CN114143148A - OFDM system channel estimation method based on neural network - Google Patents
OFDM system channel estimation method based on neural network Download PDFInfo
- Publication number
- CN114143148A CN114143148A CN202111410614.3A CN202111410614A CN114143148A CN 114143148 A CN114143148 A CN 114143148A CN 202111410614 A CN202111410614 A CN 202111410614A CN 114143148 A CN114143148 A CN 114143148A
- Authority
- CN
- China
- Prior art keywords
- neural network
- pilot frequency
- basis
- network
- channel
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 32
- 241000283153 Cetacea Species 0.000 claims abstract description 14
- 238000005457 optimization Methods 0.000 claims abstract description 14
- 238000003780 insertion Methods 0.000 claims abstract description 4
- 230000037431 insertion Effects 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 37
- 230000004044 response Effects 0.000 claims description 19
- 238000012937 correction Methods 0.000 claims description 5
- 210000002569 neuron Anatomy 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 238000005314 correlation function Methods 0.000 claims description 3
- 238000011084 recovery Methods 0.000 claims description 3
- 230000001360 synchronised effect Effects 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 abstract description 7
- 230000008569 process Effects 0.000 abstract description 6
- 238000004891 communication Methods 0.000 abstract description 3
- 239000011159 matrix material Substances 0.000 description 9
- 230000000694 effects Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001902 propagating effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
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
- H04L25/0212—Channel estimation of impulse response
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- H04L25/0254—Channel estimation channel estimation algorithms using neural network algorithms
-
- 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
- H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
- H04L25/03006—Arrangements for removing intersymbol interference
- H04L25/03821—Inter-carrier interference cancellation [ICI]
-
- 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
- H04L25/03—Shaping networks in transmitter or receiver, e.g. adaptive shaping networks
- H04L25/03006—Arrangements for removing intersymbol interference
- H04L2025/0335—Arrangements for removing intersymbol interference characterised by the type of transmission
- H04L2025/03375—Passband transmission
- H04L2025/03414—Multicarrier
Landscapes
- Engineering & Computer Science (AREA)
- Power Engineering (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Cable Transmission Systems, Equalization Of Radio And Reduction Of Echo (AREA)
Abstract
The invention discloses a neural network-based OFDM system channel estimation method, and relates to the technical field of wireless communication. The method comprises the steps of pilot frequency position optimization, pilot frequency insertion, fast time-varying channel modeling, compressed sensing channel estimation and the like. The invention obtains the pilot frequency structure which minimizes the cross correlation value by utilizing the neural network, and has low calculation complexity and higher precision. In addition, in the back propagation process of the neural network, the optimal learning rate of the network and the input of the next iteration of the optimized network are obtained by using a whale optimization algorithm, so that the calculation time is shortened.
Description
Technical Field
The invention relates to the technical field of wireless communication, in particular to an OFDM system channel estimation method based on a neural network.
Background
In a wireless communication system, a signal transmitted from a transmitting end reaches a receiving end through direct, reflected, scattered and other paths. In the OFDM system, in order to obtain better performance, channel estimation is required to obtain the state information of the channel. Although the OFDM technique can suppress the inter-symbol interference generated by the multipath effect by adding the cyclic prefix, it is extremely sensitive to the doppler effect generated by high-speed movement. Typical channel estimation methods include a least square method and a least mean square error method, however, when the relative moving speed of a transmitting end and a receiving end of a system is high, the performance of the two methods is limited, and the sparsity of a wireless channel in a delay-Doppler domain during high-speed movement is ignored.
For the channel response at the pilot frequency position of the fast time-varying channel moving at a high speed, a base expansion model can be adopted to model the channel, the parameter estimation of the complex fast time-varying channel is simplified into the estimation of a small number of base function coefficients, and the base function coefficients can be estimated by using a compressed sensing technology based on the sparsity of the fast time-varying channel. And finally, obtaining the channel response of the data position by an interpolation method according to the channel response of the pilot frequency position.
However, due to inter-subcarrier interference caused by doppler effect in high-speed movement, the description of the channel response of the data location is inaccurate, thereby affecting the system performance.
Disclosure of Invention
In view of this, the invention provides an OFDM system channel estimation method based on a neural network, which can resist inter-subcarrier interference generated by doppler effect in high-speed movement of an OFDM system, more accurately describe channel response of a data location, and improve system performance.
In order to achieve the purpose, the invention provides the following technical scheme:
an OFDM system channel estimation method based on neural network includes the following steps:
s1, pilot position optimization: optimizing the pilot frequency position by using a neural network, and reducing training times by using a whale optimization algorithm;
s2, pilot insertion: inserting the optimized pilot frequency at the transmitting end, wherein the pilot frequency comprises the pilot frequency and the virtual subcarrier;
s3, modeling a fast time-varying channel: modeling a fast time-varying channel in a high-speed moving scene by using a base extension model;
s4, compressed sensing channel estimation: estimating the channel response of the pilot frequency position by utilizing the receiving pilot frequency and the protection pilot frequency at a receiving end;
s5, error correction: correcting errors of the basis function coefficients of the basis expansion model in the step S3;
s6, interpolation: and obtaining the channel response of the data position between the pilot frequencies by utilizing an interpolation method at the receiving end to complete channel estimation.
Further, in step S1, the neural network has six layers, four of which are hidden layers, each layer including 2048 neurons; the neural network uses a Tanh function as an activation function, uses a whale optimization algorithm to search and obtain the optimal learning rate of the network, uses a cross-correlation function as a loss function of the network, and trains in a supervision learning mode.
Further, in step S3, the basis expansion model is third order, and a fourier basis is used as the basis function.
Further, in step S4, a synchronous orthogonal matching pursuit algorithm is used as the compressed sensing recovery algorithm.
Further, in step S5, the basis function coefficients of the basis-extended model in step S3 are error-corrected using the basis functions based on the discrete carhennan-loeve expansion.
Further, in step S6, the data position channel response is obtained by linear interpolation
The invention has the following beneficial effects:
(1) the invention obtains the pilot frequency structure which minimizes the cross correlation value by utilizing the neural network, and has low calculation complexity and higher precision.
(2) In the back propagation process of the neural network, the optimal learning rate of the network and the input of the next iteration of the optimized network are obtained by using a whale optimization algorithm, and the calculation time is shortened.
(3) The neural network in the invention does not need to be trained in advance, and the use precision of each time can be improved.
Drawings
Fig. 1 is a schematic diagram of an optimized pilot structure and a position of a guard pilot in an embodiment of the present invention.
Fig. 2 is a schematic diagram of the operation process of the neural network in the embodiment of the present invention.
Fig. 3 is a schematic structural diagram of a neural network according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The common channel estimation method is suitable for the situation of low moving speed, and for a high-speed moving scene, the interference between adjacent subcarriers is very serious, and the performance of the traditional channel estimation method is severely deteriorated. Based on the above, the following provides an OFDM system channel estimation method based on a neural network, which models a fast time-varying channel by using a basis expansion model, recovers a basis function coefficient by using a compressed sensing technique, and further obtains a channel response of a pilot frequency position, and for a modeling error caused by a complex exponential basis function, performs error correction on the obtained basis function coefficient by using a basis function based on discrete karhunen-lowei expansion; the channel response for the data location is obtained using interpolation.
The specific implementation mode is as follows:
firstly, obtaining an optimal pilot frequency position by using a neural network, inserting a pilot frequency into a corresponding position, and including a protection pilot frequency, wherein fig. 1 is a possible pilot frequency insertion mode mainly for explaining a relative position relationship between the protection pilot frequency and the pilot frequency; modeling a fast time-varying channel by using a third-order complex exponential basis function; recovering the basis function coefficients by utilizing an orthogonal matching pursuit algorithm in a compressed sensing technology at a receiving end according to the received pilot frequency and the protection pilot frequency, and performing error correction on the obtained basis function coefficients by using a basis function based on discrete carhennan-lowei expansion; and obtaining the channel response of the data position by interpolation according to the channel response of the pilot position.
Specifically, the error correction method for the obtained basis function coefficients using the basis functions based on the discrete carhennan-lovin expansion is as follows:
when the channel response is modeled using complex exponential basis functions, the channel response is expressed as:
wherein, BCEIs a matrix of basis functions, ILA diagonal matrix representing the number of diagonal elements as L, where L is the number of multipaths of the system, cCEIs a matrix of coefficients of the basis functions,
to model errors.
When modeling the channel response using basis functions based on the discrete carhennan-lovin expansion, the channel response is expressed as:
wherein, BDKLIn the form of a matrix of basis functions,
is a matrix of coefficients of the basis functions,
to model errors.
The two are equal, the subtraction value of the modeling error is extremely small and can be ignored, and the following can be obtained:
Specifically, the optimal pilot position is obtained by using neural network fitting as follows:
when compressed sensing is used to recover the basis coefficients, the key is that the placement of the pilots needs to make the value of the cross-correlation of the measurement matrix as small as possible, and the smaller the value, the higher the recovery accuracy. To get the optimal number and location of pilots, a neural network is used for the calculation. Firstly, the number of subcarriers and the number of OFDM blocks of a scheme are determined, then the pilot frequencies at different numbers of positions are used for training the network, and simultaneously, a gradient descent algorithm is used for updating the weight and the offset set in the network, so that the value of a loss function is minimized, and an estimation result with higher precision is achieved. The inputs to the network at this time are: a random number and location of pilots. The output of the network is: an optimal number and location of pilots. In order to enhance the representation capability of the network and avoid overfitting, and simultaneously take complexity into consideration, as shown in fig. 3, a total of 6 layers of the neural network are selected, each layer of the hidden layers comprises 2048 neurons, and the nonlinear transformation obtained by weighted summation of the network neurons in the previous layer is input to the next layer. Meanwhile, a Tanh function is used as an activation function of the network, and a cross-correlation function is used as a loss function of the network.
Specifically, when a gradient descent algorithm is used to update the weight and the bias set in the network in the back propagation process, the gradient of the loss function is calculated at first:
wherein, yi，lIs the output of the network when propagating in the forward direction.
Then updating the weight matrix W of each layer in turnlAnd a bias matrix bl：
As shown in fig. 2, for the learning rate α in the formula, a whale optimization algorithm is used to perform a search to obtain an optimal value of the network in the current iteration, and the network is updated every iteration. And in the searching process, when the output of the network does not meet the error requirement, the output of the iterative neural network is also used as the input of the whale optimization algorithm. So in each iteration, the output of the whale optimization algorithm is: the learning rate of the iterative neural network at this time and the output of the iterative neural network at the next time. In the improved whale optimization algorithm, a whale position updating formula is as follows:
in the formula, t is the number of iterations,
for the position vector associated with the optimal solution, b is a constant defining the shape of the logarithmic spiral, and l is [ -1, 1 [ ]]A random number in between. A and
as coefficient vectors:
A＝2ar1-a
C＝2r2
wherein r is1、r2Is a random number with a value range of [0, 1 ]]The larger the iteration number of the control parameter is, the smaller the numerical value is, and the more the numerical value is, and the more the numerical value is, the greater the numerical value is, the control parameter is, and the more the greater the numerical value is, and the more the greater the more the greater the numerical value is, the greater the numerical value is, the greater the numerical value is, the greater the numerical value is, the greater the control parameter is, the greater the numerical value is, the greater the numerical value is, the greater the numerical value is, the greater the numerical value is, the greater.
the method comprises the following specific steps:
1. initializing whale population Xi(i＝1，2，.·.，n)；
2. Randomly selecting an initial position
Initializing an error as 10000;
3. current position XjThe training error of the neural network of the iteration is error (j) when j is 1, 2,. cndot., n);
4、X*the optimal position;
5. while (t < maximum number of iterations)
For all individuals
Defining individuals X at a current locationj∈(0，1]
If1 error(j)＜error
error＝error(j)
And XjParticipating in location updates
end if1
Updating parameters a, A, C, l, p
if2(p＜0.5)
if3(|A|＜1)
Updating the current individual location according to equation (1)
else if3(|A|≥1)
Updating the current individual location according to equation (2)
end if3
else if2(p≥0.5)
Updating the current individual location according to equation (3)
end if2
Let the updated position be Xj+1And Xj+1∈(0,1]
end while
Output X*
X*I.e. the matrix of the optimal learning rate of the iterative network and the next input of the network。
In a word, the invention fully considers and utilizes the characteristic of high-speed movement of the fast time-varying channel, obtains the pilot frequency structure which minimizes the cross correlation value by utilizing the neural network, and has low calculation complexity and higher precision. In addition, in the back propagation process of the neural network, the optimal learning rate of the network and the input of the next iteration of the optimized network are obtained by using a whale optimization algorithm, so that the calculation time is shortened.
It should be noted that although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations can be made in the above embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (6)
1. A channel estimation method of an OFDM system based on a neural network is characterized in that: the method comprises the following steps:
s1, pilot position optimization: optimizing the pilot frequency position by using a neural network, and reducing training times by using a whale optimization algorithm;
s2, pilot insertion: inserting the optimized pilot frequency at the transmitting end, wherein the pilot frequency comprises the pilot frequency and the virtual subcarrier;
s3, modeling a fast time-varying channel: modeling a fast time-varying channel in a high-speed moving scene by using a base extension model;
s4, compressed sensing channel estimation: estimating the channel response of the pilot frequency position by utilizing the receiving pilot frequency and the protection pilot frequency at a receiving end;
s5, error correction: correcting errors of the basis function coefficients of the basis expansion model in the step S3;
s6, interpolation: and obtaining the channel response of the data position between the pilot frequencies by utilizing an interpolation method at the receiving end to complete channel estimation.
2. The method according to claim 1, wherein in step S1, the neural network has six layers, four layers are hidden layers, and each layer comprises 2048 neurons; the neural network uses a Tanh function as an activation function, uses a whale optimization algorithm to search and obtain the optimal learning rate of the network, uses a cross-correlation function as a loss function of the network, and trains in a supervision learning mode.
3. The method as claimed in claim 1, wherein in step S3, the basis expansion model is third order, and fourier basis is used as the basis function.
4. The method for estimating the channel of the OFDM system based on neural network as claimed in claim 1, wherein in step S4, a synchronous orthogonal matching pursuit algorithm is used as the compressed sensing recovery algorithm.
5. The method as claimed in claim 1, wherein in step S5, the basis function coefficients of the basis expansion model in step S3 are error corrected by using the basis functions based on the discrete karhunen-loeve expansion.
6. The method as claimed in claim 1, wherein in step S6, the data position channel response is obtained by linear interpolation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111410614.3A CN114143148A (en) | 2021-11-25 | 2021-11-25 | OFDM system channel estimation method based on neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111410614.3A CN114143148A (en) | 2021-11-25 | 2021-11-25 | OFDM system channel estimation method based on neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114143148A true CN114143148A (en) | 2022-03-04 |
Family
ID=80391636
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111410614.3A Pending CN114143148A (en) | 2021-11-25 | 2021-11-25 | OFDM system channel estimation method based on neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114143148A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109361630A (en) * | 2018-09-30 | 2019-02-19 | 上海交通大学 | Channel estimation methods and system based on compressed sensing and iterative interference cancellation |
| CN109450830A (en) * | 2018-12-26 | 2019-03-08 | 重庆大学 | Channel estimation methods based on deep learning under a kind of high-speed mobile environment |
| CN113285899A (en) * | 2021-05-20 | 2021-08-20 | 南京邮电大学 | Time-varying channel estimation method and system based on deep learning |
| CN113472706A (en) * | 2021-07-12 | 2021-10-01 | 南京大学 | MIMO-OFDM system channel estimation method based on deep neural network |
-
2021
- 2021-11-25 CN CN202111410614.3A patent/CN114143148A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109361630A (en) * | 2018-09-30 | 2019-02-19 | 上海交通大学 | Channel estimation methods and system based on compressed sensing and iterative interference cancellation |
| CN109450830A (en) * | 2018-12-26 | 2019-03-08 | 重庆大学 | Channel estimation methods based on deep learning under a kind of high-speed mobile environment |
| CN113285899A (en) * | 2021-05-20 | 2021-08-20 | 南京邮电大学 | Time-varying channel estimation method and system based on deep learning |
| CN113472706A (en) * | 2021-07-12 | 2021-10-01 | 南京大学 | MIMO-OFDM system channel estimation method based on deep neural network |
Non-Patent Citations (2)
| Title |
|---|
| 曹梦硕等: "基扩展模型下基于深度学习的双选信道估计方法", 计算机测量与控制, 25 October 2020 (2020-10-25), pages 205 - 210 * |
| 袁继耀等: "OFDM系统中联合卡尔曼滤波与导频符号的快变信道估计算法", 信息通信, 15 January 2018 (2018-01-15) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111786921B (en) | Aviation communication system base extension channel estimation method based on prior time delay information | |
| CN105227505B (en) | A kind of more symbol combination channel estimating methods under high-speed mobile environment | |
| CN102664841B (en) | Method for SC-FDE (single carrier-frequency domain equalization) system low complexity RLS self-adaption channel estimation | |
| CN111147407B (en) | TMSBL underwater acoustic OFDM time-varying channel estimation method based on channel prediction | |
| CN107395536B (en) | Method for estimating underwater sound channel impulse response function in multi-path environment | |
| CN111131097B (en) | Block diagonal sparse Bayesian channel estimation method under SC-MIMO underwater acoustic communication environment | |
| CN113242191B (en) | Improved time sequence multiple sparse Bayesian learning underwater acoustic channel estimation method | |
| CN110311876A (en) | The implementation method of underwater sound OFDM receiver based on deep neural network | |
| JP2005515698A5 (en) | ||
| CN114143148A (en) | OFDM system channel estimation method based on neural network | |
| Zhang et al. | H∞ fixed-lag smoothing for discrete linear time-varying systems | |
| CN111291511B (en) | Soft Kalman filtering iteration time-varying channel estimation method based on historical information | |
| CN111555994A (en) | Cluster sparse channel estimation method based on maximum skip rule algorithm | |
| CN113923085B (en) | Underwater acoustic communication system multi-transmitting-end parallel sparse channel estimation method | |
| CN110784423A (en) | Underwater acoustic channel estimation method based on sparse constraint | |
| Chen et al. | Hardware efficient massive MIMO detector based on the Monte Carlo tree search method | |
| CN115987722A (en) | Deep learning assisted OFDM channel estimation and signal detection method | |
| CN111695617B (en) | Distributed fire control fusion method based on improved covariance intersection algorithm | |
| CN107888537B (en) | Signal detection method for improving system complexity in large-scale antenna system | |
| CN113014341A (en) | Estimation method for nonideal sparse channel | |
| Campillo et al. | Parallel and interacting Markov chain Monte Carlo algorithm | |
| CN108279564B (en) | Robust sparse multi-task self-adaptive system and iteration method | |
| Li et al. | Pilot allocation optimization using enhanced salp swarm algorithm for sparse channel estimation | |
| CN112702286B (en) | Method for estimating downlink channel in unmanned aerial vehicle communication | |
| CN116827725B (en) | Iterative channel parameter estimation method based on genetic algorithm |
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 |