Task-based channel estimation quantification method

Abstract

The invention discloses a channel estimation quantization method based on tasks, and belongs to the field of radio frequency (RF) / VLC hybrid transmission systems and MIMO. The method specifically comprises the following steps: firstly, calculating a VLC signal matrix YVLC received by a downlink channel user by utilizing an LOS channel gain between a user side and an LED transmitting end, and converting the VLC signal matrix YVLC into a vector form to obtain a pilot frequency receiving signal yVLC; then, a VQ algorithm is adopted, the yVLC is mapped to a codebook, Bussgang decomposition is used for quantization, and a quantization output code cn and an index vector in of the quantization output code cn are obtained; uplink transmission is carried out on an index vector in, a receiving end obtains a signal matrix yRF, then, an estimator # imgabs0 # of the index vector is solved by utilizing the signal matrix yRF, a channel quantization output estimator # imgabs1 # is obtained by utilizing a password book, finally, a minimum mean square error MSE of channel estimation is introduced, and an optimal solution of a post-processing matrix D is solved, so that the channel estimation error is minimum. And the channel state is optimal. According to the invention, the error of channel estimation is effectively reduced, and the channel capacity of the system is further improved through accurate CSI estimation.

Description

Technical Field

[0001] The invention relates to the field of radio frequency (RF) / VLC hybrid transmission system and MIMO, and in particular to a task-based channel estimation and quantization method. Background Art

[0002] Visible light communication (VLC) can utilize the high-frequency switching characteristics of LEDs for data transmission while meeting lighting needs. On this basis, the MIMO technology can significantly improve system performance by increasing bandwidth. However, the performance of MIMO depends on the accurate channel state information (CSI) obtained by the base station (BS), and CSI feedback will occupy a large amount of uplink bandwidth resources, especially in the hybrid RF / VLC system of VLC. Therefore, in order to solve the problems of scarcity of RF bandwidth resources and the huge overhead of CSI feedback, it is necessary to reduce the cost of uplink VLC transmission through quantization, coding and spectrum reuse.

[0003] In the existing technology, quantization has attracted much attention due to its simplicity and high efficiency. The current quantization algorithms include scalar quantization (SQ) and vector quantization (VQ). Although SQ is simple, it has large errors; VQ can capture more complex channel characteristics, but the design is more complicated; and most previous studies have ignored the impact of quantization errors on codebook recovery and only focused on the design of quantization codebooks. In addition, due to the LOS characteristics of VLC, the channel performance is significantly affected by the distance and direction of the device, and the existing methods do not provide sufficient support for accurate CSI estimation.

[0004] Therefore, it is particularly important to study and optimize a method that can simultaneously reduce CSI feedback overhead and improve channel estimation accuracy. Summary of the invention

[0005] In view of the above problems, the present invention proposes a task-based channel estimation quantization method to alleviate the conflict between channel state information estimation accuracy and RF resource cost in MIMO field radio frequency (RF) / VLC hybrid transmission system.

[0006] The task-based channel estimation and quantization method comprises the following steps:

[0007] Step 1: Use the LOS channel gain between the user end and the LED transmitter to calculate the VLC signal matrix Y received by the downlink channel user VLC :

[0008] Y VLC =H VLC Φ T +W

[0009] H VLC is the real channel matrix, W is the real noise matrix, and Φ is the transmitted signal;

[0010] Step 2: VLC signal matrix Y VLC Converted into vector form, the pilot received signal y is obtained VLC :

[0011]

[0012] Represents the identity matrix; vec() represents the vectorization function of the matrix. represents the Kronecker product;

[0013] Step 3: Use the VQ algorithm to transform the vector y VLC Mapped to the codebook, and quantized using Bussgang decomposition to obtain the quantized output code c n and its index vector i n ;

[0014] The mapping formula is:

[0015] c n =Q B (y VLC )

[0016] Q B is a quantization function that transforms the vector y VLC Quantized into a codeword c in the codebook n ;c n represents the nth code in the codebook set B;

[0017] Using Bussgang decomposition, we get the quantitative results:

[0018] c n =Gy VLC +e

[0019] Where e is a zero-mean random variable and G is the Bussgang gain, calculated as:

[0020]

[0021] Indicates the code c n and VLC The cross-correlation matrix of Represents vector y VLC The autocorrelation matrix of .

[0022] Step 4: Index vector i n Uplink transmission, the receiving end obtains the signal matrix y RF :

[0023] y RF =H RFP RF i n +N

[0024] H RF is the RF channel matrix; P RF is the uplink signal precoding matrix; N is the noise matrix at the receiving end.

[0025] Step 5: Use the signal matrix y RF Find an estimate of the index vector Using the codebook to obtain an estimate of the channel quantization output code

[0026] First, we calculate the estimator of the index vector The formula is:

[0027] Π is the projection operation function, Used to solve the pseudo-inverse of a matrix;

[0028] Then, the estimated channel quantization output code is recovered from the received index vector according to the codebook And use the post-processing matrix D to get the estimated channel matrix

[0029]

[0030] Step 6: Introduce the minimum mean square error (MSE) of channel estimation and solve to obtain the optimal solution of the post-processing matrix D, so that the channel estimation error is minimized and the channel state is optimal.

[0031] The specific steps are as follows:

[0032] Step 601: Define the real channel vector s and the estimated value of s

[0033] They are:

[0034] Step 602: Introduction For a given y VLC The minimum mean square error estimate of time s;

[0035] Γ is the linear transformation matrix, defined as

[0036] Step 603: Using channel information s and its estimated value Calculate the minimum value of the mean square error MSE, that is:

[0037]

[0038] Step 604: Derivative the second term of the mean square error MSE expansion to obtain an optimal solution of the post-processing matrix D that minimizes the uplink CSI quantization error;

[0039] First, the second term is expanded as:

[0040] Taking the derivative, we get:

[0041] Let the derivative result be equal to 0, and get the optimal solution D of the post-optimization matrix:

[0042]

[0043] y index Represents the quantized index vector.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] 1. A task-based channel estimation quantization method can effectively reduce feedback overhead: By designing a codebook based on vector quantization (VQ), the number of CSI feedback bits can be significantly reduced under limited uplink radio frequency (RF) resources, thereby saving uplink resources.

[0046] 2. A task-based channel estimation quantization method to improve channel estimation accuracy: This method optimizes the quantization codebook and post-processing matrix, and provides a closed-form solution of the post-processing matrix based on the minimum mean square error (MMSE) criterion, effectively reducing the error of channel estimation.

[0047] 3. A task-based channel estimation quantization method that can achieve better error performance: Simulation results show that under the same quantization bit conditions, the channel estimation error (MSE) of the TQ-CE scheme is significantly lower than that of the traditional scalar quantization (SQ) and vector quantization (VQ) schemes.

[0048] 4. A task-based channel estimation quantization method, more accurate CSI estimation further improves the channel capacity of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Flow chart of the radio frequency (RF) / VLC hybrid transmission system in an embodiment of the present invention.

[0050] Figure 2 Schematic diagram of a radio frequency (RF) / VLC hybrid transmission system in an embodiment of the present invention.

[0051] Figure 3 The flowchart of uplink and downlink information transmission and changes in a radio frequency (RF) / VLC hybrid transmission system in an embodiment of the present invention.

[0052] Figure 4 This is a curve diagram comparing the mean square error of channel estimation between TQ-CE and its comparison scheme when the training signal-to-noise ratio is 5 dB in an embodiment of the present invention.

[0053] Figure 5 This is a curve diagram comparing the mean square error of channel estimation between TQ-CE and its comparison scheme when the training signal-to-noise ratio is 10 dB in an embodiment of the present invention.

[0054] Figure 6 This is a curve diagram comparing the channel capacities of TQ-CE and its comparison schemes when the training signal-to-noise ratio is 5 dB in an embodiment of the present invention.

[0055] Figure 7 This is a curve diagram showing the channel capacity comparison between TQ-CE and its comparison scheme when the training signal-to-noise ratio is 10 dB in the embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to facilitate those skilled in the art to understand and implement the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only partial embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work should fall within the scope of protection of the present invention.

[0057] The present invention proposes a task-based quantization method (TQ-CE), which minimizes the mean square error (MSE) of channel estimation by optimizing the quantization codebook and post-processing matrix, effectively solving the cost and performance problems.

[0058] Specifically, in the RF / VLC system, the visible light signal matrix received by the user is first obtained according to the downlink channel gain, and the corresponding codebook containing channel information is obtained by using the VQ algorithm and Bussgang decomposition. The user uploads the codebook information, and the LED receiving end uses the codebook and matrix transformation to obtain the estimated channel matrix vector. In order to minimize the error between the channel estimation value and the true value, the post-processing matrix is ​​optimized to achieve the optimal solution. In the whole process, the quantization codebook and the post-processing matrix are designed to minimize the mean square error (MSE) of the channel estimation, thereby taking into account the cost and accuracy of the system.

[0059] The present invention is a task-based channel estimation quantization scheme that takes into account both channel state information estimation accuracy and radio frequency resource cost in a radio frequency (RF) / VLC hybrid transmission system. The scheme aims to minimize the mean square error (MSE) of channel estimation. By means of a VQ algorithm, Bussgang decomposition and codebook information, the post-processing matrix is ​​optimized to achieve an optimal solution, thereby optimizing the cost and accuracy of the system.

[0060] like Figure 1 As shown, the following steps are included:

[0061] Step 1: Use the LOS channel gain between the user end and the LED transmitter to express the VLC signal matrix Y received by the downlink channel user VLC :

[0062] Y VLC =H VLC Φ T +W

[0063] H VLC is the real channel matrix, W is the real noise matrix, and Φ is the transmitted signal;

[0064] Step 2: VLC signal matrix Y VLC Convert to vector form and get the pilot received signal vector y VLC :

[0065]

[0066] Represents the identity matrix; vec() represents the vectorization function of the matrix. represents the Kronecker product;

[0067] Step 3: Use the VQ algorithm to transform the vector y VLC Mapped to a finite codebook and quantized using Bussgang decomposition to obtain the quantized output code c n and its index vector i n ;

[0068] Each codeword corresponds to a possible channel state of the MIMO VLC system, that is, the mapping formula is:

[0069] c n =Q B (y VLC )

[0070] Q B is a quantization function that transforms the vector y VLC Quantized into a codeword c in the codebook n ;c n represents the nth code in the codebook set B;

[0071] By using Bussgang decomposition, the nonlinear quantization process is equivalent to a linear operation, and the quantization result is obtained:

[0072] c n =Gy VLC +e

[0073] Where e is a zero-mean random variable and G is the Bussgang gain, calculated as:

[0074]

[0075] Indicates the code c n and VLC The cross-correlation matrix of Represents vector y VLC The autocorrelation matrix of .

[0076] Step 4: Output the quantized CSI codebook set B into code c n The index vector i n Uplink transmission, the receiving end obtains the signal matrix y RF :

[0077] y RF =H RF P RF i n +N

[0078] H RF is the RF channel matrix; P RF is the uplink signal precoding matrix; N is the noise matrix at the receiving end (the elements have a mean of zero and a variance of real-valued Gaussian variable).

[0079] Step 5: Use the signal matrix y RF Find an estimate of the index vector Using the codebook to obtain an estimate of the channel quantization output code

[0080] First, calculate the index vector of the received estimate The formula is:

[0081] Π is the projection operation function, Used to solve the pseudo-inverse of a matrix;

[0082] Then, the estimated channel quantization output code is recovered from the received index vector according to the codebook And use the post-processing matrix D to get the estimated channel matrix

[0083]

[0084] Step 6: In order to minimize the error between the channel estimate and the true value, the minimum mean square error (MSE) of the channel estimate is introduced, and the optimal solution of the post-processing matrix D is obtained, so that the channel state reaches the optimal state.

[0085] The specific steps are as follows:

[0086] Step 601: Define the real channel vector s and the estimated value of s

[0087] They are:

[0088] Step 602: Introduction For a given y VLC The minimum mean square error estimate for time s

[0089] Γ is the linear transformation matrix, defined as

[0090] Step 603: Using channel information s and its estimated value Calculate the minimum value of the mean square error MSE, that is:

[0091]

[0092] Step 604: Derivative the second term of the mean square error MSE expansion to obtain an optimal solution of the post-processing matrix D that minimizes the uplink CSI quantization error;

[0093] The main goal of the present invention is to minimize the uplink CSI quantization error, and the uplink CSI quantization error is only related to the second term of the mean square error expansion, so only the second term can be derived and the first term can be regarded as an irrelevant constant.

[0094] First, the second term is expanded as:

[0095]

[0096] Taking the derivative, we get:

[0097] Let the derivative result be equal to 0, and get the optimal solution D of the post-optimization matrix:

[0098]

[0099] y index represents the quantized index vector, and from step 3 we can set

[0100] Using the above principles, the optimal solution of the corresponding post-optimization matrix is ​​obtained and a simulation is performed.

[0101] Example:

[0102] This embodiment constructs a scenario of radio frequency (RF) / VLC hybrid information transmission in the MIMO field, such as Figure 2As shown, the VQ algorithm is used to map the vector channel information to a finite codebook; the RF channel matrix and the uplink signal precoding matrix are used to minimize the mean square error of the channel estimation to obtain the optimal solution of the post-optimization matrix; and the channel information is restored and verified by simulation.

[0103] Principle Figure 3 As shown in Figure 1, the system is divided into downlink and uplink. In the downlink, the light source acts as the transmitter to send a signal Φ, which passes through the channel state information matrix H. VLC After being affected by the user end, y p , and perform the corresponding vector form conversion. Then, the vector is quantized and decomposed to obtain c n and its corresponding index vector i n In the uplink, the user end will i n Send, through the RF channel matrix H RF After being affected by the light source, we get Y q , using Y q Recover the estimated index vector Estimator of the channel quantization output Finally, the channel state information matrix estimate is obtained by post-optimizing the matrix D

[0104] The specific steps are as follows:

[0105] Step 1: Use the LOS channel gain between the user end and the LED transmitting end to represent the VLC signal matrix received by the downlink channel user;

[0106] Specifically: For the downlink channel, assume that all LEDs send pilot signals to the user at the same time (N T is the number of LEDs in each BS base station) Since the LOS channel of the wireless optical channel contains most of the transmission energy, this example only considers the LOS channel. In particular, the user receiver i th and LED lights th The LOS channel gain between

[0107]

[0108] in φ is the radiation angle, ψ is the incident angle, Ψ C is the incident angle, the half angle of the receiver field of view, A is the detector area, d i,j It is the user receiver PDi th and LED lights th The distance between them, T(ψ) is the gain of the optical filter, g(ψ) is the gain of the light concentrator, and hi,j Written in matrix form as H VLC Then we get the VLC signal matrix Y received by the downlink channel user: VLC ;

[0109] Step 2: Convert the signal matrix into a vector form;

[0110] Step 3: Using VQ algorithm, vector y VLC is mapped onto a finite codebook;

[0111] Each codeword corresponds to a possible channel state of the MIMO VLC system. For example, the quantized signal obtained after B-bits VQ can be expressed as

[0112] c n =Q B (y VLC )

[0113] c n Represents the nth code in the codebook set B; the number of elements in set B is 2 B indivual;

[0114] Step 4: Use the RF channel matrix, uplink signal precoding matrix and noise matrix to output code c in codebook set B representing the quantized CSI n The index vector i n Uplink transmission (representing quantized CSI); get:

[0115] y RF =H RF P RF i n +N

[0116] According to the codebook, the received index vector Restore the estimated channel quantization output And use the post-processing matrix to get the estimated channel matrix

[0117]

[0118] Step 5: In order to minimize the error between the channel estimate and the true value, the mean square error (MSE) of the channel estimate is introduced, and the task is to minimize the mean square error, that is,

[0119]

[0120] s is the task vector, that is, s = vec(H VLC ), represents the pilot received signal y VLC The estimated value of s when known;

[0121] Specifically, in the mean square error expression, MSE is divided into the sum of two terms. The first term represents the CSI s of the VLC link and its approximation The second term represents the estimation error between Since the main goal of this scheme is to minimize the uplink CSI quantization error, and the uplink CSI quantization error is only related to the second term, the second term is optimized, and the first term is regarded as an irrelevant constant.

[0122] Step 6: Expand the second term:

[0123]

[0124] Because the optimal solution of D is required to minimize the mean square error, the expansion is derived with respect to D, and the derivative result is:

[0125]

[0126] Set it equal to 0 to obtain the optimal solution of the post-optimization matrix

[0127]

[0128] Step 7: Using the above principles, find the optimal solution of the corresponding post-optimization matrix and perform simulation:

[0129] Specifically, the effectiveness of the TQ-CE algorithm was first verified by simulation, and the traditional SQ-CE and VQ-CE schemes were selected for comparison; an RF / VLC hybrid MIMO system was constructed, in which the downlink communication was based on VLC and the uplink communication was based on RF.

[0130] according to Figure 4 It can be seen that when the training signal-to-noise ratio (SNR) is 5dB: As the number of quantization bits increases, the MSE of all schemes gradually decreases. However, under the same number of quantization bits, the MSE of TQ-CE is significantly better than that of traditional scalar quantization (SQ-CE) and vector quantization (VQ-CE) schemes. For example, when the actual pilot transmission SNR is 45dB and the quantization bit is 4, the MSE of the TQ-CE scheme is 6.1×10 lower than that of the 18-bit SQ-CE scheme. -4 , which is 2.78×10 lower than the 4-bit VQ-CE solution. -3 .

[0131] according to Figure 5 It can be seen that when the training SNR is 10dB: TQ-CE can still achieve a lower MSE with fewer quantization bits. For example: when the actual pilot signal-to-noise ratio is 45dB and the quantization bit is 4, the MSE under the TQ-CE scheme is reduced by 4×10 compared with the 18-bit SQ-CE scheme.-3 , which is 1.07×10 lower than the 4-bit VQ-CE solution. -3 .

[0132] according to Figure 6 It can be seen that when the training SNR is 5dB: the channel capacity of TQ-CE is significantly higher than that of the comparison scheme. For example, when the pilot SNR is 45dB, the channel capacity of the 4-bit TQ-CE scheme is about 0.25Mbit / s higher than that of the 18-bit SQ-CE scheme, and about 0.62Mbit / s higher than that of the 4-bit VQ-CE scheme.

[0133] according to Figure 7 It can be seen that when the training SNR is 10dB, the efficiency of the TQ-CE scheme is further reflected. For example, when the pilot SNR is 45dB, the channel capacity of the 4-bit TQ-CE scheme is increased by about 1.04Mbit / s compared with the 18-bit SQ-CE scheme, and by about 0.31Mbit / s compared with the 4-bit VQ-CE scheme.

[0134] The above simulation results show that the TQ-CE scheme can significantly reduce the channel estimation error and improve the system channel capacity at a lower feedback overhead (a small number of quantization bits). Compared with the traditional SQ-CE and VQ-CE schemes, TQ-CE has better performance under the same conditions and is a better choice for processing CSI feedback in MIMO VLC systems.

[0135] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A task-based channel estimation quantization method, characterized in that: The specific steps are as follows: Step 1: Use the LOS channel gain between the user end and the LED transmitter to calculate the VLC signal matrix Y received by the downlink channel user VLC : Y VLC =H VLC Φ T +W H VLC is the real channel matrix, W is the real noise matrix, and Φ is the transmitted signal; Step 2: VLC signal matrix Y VLC Convert to vector form and get the pilot received signal vector y VLC : Represents the identity matrix; vec() represents the vectorization function of the matrix; represents the Kronecker product; Step 3: Use the VQ algorithm to transform the vector y VLC Mapped to the codebook, and quantized using Bussgang decomposition to obtain the quantized output code c n and its index vector i n ; Step 4: Index vector i n Uplink transmission, the receiving end obtains the signal matrix y RF : y RF =H RF P RF i n +N H RF is the RF channel matrix; P RF is the uplink signal precoding matrix; N is the noise matrix at the receiving end; Step 5: Using the signal matrix y RF Find an estimate of the index vector Using the codebook to obtain an estimate of the channel quantization output code Step 6: Introduce the minimum mean square error (MSE) of channel estimation and solve to obtain the optimal solution of the post-processing matrix D, so that the channel estimation error is minimized and the channel state is optimal. The specific steps are as follows: Step 601: Define the real channel vector s and the estimated value of s They are: Step 602: Introduction For a given y VLC The minimum mean square error estimate of time s; Γ is the linear transformation matrix, defined as Step 603: Using channel information s and its estimated value Calculate the minimum value of the mean square error MSE, that is: Step 604: Derivative the second term of the mean square error MSE expansion to obtain an optimal solution of the post-processing matrix D that minimizes the uplink CSI quantization error; First, the second term is expanded as: Taking the derivative, we get: Let the derivative result be equal to 0, and get the optimal solution D of the post-optimization matrix: y index Represents the quantized index vector.

2. A task-based channel estimation quantization method as claimed in claim 1, characterized in that: In step 3, the mapping formula is: c n =Q B (y VLC ) Q B is a quantization function that transforms the vector y VLC Quantized into a codeword c in the codebook n ;c n represents the nth code in the codebook set B; Using Bussgang decomposition, we get the quantitative results: c n =Gy VLC +e Where e is a zero-mean random variable and G is the Bussgang gain, calculated as: Indicates the code c n and VLC The cross-correlation matrix, C yVLC Represents vector y VLC The autocorrelation matrix of .

3. A task-based channel estimation and quantization method as claimed in claim 1, characterized in that: In step 5, first, the estimated value of the index vector is calculated The formula is: Π is the projection operation function, Used to solve the pseudo-inverse of a matrix; Then, the estimated channel quantization output code is recovered from the received index vector according to the codebook And use the post-processing matrix D to get the estimated channel matrix

Images