Channel estimation method, device, equipment, medium and product
Through the RWKV neural network model combined with the DPA process, the accuracy of channel estimation is improved and the bit error rate is reduced, and the problems of high channel estimation error rate and high calculation complexity in the prior art are solved, which meets the vehicle communication environment High reliability requirements.
Patent Information
- Application Number
- CN202411143167.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-20
- Publication Date
- 2024-12-03
AI Technical Summary
The existing channel estimation method has high bit error rate, low estimation accuracy, and high computational complexity, which cannot meet the high reliability requirements of inter-vehicle communication.
Using the RWKV neural network model combined with the DPA process, the RWKV prediction of the data subcarrier of the current OFDM symbol is obtained through the DPA prediction channel frequency domain response of the data subcarrier of the previous OFDM symbol and the frequency domain response of the pilot subcarrier of the current OFDM symbol. Channel frequency domain response and DPA estimation is performed to improve channel estimation accuracy and reduce bit error rate.
It significantly reduces the bit error rate and normalized mean square error, improves the accuracy of channel estimation, and reduces the computational complexity, meeting the high reliability requirements in the vehicle communication environment.
Smart Images

Figure CN119071112A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of wireless communications, and in particular to a channel estimation method, device, equipment, medium and product. Background Art
[0002] In the vehicle communication environment, channel estimation for vehicle network communication is mostly based on the IEEE 802.11p standard.
[0003] At present, when performing channel estimation, a deep neural network is introduced based on the STA estimator to use DNN for error compensation. After error compensation, STA is used to predict the channel estimation value of the next symbol.
[0004] However, the bit error rate of this channel estimation method is very high. Summary of the invention
[0005] The purpose of this application is to provide a channel estimation method, device, equipment, medium and product, which can improve the accuracy of channel estimation.
[0006] To achieve the above objectives, this application provides the following solutions:
[0007] In a first aspect, the present application provides a channel estimation method, comprising:
[0008] According to the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol and the frequency domain response of the pilot subcarrier of the current OFDM symbol, the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol is obtained through the trained RWKV neural network model; if the current OFDM symbol is the first OFDM symbol, the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol corresponding to the first OFDM symbol is the channel frequency domain response at the long training symbol of the preamble code in the transmission frame where the first OFDM symbol is located;
[0009] Perform DPA estimation on the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol to obtain the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol; wherein the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol is the channel estimation value corresponding to the data subcarrier of the current OFDM symbol.
[0010] In a second aspect, the present application provides a channel estimation device, comprising:
[0011] The first acquisition module is used to obtain the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol according to the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol and the frequency domain response of the pilot subcarrier of the current OFDM symbol through the trained RWKV neural network model; if the current OFDM symbol is the first OFDM symbol, the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol corresponding to the first OFDM symbol is the channel frequency domain response at the long training symbol of the preamble code in the transmission frame where the first OFDM symbol is located;
[0012] The second acquisition module is used to perform DPA estimation on the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol, and obtain the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol; wherein the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol is the estimated channel estimation value corresponding to the data subcarrier of the current OFDM symbol.
[0013] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any one of the channel estimation methods described above.
[0014] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any one of the channel estimation methods described above.
[0015] In a fifth aspect, the present application provides a computer program product, including a computer program, which implements the steps of any one of the channel estimation methods described above when executed by a processor.
[0016] According to the specific embodiments provided in this application, this application discloses the following technical effects:
[0017] The present application provides a channel estimation method, apparatus, device, medium and product. The method obtains the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol through a trained RWKV neural network model based on the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol and the frequency domain response of the pilot subcarrier of the current OFDM symbol; if the current OFDM symbol is the first OFDM symbol, the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol corresponding to the first OFDM symbol is the channel frequency domain response at the long training symbol of the preamble code in the transmission frame where the first OFDM symbol is located; DPA estimation is performed on the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol to obtain the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol; wherein the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol is the channel estimation value corresponding to the data subcarrier of the current OFDM symbol. Among them, the RWKV neural network itself will judge the gain of the data subcarriers of different OFDM symbols, thereby improving the accuracy of the predicted value of the channel frequency domain response at the data subcarrier, and then associate the adjacent OFDM symbols through the DPA process to further extract the correlation between OFDM symbols and OFDM symbols, thereby further improving the accuracy of channel estimation and reducing the bit error rate. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0019] Figure 1 This is a schematic diagram of a transmission frame structure under IEEE 802.11p in an embodiment of the present application;
[0020] Figure 2 This is an application environment diagram of a channel estimation method in one embodiment of the present application;
[0021] Figure 3 is a flow chart of a channel estimation method according to an exemplary embodiment;
[0022] Figure 4 is a schematic diagram of the structure of a RWKV-DPA estimation method according to an exemplary embodiment;
[0023] Figure 5 is a schematic diagram showing a comparison of bit error rate performances of different estimators at various signal-to-noise ratios according to an exemplary embodiment;
[0024] Figure 6 is a schematic diagram showing a comparison of normalized mean square error performances of different estimation methods at various signal-to-noise ratios according to an exemplary embodiment;
[0025] Figure 7 is a schematic diagram showing a comparison of bit error rates of the proposed RWKV-DPA estimation method under three different Doppler frequency shifts according to an exemplary embodiment;
[0026] Figure 8 is a schematic diagram showing a comparison of normalized mean square errors of the proposed RWKV-DPA estimation method under three different Doppler frequency shifts according to an exemplary embodiment;
[0027] Fig. 9 is a schematic diagram showing a comparison of computational complexity of different estimation methods according to an exemplary embodiment;
[0028] Fig.10 A schematic diagram of functional modules of a channel estimation device provided in one embodiment of the present application;
[0029] Fig.11 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION
[0030] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0031] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.
[0032] With the continuous development of modern society, communication wireless network services are also constantly improving. Future wireless networks need to meet the high demands of emerging applications in high speed, low latency, high reliability and massive connections. Autonomous driving technology has sparked heated discussions in recent years. One of the important issues of autonomous driving technology is to solve the communication problem of cars driving at high speed. When vehicles are driving at high speed, the high-speed mobility brings huge Doppler frequency shift and the multipath effect of the wireless channel causes the channel between the vehicle and other objects to selectively fade at certain frequencies and times. Compared with general wireless channels, communication in vehicle scenarios is more complex. As an important part of wireless communication, channel estimation has a direct impact on the quality of the communication system due to its estimation accuracy. Therefore, it is very important to accurately estimate the dual selective fading channel in vehicle communication scenarios.
[0033] Currently, in the vehicle communication environment, most of the vehicle network communication is based on the IEEE 802.11p standard, which defines the transmission format and parameter settings of the signal during the communication process. The IEEE 802.11p standard is Wireless Access in Vehicle Environments (WAVE), which is specially developed to meet the needs of VANET and support intelligent transmission systems (ITS). The operating frequency band is 5.85~5.95GHZ, with a transmission bandwidth of 10MHZ. Its physical layer is based on the OFDM technology of 64-point FFT, and the transmission frame structure consists of a preamble, an information field, and a data field, such as Figure 1 As shown in Figure 1. The preamble consists of 12 training symbols, including 10 short training symbols and 2 long training symbols, which are mainly used for diversity selection, automatic gain control, timing synchronization, etc. The information field contains the modulation type and coding rate of the subsequent data field. The data field is used to transmit data information and an indefinite number of data symbols.
[0034] In the related art, based on the IEEE 802.11p standard, channel estimation methods are generally divided into two categories: traditional estimation methods and deep learning-based estimation methods.
[0035] 1. Traditional estimation methods
[0036] Usually, there are only four pilots in the data domain symbol. In order to solve the problem of insufficient pilots, many researchers have considered the data-pilot data assisted estimation (DPA) method. The main idea of this method is to regard the demodulated data symbol as a data pilot symbol and use the demodulated data subcarriers outside the pilot subcarriers to estimate the current channel. However, due to the fast time-varying nature of the dual-selective fading channel, the estimation performance of this method is extremely poor.
[0037] 2. Estimation method based on deep learning
[0038] In order to better estimate the dual-selective fading channel, deep learning technology has been widely used in the field of channel estimation in recent years, capturing complex channel characteristics by training neural network models. However, these methods still face many challenges when dealing with time-frequency dual-selective fading channels, such as insufficient estimation accuracy and high model complexity. Therefore, a new channel estimation method is urgently needed to solve the above problems and improve the channel estimation performance in the Internet of Vehicles environment.
[0039] S. Han et al. published a paper titled "A deep learning based channel estimation scheme for IEEE 802.11p systems" at the IEEE International Conference on Communication, May 2019. The paper introduced a deep neural network based on the STA estimator and used DNN for error compensation. DNN performed error compensation, and STA predicted the channel estimation value of the next symbol. However, DNN has a very poor ability to capture the characteristics of a single OFDM symbol, resulting in a very high bit error rate of the STA-DNN estimation method.
[0040] AKGizzini et al. published a paper titled "Temporal averaging LSTM-based channel estimation scheme for IEEE 802.11p standard" at the IEEE Global Communications Conference, Dec. 2021. The paper uses the sequence model LSTM for error compensation, and then uses DPA to predict the next associated symbol, and uses time averaging (TA) to mitigate the mutation error caused by the time step. However, this method uses three steps of LSTM-DPA-TA, and LSTM relies on the gating mechanism, which makes the calculation of each time step more complicated. Since LSTM depends on the output of the previous time step, its training and inference process are difficult to parallelize. The computational complexity of the LSTM-DPA-TA estimator is very high, and the estimation accuracy is limited.
[0041] J. Hou et al. published a paper titled “GRU-Based Deep Learning Channel Estimation Scheme for the IEEE 802.11p Standard” in IEEE Wireless Commun. Lett., vol. 12, no. 10, pp. 764-768, May 2023, which uses a DPA-GRU estimation method with very high bit error rate in dual-selection channels.
[0042] From the above analysis, we can see that the existing channel estimation method has the following problems:
[0043] (1) The accuracy of existing estimators is very low. The bit error rate (BER) and normalized mean square error (NMSE) are usually used to evaluate the accuracy of the estimator. Under the IEEE 802.11p standard, when the signal-to-noise ratio (SNR) is 25dB, the bit error rate (BER) of the LSTM-DPA-TA estimation method is less than 10 -3 The STA-DNN estimation method has a bit error rate of 10 -2 This bit error rate cannot guarantee the communication between vehicles and between vehicles and objects in the environment of Internet of Vehicles. When the signal-to-noise ratio SNR = 25dB, the normalized mean square error (NMSE) of the STA-DNN estimation method is around 10 -1 ~10 -2 The error cannot guarantee the estimation accuracy, which in turn affects the communication quality of the Internet of Vehicles. The RWKV-DPA estimation method proposed in this disclosure has a bit error rate of 10 when SNR = 25dB. -4 Around, NMSE is 10 -3 It greatly improves the channel estimation accuracy under the IEEE 802.11p standard and ensures the communication quality between vehicles and between vehicles and objects.
[0044] (2) Existing estimators lack criteria for information screening and extraction. Whether it is DNN or LSTM, traditional neural networks or sequence models process input features equally, but the importance of wireless channel gain to all subcarriers is not necessarily equal, and the correlation matrix of wireless channel gain is not a diagonal matrix. Some elements may be more important than other elements in DPA estimation. Especially in doubly selective fading channels, due to the rapid changes in the channel, the channel features are over-represented and a large amount of redundant information is generated. Therefore, screening and extracting effective information becomes crucial.
[0045] (3) The complexity of the estimation method is very high while ensuring good estimation accuracy. Although the LSTM-DPA-TA estimation method is ahead of other estimation methods in terms of bit error rate and normalized mean square error, it uses three modules, LSTM, DPA, and TA, for channel estimation, which makes its computational complexity very high.
[0046] In order to solve the above technical problems, the present disclosure proposes a channel estimation method. The channel estimation method provided in the embodiment of the present application can be applied to Figure 2In the application environment shown, the terminal 102 communicates with the server 104 through a network. The data storage system can store data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 may send the channel estimation requirement to the server 104, and the server 104 may obtain the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol based on the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol and the frequency domain response of the pilot subcarrier of the current OFDM symbol through the trained RWKV neural network model; if the current OFDM symbol is the first OFDM symbol, the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol corresponding to the first OFDM symbol is the channel frequency domain response at the long training symbol of the preamble code in the transmission frame where the first OFDM symbol is located; perform DPA estimation on the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol, and obtain the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol; wherein the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol is the channel estimation value corresponding to the data subcarrier of the current OFDM symbol. The server 104 may feed back the obtained channel estimation result to the terminal 102. In addition, in some embodiments, the channel estimation method can also be implemented independently by the server 104 or the terminal 102. For example, the terminal 102 can directly estimate the channel, or the server 104 can obtain the channel estimation requirements of the terminal device from the data storage system and perform visual channel estimation based on the channel estimation requirements.
[0047] The terminal 102 may be, but is not limited to, various desktop computers, laptop computers, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.
[0048] Figure 3 is a flow chart of a channel estimation method according to an exemplary embodiment. Figure 3 As shown, the channel estimation method includes the following steps S201-S202:
[0049] In step S201, based on the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol and the frequency domain response of the pilot subcarrier of the current OFDM symbol, the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol is obtained through the trained RWKV neural network model; if the current OFDM symbol is the first OFDM symbol, the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol corresponding to the first OFDM symbol is the channel frequency domain response at the long training symbol of the preamble code in the transmission frame where the first OFDM symbol is located.
[0050] In step S202, DPA estimation is performed on the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol to obtain the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol; wherein the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol is the channel estimation value corresponding to the data subcarrier of the current OFDM symbol.
[0051] Specifically, step S202 includes the following sub-steps A1-A2:
[0052] A1. Demodulate the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol to obtain a demodulated data pilot at the data subcarrier of the current OFDM symbol.
[0053] A2. Obtain a DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol according to the received value of the data subcarrier of the current OFDM symbol and the demodulated data pilot of the data subcarrier of the current OFDM symbol.
[0054] For the wireless channel in IEEE 802.11p, due to the mobility of the transceiver equipment in the communication and the refraction and reflection of the signal in the wireless propagation, the wireless channel presents a fast time-varying characteristic, and the channel gain of the data subcarrier of each OFDM symbol is different. However, the DNN network in the related technology treats all data subcarriers equally and cannot meet the existing channel requirements. The RWKV neural network itself will judge the gain of the data subcarriers of different OFDM symbols, thereby improving the accuracy of the predicted value of the channel frequency domain response at the data subcarrier, and then associate the adjacent OFDM symbols through the DPA process to further extract the correlation between OFDM symbols, thereby further improving the accuracy of channel estimation and reducing the bit error rate.
[0055] In the present disclosure, after the channel estimation values corresponding to the data subcarriers of all OFDM symbols are obtained, the channel estimation values corresponding to the transmission frame can be obtained based on the channel estimation values corresponding to the data subcarriers of all OFDM symbols.
[0056] Before executing step S201 in the present disclosure, the RWKV neural network model needs to be trained to obtain the trained RWKV neural network model in step S201. The training method of the RWKV neural network model includes the following sub-steps:
[0057] B1. Initialize the RWKV neural network model.
[0058] B2. Acquire a training set, where the training set includes: preamble information of a training frame and a real channel frequency domain response at a training subcarrier of each OFDM training symbol in the training frame.
[0059] The training subcarriers include: training data subcarriers and training pilot subcarriers.
[0060] B3. According to the DPA predicted channel frequency domain response of the training data subcarrier of the previous OFDM training symbol and the training pilot subcarrier of the current OFDM training symbol, the RWKV predicted channel frequency domain response at the training subcarrier of the current OFDM training symbol is obtained through the initialized RWKV neural network model.
[0061] If the current OFDM training symbol is the first OFDM training symbol, the DPA predicted channel frequency domain response of the training data subcarrier of the previous OFDM training symbol corresponding to the first OFDM training symbol is the channel frequency domain response at the long training symbol in the preamble information in the training frame where the first OFDM training symbol is located.
[0062] like Figure 4 As shown, the least squares (LS) estimation algorithm can be used to obtain the frequency domain response of the kth training data subcarrier of the long training symbol in the preamble information of the training frame where the first OFDM training symbol is located, and use it as the DPA predicted channel frequency domain response of the kth training data subcarrier in the previous OFDM training symbol when predicting the RWKV predicted channel frequency domain response at the kth training data subcarrier in the first OFDM training symbol.
[0063] Among them, the channel frequency domain response at the long training symbol in the preamble information in the training frame where the first OFDM training symbol is located can be obtained by the following formula:
[0064]
[0065] in, represents the frequency domain response of the kth training subcarrier of the long training symbol in the preamble information in the training frame where the first OFDM training symbol is located, p[k] represents the long training symbol of the preamble at the kth training subcarrier of the long training symbol in the preamble information in the training frame, and represents the preamble length training symbol received at the kth training subcarrier, S on Indicates the position index of the training subcarrier under the IEEE 802.11p standard.
[0066] The specific algorithm formula of the initialized RWKV neural network model includes:
[0067]
[0068] Wherein, i is a positive integer greater than or equal to 1, It includes the real and imaginary parts of 52 effective subcarriers, and concatenates the real and imaginary parts into a 2S on ×1 vector. S d and S p Respectively represent the position index of the training data subcarrier and the training pilot subcarrier under the IEEE 802.11p standard. on YesS d and S p A set of, representing the position index of the training subcarrier. represents the DPA predicted channel frequency domain response of the kth training subcarrier in the i-1th OFDM training symbol.
[0069] When i=1:
[0070]
[0071] in, It represents the channel frequency domain response obtained by estimating the training pilot subcarrier in the i-th OFDM training symbol using the LS algorithm:
[0072]
[0073] Among them, x i,p [k] represents the training pilot subcarrier in the i-th OFDM training symbol transmitted, y i,p [k] represents the training pilot subcarrier in the received i-th training OFDM symbol.
[0074] Use the RWKV neural network model to process input s i [k] and then the RWKV channel frequency domain response at the k-th training subcarrier in the i-th OFDM training symbol is obtained:
[0075]
[0076] Among them, f RWKV Represents the processing of the RWKV neural network model, with a total weight of Θ. The specific processing of the RWKV neural network model is as follows:
[0077]
[0078] Among them, r k Represents the acceptance vector of past information, W represents the Hadamard product, and σ represents the ReLU activation function, as follows:
[0079] σ(r k )=max(0,r k );
[0080] Among them, r k and wKv k The update comes from:
[0081] r k =W r ·(μ k x k +(1-μ k )x k-1 );
[0082] x k =s i [k];
[0083] x k-1 =s i [k-1];
[0084]
[0085] K k =W K ·(μ K x k +(1-μ K )x k-1 );
[0086] v k =W v ·(μ v x k +(1-μ k )x k-1 );
[0087] Among them, x k and x k-1 is the input of the RWKV neural network model, s i[k] represents the DPA predicted channel frequency domain response of the kth training subcarrier in the i-1th OFDM training symbol and the frequency domain response of the kth training pilot subcarrier in the i-th OFDM training symbol, s i [k-1] represents the DPA predicted channel frequency domain response of the k-1th training subcarrier in the i-1th OFDM training symbol and the frequency domain response of the k-1th training pilot subcarrier in the i-th OFDM training symbol. When k=1, x k-1 The initial value of wKv is 0; k represents the attention score calculation of the RWKV neural network model, which measures the correlation between tokens. As the number of tokens increases, the correlation tends to decay; w represents the position weight decay vector, a trainable model parameter; u represents the special weighting factor applied to the current input; K k Indicates the Key value in the attention score; v k Represents the calculation of the Value value in the attention score. o , W r , W K and W v They represent the projection matrix of the output vector, the projection matrix of the received vector, the projection matrix of the Key value, and the projection matrix of the Value value. r , μ K and μ v They represent the mixing factors of the accepted vector, the mixing factors of the Key value, and the mixing factors of the Value value respectively.
[0088] B4. Calculate the value of the loss function according to the RWKV predicted channel frequency domain response at the training subcarrier of the current OFDM training symbol and the actual channel frequency domain response at the training subcarrier of the current OFDM training symbol in the training set.
[0089] Among them, the mean square error (MSE) function is used as the loss function during training, as shown in the following formula:
[0090]
[0091] Where L represents the total number of OFDM training symbols involved in the training, i.e., the number of training samples, and h i [k] represents the true value of the channel frequency domain response at the kth training subcarrier in the i-th OFDM training symbol, represents the channel frequency domain response predicted by RWKV at the kth training subcarrier of the i-th OFDM training symbol.
[0092] B5. Based on the value of the loss function, update the parameters in the RWKV neural network model to obtain a trained RWKV neural network model.
[0093] After obtaining the value of the loss function, the internal parameters can be updated through the back propagation algorithm. The internal parameters include w, u, W o , W r , W K , W v , μ r , μ K and μ v , the total weight of the model after training is Θ.
[0094] B6. Perform DPA estimation on the RWKV predicted channel frequency domain response at the training subcarrier of the current OFDM training symbol to obtain the DPA predicted channel frequency domain response at the training data subcarrier of the current OFDM training symbol.
[0095] Specifically, after being processed by the RWKV neural network model, the DPA estimation is performed on the RWKV frequency domain response at the k-th training subcarrier in the i-th OFDM training symbol, as shown in the following formula:
[0096]
[0097] in, represents the DPA predicted channel frequency domain response at the kth training data subcarrier in the i-th OFDM training symbol, y i,d [k] represents the kth training data subcarrier of the received i-th OFDM training symbol, represents the data pilot obtained by demodulating the kth training data subcarrier in the i-th OFDM training symbol, Refers to the demodulation operation performed according to the modulation scheme QPSK adopted, It represents the RWKV predicted channel frequency domain response at the kth training data subcarrier of the ith training symbol obtained after prediction by the RWKV neural network model.
[0098] In the present disclosure, the specific implementation of S101 is similar to that of B3, and the specific implementation of S102 is similar to that of B4, which will not be described in detail here.
[0099] From the above analysis, it can be seen that for the wireless channel in IEEE 802.11p, due to the mobility of the transceiver equipment in the communication and the refraction and reflection of the signal in the wireless propagation, the wireless channel presents a fast time-varying characteristic, and the channel gain of each subcarrier is different. However, the DNN network treats all subcarriers equally and cannot meet the existing channel requirements. This method uses the attention mechanism of the RWKV network to calculate the attention score of the subcarrier of each symbol, and further extracts the correlation between symbols, thereby improving the accuracy of channel estimation and reducing the bit error rate.
[0100] Specifically, the method disclosed herein can bring the following beneficial effects:
[0101] (1) The bit error rate is significantly reduced. Figure 5 and Figure 7 As shown in the figure, the proposed RWKV-DPA estimation method is compared with other algorithms under different signal-to-noise ratio environments. The bit error rate of the RWKV-DPA estimation method is significantly lower than that of other estimation methods, which proves the effectiveness of the RWKV-DPA estimation method in the present disclosure. -4 When , the required SNR of RWKV-DPA is 25dB, which is 5dB less than that of the LSTM-DPA-TA estimation method and much less than that of the STA-DNN estimation method.
[0102] (2) The normalized mean square error is significantly reduced. Figure 6 and Figure 8 As shown in the figure, the proposed RWKV-DPA estimation method is compared with other methods under different signal-to-noise ratio environments. The RWKV-DPA estimation method is significantly smaller than other methods in normalized mean square. When the signal-to-noise ratio is 25dB, the normalized mean square error of the RWKV-DPA estimation method is 70% lower than that of the LSTM-DPA-TA estimation method and 86% lower than that of the STA-DNN estimation method.
[0103] (3) The computational complexity is low. Fig. 9 As shown, the computational complexity of the RWKV-DPA estimation algorithm of the present invention is compared with similar algorithms. The complexity of the RWKV-DPA estimation algorithm is 24.9% lower than that of the LSTM-DPA-TA estimation algorithm.
[0104] The channel estimation method in the present disclosure is described in detail below through specific embodiments.
[0105] The present disclosure simulates the IEEE 802.11p standard during the experiment and selects the VTV-SDWW channel model, which describes the communication channel between vehicles in a highway scenario and there is a center wall between the oncoming lanes. The channel model is measured at different locations along various highways in a certain city. The specific parameters of the VTV-SDWW channel are shown in Table 1 below:
[0106] Table 1
[0107]
[0108] In this embodiment, QPSK modulation is adopted and each training frame signal is defined to contain 20 OFDM training symbols. In order to verify the effectiveness of the algorithm for dual-selective channel estimation, the VTV-SDWW maximum Doppler shift is adopted. The RWKV network parameter settings are shown in Table 2:
[0109] Table 2
[0110] Model Type R Model Layers 2 Number of attention mechanism taps 2 Number of training rounds 3000 Number of training samples 100000 Number of test samples 2500 Batch size 64 Optimizer AdamW Loss Function MSE Learning Rate 1e-4 Training signal-to-noise ratio 40
[0111] Specific steps:
[0112] According to the VTV-SDWW channel parameters, the channel is simulated and the transmitted OFDM symbols are generated. After passing through the VTV-SDWW channel, each OFDM symbol is estimated in turn:
[0113] Channel estimation for the first OFDM symbol:
[0114] By receiving the preamble and And send the predefined preamble mid-length symbol p[k] in the OFDM symbol to perform initial channel estimation:
[0115]
[0116] Get the initial channel estimate
[0117] Using the initial channel estimate And the frequency domain response of the pilot subcarrier of the first OFDM symbol is used to predict and compensate the channel frequency domain response at the kth subcarrier of the first OFDM symbol:
[0118]
[0119] Among them, s1[k] is the input of the RWKV neural network model corresponding to the kth subcarrier of the first OFDM symbol, The kth subcarrier of the first OFDM symbol is obtained by applying the RWKV neural network model. For the output of the RWKV neural network model, only the channel frequency domain response at the data subcarrier is retained, that is,
[0120] Utilize the proceeds And receive the signal to perform the DPA process:
[0121]
[0122] in, is the channel estimation value of the kth data subcarrier in the first OFDM symbol, and the estimation of the first OFDM symbol is completed.
[0123] Channel estimation for the second OFDM symbol:
[0124] For the second OFDM symbol, the RWKV neural network model is used for prediction and error compensation. The input of the RWKV neural network model is as follows:
[0125]
[0126]
[0127] The DPA process is then performed using the channel frequency domain response at the kth data subcarrier of the second OFDM symbol and the received signal:
[0128]
[0129] is the channel estimation value of the kth data subcarrier in the second OFDM symbol, and the estimation of the second OFDM symbol is completed.
[0130] …
[0131] The third OFDM symbol is estimated in the same way until all 20 OFDM symbols are estimated.
[0132] Computational complexity analysis of this disclosure:
[0133] The complexity of the estimator is usually measured by analyzing the number of real-valued operations. The present disclosure selects the complexity index of the RWKV-DPA estimator measured by the number of real-valued mathematical operations required to estimate the channel frequency response of a single received OFDM symbol. First, for two long symbols, LS estimation requires 2S on Division and 2S on The second step of the DPA process involves complex value division at each data subcarrier, a total of 22S on The complexity of RWKV mainly depends on the calculation of the attention mechanism, but the advantage of RWKV is that it does not perform operations between inputs, but calculates through offsets, resulting in a total of 13DL operations. D represents the size of RWKV, and L represents the number of layers. In RWKV-DPA, the size of RWKV is consistent with the size of the input vector. t , k t and v t Each vector update in the new process generates 2DL+D 2 L multiplications and DL additions. Also output o t Produce (D 2 +D)L multiplications. The total computational complexity of the RWKV-DPA estimator is 26S on +4D 2 L+20DL.
[0134] In this example, the size D of RWKV is 104, the number of layers L is 2, and S on The number of training subcarriers is 52, and finally the real-valued operation number of the RWKV-DPA estimator is calculated to be 92040.
[0135] Based on the same inventive concept, the embodiment of the present application also provides a channel estimation device for implementing the channel estimation method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more channel estimation device embodiments provided below can refer to the limitations on the channel estimation method above, and will not be repeated here.
[0136] In an exemplary embodiment, Fig.10 As shown, a channel estimation device is provided, comprising:
[0137] The first acquisition module 301 is used to obtain the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol according to the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol and the frequency domain response of the pilot subcarrier of the current OFDM symbol through the trained RWKV neural network model; if the current OFDM symbol is the first OFDM symbol, the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol corresponding to the first OFDM symbol is the channel frequency domain response at the long training symbol of the preamble code in the transmission frame where the first OFDM symbol is located;
[0138] The second acquisition module 302 is used to perform DPA estimation on the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol, and obtain the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol; wherein the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol is the channel estimation value corresponding to the data subcarrier of the current OFDM symbol.
[0139] As an optional implementation manner, the channel estimation device further includes:
[0140] The third acquisition module is used to acquire the channel estimation value corresponding to the transmission frame according to the channel estimation values corresponding to the data subcarriers of all OFDM symbols.
[0141] As an optional implementation manner, the second acquisition module includes:
[0142] A demodulation submodule, configured to demodulate the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol to obtain a demodulated data pilot at the data subcarrier of the current OFDM symbol;
[0143] The first acquisition submodule is used to acquire the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol according to the received value of the data subcarrier of the current OFDM symbol and the demodulated data pilot of the data subcarrier of the current OFDM symbol.
[0144] As an optional implementation manner, the channel estimation device further includes:
[0145] Initialization module, used to initialize the RWKV neural network model;
[0146] A third acquisition module is used to acquire a training set, wherein the training set includes: training frame preamble information and a real channel frequency domain response at a training subcarrier of each OFDM training symbol in the training frame; wherein the training subcarrier includes: a training data subcarrier and a training pilot subcarrier;
[0147] The fourth acquisition module is used to obtain the RWKV predicted channel frequency domain response at the training subcarrier of the current OFDM training symbol according to the DPA predicted channel frequency domain response of the training data subcarrier of the previous OFDM training symbol and the frequency domain response of the training pilot subcarrier of the current OFDM training symbol through the initialized RWKV neural network model; if the current OFDM training symbol is the first OFDM training symbol, the DPA predicted channel frequency domain response of the training data subcarrier of the previous OFDM training symbol corresponding to the first OFDM training symbol is the channel frequency domain response at the long training symbol in the preamble code information in the training frame where the first OFDM training symbol is located;
[0148] A fifth acquisition module, configured to calculate a value of a loss function according to a RWKV predicted channel frequency domain response at a training subcarrier of the current OFDM training symbol and a real channel frequency domain response at a training subcarrier of the current OFDM training symbol in the training set;
[0149] A sixth acquisition module, used for updating the parameters in the RWKV neural network model based on the value of the loss function to obtain the trained RWKV neural network model;
[0150] The seventh acquisition module is used to perform DPA estimation on the RWKV predicted channel frequency domain response at the training subcarrier of the current OFDM training symbol, and obtain the DPA predicted channel frequency domain response at the training data subcarrier of the current OFDM training symbol.
[0151] As an optional implementation, the loss function is a mean square error function.
[0152] As an optional implementation manner, the seventh acquisition module includes:
[0153] The second acquisition submodule is used to update the parameters in the RWKV neural network model using a back propagation algorithm based on the value of the loss function.
[0154] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig.11 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data required for channel estimation. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a channel estimation method is implemented.
[0155] Those skilled in the art will understand that Fig.11 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0156] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.
[0157] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.
[0158] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.
[0159] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.
[0160] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).
[0161] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.
[0162] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0163] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.
Claims
1. A channel estimation method, characterized in that: The channel estimation method comprises: According to the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol and the frequency domain response of the pilot subcarrier of the current OFDM symbol, the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol is obtained through the trained RWKV neural network model; if the current OFDM symbol is the first OFDM symbol, the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol corresponding to the first OFDM symbol is the channel frequency domain response at the long training symbol of the preamble code in the transmission frame where the first OFDM symbol is located; Perform DPA estimation on the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol to obtain the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol; wherein the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol is the channel estimation value corresponding to the data subcarrier of the current OFDM symbol.
2. The channel estimation method according to claim 1, characterized in that The method further comprises: The channel estimation value corresponding to the transmission frame is obtained according to the channel estimation values corresponding to the data subcarriers of all OFDM symbols.
3. The channel estimation method according to claim 1, characterized in that: The performing DPA estimation on the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol to obtain the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol includes: Demodulating the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol to obtain a demodulated data pilot at the data subcarrier of the current OFDM symbol; A DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol is obtained according to the received value of the data subcarrier of the current OFDM symbol and the demodulated data pilot of the data subcarrier of the current OFDM symbol.
4. The channel estimation method according to claim 1, characterized in that: The channel estimation method comprises: Initialize the RWKV neural network model; Acquire a training set, the training set comprising: training frame preamble information and a real channel frequency domain response at a training subcarrier of each OFDM training symbol in the training frame; wherein the training subcarrier comprises: a training data subcarrier and a training pilot subcarrier; According to the DPA predicted channel frequency domain response of the training data subcarrier of the previous OFDM training symbol and the frequency domain response of the training pilot subcarrier of the current OFDM training symbol, the RWKV predicted channel frequency domain response at the training subcarrier of the current OFDM training symbol is obtained through the initialized RWKV neural network model; if the current OFDM training symbol is the first OFDM training symbol, the DPA predicted channel frequency domain response of the training data subcarrier of the previous OFDM training symbol corresponding to the first OFDM training symbol is the channel frequency domain response at the long training symbol in the preamble code information in the training frame where the first OFDM training symbol is located; Calculate the value of the loss function according to the RWKV predicted channel frequency domain response at the training subcarrier of the current OFDM training symbol and the real channel frequency domain response at the training subcarrier of the current OFDM training symbol in the training set; Based on the value of the loss function, updating the parameters in the RWKV neural network model to obtain the trained RWKV neural network model; Perform DPA estimation on the RWKV predicted channel frequency domain response at the training subcarrier of the current OFDM training symbol to obtain the DPA predicted channel frequency domain response at the training data subcarrier of the current OFDM training symbol.
5. The channel estimation method according to claim 4, characterized in that: The loss function is a mean square error function.
6. The channel estimation method according to claim 4, characterized in that: The updating of the parameters in the RWKV neural network model based on the value of the loss function includes: Based on the value of the loss function, a back propagation algorithm is used to update the parameters in the RWKV neural network model.
7. A channel estimation device, characterized in that: The channel estimation device comprises: The first acquisition module is used to obtain the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol according to the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol and the frequency domain response of the pilot subcarrier of the current OFDM symbol through the trained RWKV neural network model; if the current OFDM symbol is the first OFDM symbol, the DPA predicted channel frequency domain response of the data subcarrier of the previous OFDM symbol corresponding to the first OFDM symbol is the channel frequency domain response at the long training symbol of the preamble code in the transmission frame where the first OFDM symbol is located; The second acquisition module is used to perform DPA estimation on the RWKV predicted channel frequency domain response at the data subcarrier of the current OFDM symbol, and obtain the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol; wherein the DPA predicted channel frequency domain response of the data subcarrier of the current OFDM symbol is the estimated channel estimation value corresponding to the data subcarrier of the current OFDM symbol.
8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the channel estimation method according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the channel estimation method according to any one of claims 1 to 6 are implemented.
10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the channel estimation method according to any one of claims 1 to 6 are implemented.
Citation Information
Cited By
Channel estimation simulation method based on multi-feature input GRU
CN120050142A
