition of the M steps in the EM algorithm:
the iteration format is:
Figure FDA0002357290750000053
Figure FDA0002357290750000054
Figure FDA0002357290750000055
wherein:
Figure FDA0002357290750000056
Figure FDA0002357290750000057
Figure FDA0002357290750000058
Figure FDA0002357290750000059
wherein
Figure FDA00023572907500000510
Nt,s，
Figure FDA00023572907500000511
Is an intermediate variable defined to simplify the formula;
the iteration termination condition is as follows:
Figure FDA0002357290750000061
II thereint,s+1,∑t,s+1,Mt,s+1The mixing coefficient, variance vector and mean vector, pi, representing the corresponding s +1 th iterative mixing gaussians at time tt,s,∑t,s,Mt,sRepresenting a mixing coefficient, a variance vector and a mean vector of the corresponding s-time iterative mixed gaussians at the t moment, wherein L (-) is an objective function defined by the formula (8);
c) setting an iteration initial value:
using the mixing coefficient, variance vector and mean vector information pi of the mixed gaussians in the last time period pre-stored in the base station for the initial time period datat,0,∑t,0,Mt,0；
d) Performing iterative operations of the formulas (9) to (12) until the condition of terminating the formula (13) is satisfied;
e) for the data in the t-th time period, updating the parameter IIt,∑t,MtOn the basis of the following model for the channel Ht-1Fine tuning to obtain updated Ht：
Figure FDA0002357290750000062
Wherein WtAn indication matrix representing a t-th period,
Figure FDA0002357290750000063
⊙ represents a dot product;
the model (14) has the following solution:
Figure FDA0002357290750000064
wherein
Figure FDA0002357290750000065
Represents HtRow i, a, b are hyper-parameters,
Figure FDA0002357290750000066
and
Figure FDA0002357290750000067
the expression of (a) is as follows:
Figure FDA0002357290750000068
Figure FDA0002357290750000071
f) updating relevant information in the base station for the estimation of the channel state information in the next time period;
the information saved includes: current time period optimal channel state information estimation H*And optimal environment noise parameter recovery measurement parameter pi*,∑*,M*。
CN201810919313.5A 2018-08-14 2018-08-14 Robust online channel state estimation method for dynamic change of self-adaptive noise environment Active CN109150775B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810919313.5A CN109150775B (en) 2018-08-14 2018-08-14 Robust online channel state estimation method for dynamic change of self-adaptive noise environment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810919313.5A CN109150775B (en) 2018-08-14 2018-08-14 Robust online channel state estimation method for dynamic change of self-adaptive noise environment

Publications (2)

Publication Number Publication Date
CN109150775A CN109150775A (en) 2019-01-04
CN109150775B true CN109150775B (en) 2020-03-17

Family

ID=64793065

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810919313.5A Active CN109150775B (en) 2018-08-14 2018-08-14 Robust online channel state estimation method for dynamic change of self-adaptive noise environment

Country Status (1)

Country Link
CN (1) CN109150775B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4152803A4 (en) * 2020-05-29 2023-06-14 Guangdong Oppo Mobile Telecommunications Corp., Ltd. Channel state information processing method, electronic device, and storage medium

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112422212B (en) * 2020-11-19 2022-12-30 杭州勒贝格智能系统股份有限公司 Data self-adaptive wireless communication channel prediction method, storage medium and equipment
CN114844785B (en) * 2021-02-01 2024-02-06 大唐移动通信设备有限公司 Model updating method, device and storage medium in communication system
CN115604731A (en) * 2021-07-09 2023-01-13 华为技术有限公司（Cn） Communication method and device
WO2023279366A1 (en) * 2021-07-09 2023-01-12 Oppo广东移动通信有限公司 Noise reduction method based on transfer learning, terminal device, network device and storage medium
CN114696921B (en) * 2021-12-23 2023-06-23 清华大学 Calculation method and system for channel information loss discussion
CN116193504B (en) * 2023-04-18 2023-07-21 南京云程半导体有限公司 Channel state information reporting method, electronic equipment and storage medium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106850472A (en) * 2017-04-10 2017-06-13 中山大学 A kind of OFDM channel estimation method based on Kalman and blind estimate
CN108111441A (en) * 2018-01-12 2018-06-01 电子科技大学 Channel estimation methods based on variational Bayesian
CN108259398A (en) * 2018-01-12 2018-07-06 电子科技大学 The channel estimation methods of COMPLEX MIXED model based on variational Bayesian

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170212210A1 (en) * 2014-07-17 2017-07-27 Origin Wireless, Inc. Wireless positioning systems
EP3195541B1 (en) * 2014-07-17 2021-11-03 Origin Wireless, Inc. Wireless positioning systems
CN106815854B (en) * 2016-12-30 2020-05-15 西安交通大学 On-line video foreground and background separation method based on regular error modeling
CN107909548B (en) * 2017-05-09 2020-05-15 西安交通大学 Video rain removing method based on noise modeling
CN108171261A (en) * 2017-12-21 2018-06-15 苏州大学 Adaptive semi-supervision image classification method, device, equipment and the medium of robust

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106850472A (en) * 2017-04-10 2017-06-13 中山大学 A kind of OFDM channel estimation method based on Kalman and blind estimate
CN108111441A (en) * 2018-01-12 2018-06-01 电子科技大学 Channel estimation methods based on variational Bayesian
CN108259398A (en) * 2018-01-12 2018-07-06 电子科技大学 The channel estimation methods of COMPLEX MIXED model based on variational Bayesian

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP4152803A4 (en) * 2020-05-29 2023-06-14 Guangdong Oppo Mobile Telecommunications Corp., Ltd. Channel state information processing method, electronic device, and storage medium

Also Published As

Publication number Publication date
CN109150775A (en) 2019-01-04

Similar Documents

Publication Publication Date Title
CN109150775B (en) Robust online channel state estimation method for dynamic change of self-adaptive noise environment
Wang Quantum algorithm for linear regression
CN110632572B (en) Radar radiation source individual identification method and device based on unintentional phase modulation characteristics
Antoniadis et al. Regularization of wavelet approximations
Hang et al. A Bernstein-type inequality for some mixing processes and dynamical systems with an application to learning
CN114580498A (en) Federal learning method with high communication efficiency in wireless communication scene
CN110971457B (en) Time synchronization method based on ELM
CN115731424A (en) Image classification model training method and system based on enhanced federal domain generalization
CN113676266B (en) Channel modeling method based on quantum generation countermeasure network
Li et al. Signal denoising with random refined orthogonal matching pursuit
CN114745233B (en) Joint channel estimation method and device based on pilot frequency design
CN109858356B (en) Method and device for detecting input signal of unknown complex system
Kong et al. A transformer-based CTDNN structure for automatic modulation recognition
Han et al. SLANTS: Sequential adaptive nonlinear modeling of time series
CN115169575A (en) Personalized federal learning method, electronic device and computer readable storage medium
Jana et al. Support recovery with stochastic gates: Theory and application for linear models
Jensen et al. Simulation of conditioned semimartingales on riemannian manifolds
Li et al. Automatic modulation recognition based on a new deep K-SVD denoising algorithm
Băltoiu et al. Sparse Bayesian learning algorithm for separable dictionaries
Jana et al. Support recovery with Projected Stochastic Gates: Theory and application for linear models
Wei et al. A multi-resolution channel structure learning estimation method of geometry-based stochastic model with multi-scene
Maurya A joint convex penalty for inverse covariance matrix estimation
WO2021117163A1 (en) Optical communication status estimation method, device and program
CN112749807A (en) Quantum state chromatography method based on generative model
Zheng et al. A Model-Driven Network Based on ISTA for Massive MIMO Signal Detection

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant