A method for acquiring massive MIMO-OFDM channel information based on time-frequency two-dimensional pilots

Abstract

This invention discloses a method for acquiring channel information in large-scale MIMO-OFDM based on time-frequency two-dimensional pilots. It proposes a time-frequency two-dimensional pilot method for wireless communication systems, where the time-frequency two-dimensional pilot sequence consists of a frequency-domain pilot sequence and a time-domain pilot sequence. Both the frequency-domain and time-domain pilot sequences are obtained by modulating a Zadoff-Chu sequence using a phase-shift sequence. Utilizing a triple-beambase channel tensor model for large-scale MIMO-OFDM systems, the invention proposes a method for acquiring channel information based on time-frequency two-dimensional pilots. Each user terminal transmits a known two-dimensional pilot signal to the base station on selected time and frequency resources. The base station uses the received time-frequency signals to perform channel estimation and prediction using an information geometry algorithm to obtain the channel information for each user. The proposed method can significantly improve the pilot capacity and channel information acquisition accuracy of large-scale MIMO-OFDM systems, especially exhibiting superior performance in communication scenarios with a large number of users and high connection density.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and in particular to a method for acquiring large-scale MIMO-OFDM channel information based on time-frequency two-dimensional pilots. Background Technology

[0002] The performance of a large-scale multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) system is closely related to the quality of channel state information (CSI) acquisition. Conventional CSI acquisition methods employ pilot-assisted channel acquisition, where the transmitting device sends pilot signals, and the receiving device uses the received pilot signals to estimate channel parameters. Therefore, pilot design and channel estimation have become important aspects of theoretical research in large-scale MIMO-OFDM channel acquisition.

[0003] Phase-Shifted Orthogonal Pilot (PSOP), a conventional pilot scheme, has been widely used in 4G and 5G mobile communication systems. However, the phase-shift orthogonality of PSOP leads to a linear increase in pilot overhead with the number of users, constituting a system bottleneck. To address this issue, Adjustable Phase-Shifted Pilot (APSP) was proposed, generating more usable pilots for users within a limited overhead, thus improving the system's pilot capacity. Although APSP no longer satisfies phase-shift orthogonality, under the space-frequency beamforming channel model, by exploiting the channel's distribution characteristics in the angle and delay domains, APSP can achieve the same estimation performance as PSOP under certain conditions. With the rapid increase in user capacity and connection density in future 6G mobile communication systems, higher demands are placed on pilot capacity. Continuing to use existing pilot methods will result in larger pilot periods or pilot interference; therefore, new pilot methods are urgently needed to meet the requirements of 6G systems.

[0004] The receiving device can obtain the optimal estimation result by using the minimum mean square error (MMSE) estimation based on the received pilot signal. However, in large-scale MIMO-OFDM systems, MMSE estimation involves inversion operations, leading to high computational complexity. Therefore, based on the space-frequency beamforming channel model, the channel sparsity in the angle and time delay domains is utilized to transform the original channel estimation problem into a sparse signal recovery problem. Statistical inference methods, such as confidence propagation, expectation propagation, and approximate message propagation algorithms, are then used to solve this problem, reducing the complexity of channel estimation. With the continuous development of mobile communication systems, the number of base station antennas and users will increase exponentially, further increasing the difficulty of channel estimation. At the same time, the multiple sparsity of large-scale MIMO-OFDM channels along the angle, time delay, and Doppler domains also makes it possible to estimate channel information for a large number of users with limited time-frequency resources. Therefore, how to fully utilize the sparsity characteristics of the channel and adopt appropriate channel estimation methods to complete channel estimation with lower complexity is crucial for 6G system design. Summary of the Invention

[0005] This invention provides a method and system for acquiring large-scale MIMO-OFDM channel information based on time-frequency two-dimensional pilots, which improves the system's pilot capacity and channel information acquisition accuracy to meet the needs of 6G mobile communication systems.

[0006] A first aspect of the present invention provides a time-frequency two-dimensional pilot method, comprising the following steps:

[0007] In a wireless communication system, the transmitting device sends a known two-dimensional signal to the receiving device on selected time and frequency resources. This two-dimensional signal is called a time-frequency two-dimensional pilot.

[0008] The receiving device uses the received time-frequency two-dimensional pilot signal to perform channel estimation and obtain channel parameters.

[0009] Optionally, in one embodiment of the present invention, the time-frequency two-dimensional pilot consists of a frequency domain pilot sequence and a time domain pilot sequence, and a time-frequency phase-shifted pilot is adopted, that is, both the frequency domain pilot sequence and the time domain pilot sequence are obtained by modulating the Zadoff-Chu sequence through a phase-shifted sequence; when the time-frequency two-dimensional pilot is a time-frequency phase-shifted pilot, different transmitting devices send time-frequency phase-shifted pilots with different phase shifts or the same phase shift to the receiving device.

[0010] A second aspect of the present invention provides a method for acquiring large-scale MIMO-OFDM channel information based on time-frequency two-dimensional pilots, comprising the following steps:

[0011] In large-scale MIMO-OFDM systems, the space-frequency-time domain channel is represented by a tensor, and the angle-delay-Doppler domain is sampled to establish a triple beambase channel tensor model for the space-frequency-time domain channel. The space-frequency-time domain channel is characterized as the modulus product of the space-time-frequency beam matrix and the triple beam domain channel.

[0012] Using the statistical channel information of each user terminal in the triple beam domain, the base station schedules the corresponding time-frequency two-dimensional pilot signal for each user terminal.

[0013] All user terminals send scheduled time-frequency two-dimensional pilot signals to the base station on the selected time and frequency resources. The base station estimates the triple beam domain channel based on the received signals.

[0014] The estimated triple beam domain channel is mapped to the space-frequency-time domain for channel estimation and prediction.

[0015] Optionally, in one embodiment of the present invention, during the sampling of the angle-delay-Doppler domain, the angle domain sampling is performed in the direction cosine domain, the direction cosine domain range is [-0.5, 0.5), and the ranges of the delay domain and the Doppler domain are both related to the symbol time length; the direction cosine domain range, the delay domain range, and the Doppler domain range are uniformly segmented for sampling, and the corresponding number of uniform sampling points is greater than or equal to the number of antenna elements on the base station side, the number of effective subcarriers, and the number of time slots in one frame; the sampled and quantized angle-delay-Doppler domain is called the triple beam domain, and the quantized angle domain, delay domain, and Doppler domain are called the spatial beam domain, the frequency beam domain, and the time beam domain, respectively.

[0016] Optionally, in one embodiment of the present invention, the spatial-frequency-time domain channel is a large-scale MIMO-OFDM transmission channel on the current frame, which is composed of the current time slot and multiple previous time slots; the current frame is composed of multiple time slots, each time slot contains multiple OFDM symbols, of which one or more OFDM symbols are used to transmit pilot signals, and the remaining symbols are used for data transmission; within the current frame, the spatial-frequency-time domain channel between each user terminal and the base station is represented as a three-dimensional tensor, the size of which corresponds to the number of antenna elements on the base station side, the number of effective subcarriers, and the number of OFDM symbols in the frame, respectively.

[0017] Optionally, in one embodiment of the present invention, the triple beam base channel tensor is a three-dimensional tensor, with the three dimensions corresponding to the number of spatial beams, the number of frequency beams, and the number of time beams, respectively. Each element represents the channel gain on the corresponding triple beam, and the channel gains between different beams are independent of each other.

[0018] The space-frequency-time domain channel tensor is represented as the modulus product of the triple beam domain channel tensor and the space-time-frequency beam matrix. The space-time-frequency beam matrices are the space beam matrix, the frequency beam matrix, and the time beam matrix, respectively, which are composed of the sampled space domain rudder vector, the sampled frequency domain rudder vector, and the sampled time domain rudder vector. The rudder vectors correspond to the space beam, the frequency beam, and the time beam, respectively pointing to the sampling angle, the time delay, and the Doppler frequency shift. The matrices formed by these vectors are the space beam matrix, the frequency beam matrix, and the time beam matrix, respectively.

[0019] Optionally, in one embodiment of the present invention, the statistical channel information in the triple beam domain is the power distribution of the triple beam domain channel, represented as a three-dimensional tensor with the same dimension as the triple beam domain channel tensor and having sparsity; the statistical channel information in the triple beam domain is obtained through a channel map, which is a channel knowledge database with location index, or obtained by using a probe signal through an online statistical channel information acquisition method.

[0020] Optionally, in one embodiment of the present invention, when the time-frequency two-dimensional pilot is a time-frequency phase-shift pilot, the step of the base station scheduling the corresponding time-frequency two-dimensional pilot signal for each user terminal is as follows:

[0021] All user terminals in a large-scale MIMO-OFDM system are grouped according to the triple beam domain statistical channel information of different user terminals. User terminals whose triple beam domain power distributions do not overlap or whose overlap is less than a set threshold are grouped together. User terminals in the same group can reuse the same time-frequency two-dimensional pilot.

[0022] After the user terminals are grouped, pilot frequencies are assigned to the user terminals in different groups so that the equivalent power distribution of the triple beam domain between different user terminal groups does not overlap or the overlap is less than a set threshold.

[0023] The triple beam domain equivalent power distribution refers to the channel tensor obtained by shifting all elements of the triple beam domain channel power distribution tensor simultaneously along the frequency beam domain and the time beam domain. The shift length and shift direction are determined by the phase shift factor of the time-frequency phase shift pilot.

[0024] Overlap refers to the linear correlation between two power distribution tensors in the triple beam domain. When the positions of the non-zero elements in the two power distribution tensors are completely offset, the overlap is 0.

[0025] Optionally, in one embodiment of the present invention, the triple beam channel estimation employs a tensor-based information geometry algorithm. By constructing a target manifold and an auxiliary manifold, the posterior probability density of the triple beam domain channel information is projected onto the target manifold to obtain the target probability density distribution. In this algorithm, the multiplication operations involved in the spatial time-frequency beam matrix are implemented quickly using FFT. The expected value of the target probability density distribution is used as the estimate of the triple beam domain channel tensor.

[0026] Optionally, in one embodiment of the present invention, channel estimation and channel prediction utilize the mapping relationship between the space-frequency-time domain channel and the triple beam domain channel to perform a modulus product operation on the estimated value of the triple beam domain channel tensor and the space-time-frequency beam matrix to obtain the channel information of the pilot band and data segment in the current frame, which are respectively used as the results of channel estimation and channel prediction.

[0027] A third aspect of the present invention provides a large-scale MIMO-OFDM channel information acquisition system based on time-frequency two-dimensional pilot signals, including a base station and multiple user terminals; the base station is used to establish a triple-beambase channel tensor model of the space-frequency-time domain channel, representing the space-frequency-time domain channel as the modulus product of the space-time-frequency beam matrix and the triple-beam domain channel; using the statistical channel information of each user terminal in the triple-beam domain, corresponding time-frequency two-dimensional pilot signals are scheduled for each user; in the uplink, the triple-beam domain channel tensor is estimated based on the received signal, and the estimated triple-beam domain channel is mapped to the space-frequency-time domain to complete channel estimation and channel prediction;

[0028] The user terminal is used to send known time-frequency two-dimensional pilot signals to the base station on selected time and frequency resources in the uplink.

[0029] The method for acquiring large-scale MIMO-OFDM channel information based on time-frequency two-dimensional pilots according to the embodiments of the present invention has the following beneficial effects:

[0030] 1. This invention proposes a time-frequency two-dimensional pilot method, which improves the pilot capacity of the system without increasing pilot overhead.

[0031] 2. The pilot scheduling algorithm proposed in this invention schedules corresponding pilot signals for different users in a large-scale MIMO-OFDM system with low complexity, effectively suppressing interference between users and improving channel estimation performance.

[0032] 2. This invention establishes a channel tensor model and proposes a tensor-based information geometry channel estimation algorithm, which fully utilizes the sparsity of the channel in the triple beam domain and the structural characteristics of the triple beam matrix, effectively reducing the complexity of large-scale MIMO-OFDM channel estimation.

[0033] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0034] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0035] Figure 1 This is a schematic diagram of the frame structure according to an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of a two-dimensional time-frequency pilot signal in an embodiment of the present invention;

[0037] Figure 3 A flowchart illustrating a method for acquiring large-scale MIMO-OFDM channel information based on time-frequency two-dimensional pilots according to an embodiment of the present invention;

[0038] Figure 4 This is a performance comparison chart between the channel acquisition algorithm proposed in this embodiment and existing channel acquisition algorithms;

[0039] Figure 5 This is a performance graph of the channel prediction algorithm proposed in the embodiments of the present invention;

[0040] Figure 6 This is a comparison chart of the computational complexity of the channel estimation algorithm proposed in this embodiment of the invention and existing channel estimation algorithms;

[0041] Figure 7 This is a schematic diagram of the electronic device structure according to an embodiment of the present invention. Detailed Implementation

[0042] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0043] To address the shortcomings of existing channel acquisition technologies and meet the evolving needs of future large-scale MIMO-OFDM systems, this invention proposes a time-frequency two-dimensional pilot method for wireless communication systems, and a large-scale MIMO-OFDM channel information acquisition method based on this method. In the time-frequency two-dimensional pilot method, the system's pilot capacity is improved by extending the frequency domain pilot method to the time-frequency two-dimensional domain. In the large-scale MIMO-OFDM channel information acquisition method based on the time-frequency two-dimensional pilot, a triple-beambase channel tensor model for the large-scale MIMO-OFDM system is first established. The spatial-frequency-time domain channel between each user and the base station is represented as the modulus product of the triple-beam domain channel tensor and the space-time-frequency beam matrix. Time-frequency two-dimensional pilots are designed, and a corresponding pilot scheduling algorithm is proposed to schedule appropriate time-frequency two-dimensional pilots for different user terminals. Furthermore, utilizing the special structure of the beam matrix, a tensor-based information geometry algorithm is adopted to complete channel estimation and prediction with low complexity, obtaining the channel information for each user. Compared with existing methods, the proposed channel acquisition method significantly improves the pilot capacity and channel information acquisition accuracy of the system, and has superior performance.

[0044] This invention provides a two-dimensional time-frequency pilot method, comprising the following steps:

[0045] In a wireless communication system, the transmitting device sends a known two-dimensional signal to the receiving device on selected time and frequency resources. This two-dimensional signal is called a time-frequency two-dimensional pilot.

[0046] The receiving device uses the received time-frequency two-dimensional pilot signal to perform channel estimation and obtain channel parameters.

[0047] The time-frequency two-dimensional pilot consists of a frequency domain pilot sequence and a time domain pilot sequence. It adopts a time-frequency phase-shifted pilot, that is, both the frequency domain pilot sequence and the time domain pilot sequence are obtained by modulating the Zadoff-Chu sequence with a phase-shifted sequence. When the time-frequency two-dimensional pilot is a time-frequency phase-shifted pilot, different transmitting devices send time-frequency phase-shifted pilots with different or the same phase shifts to the receiving device.

[0048] Figure 1 A schematic diagram of the frame structure used in an embodiment of the present invention is given. The current frame consists of the current time slot and the previous (N) time slot. p -1) time slots, i.e., N p It is composed of N time slots. b There are 3 OFDM symbols, with the first OFDM symbol used for pilot transmission and the remaining OFDM symbols used for data transmission. Combining the above frame structure, Figure 2 A schematic diagram of the time-frequency two-dimensional pilot method disclosed in an embodiment of the present invention is provided. The time-frequency two-dimensional pilot is transmitted along the effective subcarriers in the frequency domain and the pilot segment on the time-domain radio frame. (See figure.) This represents the frequency domain pilot sequence corresponding to transmitting device u; on the nth pilot band, the sequence currently transmitted by the user along the effective subcarrier is... That is, in the frequency domain pilot sequence Multiplied by a time-related coefficient For different pilot bands n, the corresponding The values ​​of can be the same or different. Along the time domain, the coefficients on all pilot bands... It can form a continuous time-domain pilot sequence. The aforementioned frequency-domain pilot sequences and time-domain pilot sequences together constitute a two-dimensional time-frequency pilot. The two-dimensional time-frequency pilot can be seen as an extension of the traditional frequency-domain pilot to a two-dimensional time-frequency system. Correspondingly, the frequency-domain pilot method can also be seen as a two-dimensional time-frequency pilot method that holds true for all non-negative integers n. An exception at the time of its establishment.

[0049] Figure 3 This is a flowchart illustrating a method for acquiring large-scale MIMO-OFDM channel information based on time-frequency two-dimensional pilots according to an embodiment of the present invention.

[0050] like Figure 3 As shown, the method for acquiring large-scale MIMO-OFDM channel information based on time-frequency two-dimensional pilots includes the following steps:

[0051] In step S101, in a large-scale MIMO-OFDM system, the space-frequency-time domain channel is represented by a tensor, the angle-delay-Doppler domain is sampled, a triple beam base channel tensor model of the space-frequency-time domain channel is established, and the space-frequency-time domain channel is characterized as the modulus product of the space-time-frequency beam matrix and the triple beam domain channel.

[0052] In step S102, the base station schedules corresponding time-frequency two-dimensional pilot signals for each user terminal using the statistical channel information of each user terminal in the triple beam domain.

[0053] In step S103, all user terminals send scheduled time-frequency two-dimensional pilot signals to the base station on the selected time and frequency resources. The base station estimates the triple beam domain channel based on the received signals.

[0054] In step S104, the estimated triple beam domain channel is mapped to the space-frequency-time domain to perform channel estimation and channel prediction.

[0055] In a wireless communication system, the pilot signals of each user terminal are time-frequency two-dimensional signals. The user terminal sends the known time-frequency two-dimensional signals to the receiving device on the selected time and frequency resources. The base station uses the received time-frequency two-dimensional pilot signals to estimate the channel parameters.

[0056] Optionally, in one embodiment of the present invention, the spatial-frequency-time domain channel is a large-scale MIMO-OFDM transmission channel on the current frame, which is composed of the current time slot and multiple previous time slots; the current frame is composed of multiple time slots, each time slot contains multiple OFDM symbols, of which one or more OFDM symbols are used to transmit pilot signals, and the remaining symbols are used for data transmission; within the current frame, the spatial-frequency-time domain channel between each user terminal and the base station is represented as a three-dimensional tensor, the size of which corresponds to the number of antenna elements on the base station side, the number of effective subcarriers, and the number of OFDM symbols in the frame, respectively.

[0057] Optionally, in one embodiment of the present invention, during the sampling of the angle-delay-Doppler domain, the angle domain sampling is performed in the direction cosine domain, the direction cosine domain range is [-0.5, 0.5), and the ranges of the delay domain and the Doppler domain are both related to the symbol time length; the direction cosine domain range, the delay domain range, and the Doppler domain range are uniformly segmented for sampling, and the corresponding number of uniform sampling points is greater than or equal to the number of antenna elements on the base station side, the number of effective subcarriers, and the number of time slots in one frame; the sampled and quantized angle-delay-Doppler domain is called the triple beam domain, and the quantized angle domain, delay domain, and Doppler domain are called the spatial beam domain, the frequency beam domain, and the time beam domain, respectively.

[0058] Optionally, in one embodiment of the present invention, the triple beam-based channel tensor is a three-dimensional tensor, with the three dimensions corresponding to the number of spatial beams, the number of frequency beams, and the number of time beams, respectively. Each element represents the channel gain on the corresponding triple beam, and the channel gains between different beams are independent of each other. The triple beam domain channel tensor is a sparse tensor, that is, it has only a few non-zero elements.

[0059] Optionally, in one embodiment of the present invention, the space-frequency-time domain channel tensor is represented as the modulus product of the triple beam domain channel tensor and the space-time-frequency beam matrix; the space-time-frequency beam matrix is ​​the space beam matrix, the frequency beam matrix, and the time beam matrix, respectively, which are composed of the sampled space domain rudder vector, the sampled frequency domain rudder vector, and the sampled time domain rudder vector, respectively. The rudder vectors correspond to the space beam, the frequency beam, and the time beam, respectively pointing to the sampling angle, the time delay, and the Doppler frequency shift, and the matrices formed are the space beam matrix, the frequency beam matrix, and the time beam matrix, respectively.

[0060] Optionally, in one embodiment of the present invention, the statistical channel information in the triple beam domain is the power distribution of the triple beam domain channel, represented as a three-dimensional tensor with the same dimension as the triple beam domain channel tensor and having sparsity; the statistical channel information in the triple beam domain is obtained through a channel map, which is a channel knowledge database with location index, or obtained by using a probe signal through an online statistical channel information acquisition method.

[0061] Optionally, in one embodiment of the present invention, when the time-frequency two-dimensional pilot is a time-frequency phase-shift pilot, the step of the base station scheduling the corresponding time-frequency two-dimensional pilot signal for each user terminal is as follows:

[0062] All user terminals in a large-scale MIMO-OFDM system are grouped according to the triple beam domain statistical channel information of different user terminals. User terminals whose triple beam domain power distributions do not overlap or whose overlap is less than a set threshold are grouped together. User terminals in the same group can reuse the same time-frequency two-dimensional pilot.

[0063] After the user terminals are grouped, pilot frequencies are assigned to the user terminals in different groups so that the equivalent power distribution of the triple beam domain between different user terminal groups does not overlap or the overlap is less than a set threshold.

[0064] The triple beam domain equivalent power distribution refers to the channel tensor obtained by shifting all elements of the triple beam domain channel power distribution tensor simultaneously along the frequency beam domain and the time beam domain. The shift length and shift direction are determined by the phase shift factor of the time-frequency phase shift pilot.

[0065] Overlap refers to the linear correlation between two power distribution tensors in the triple beam domain. When the positions of the non-zero elements in the two power distribution tensors are completely offset, the overlap is 0.

[0066] Optionally, in one embodiment of the present invention, the triple beam channel estimation employs a tensor-based information geometry algorithm. By constructing a target manifold and an auxiliary manifold, the posterior probability density of the triple beam domain channel information is projected onto the target manifold to obtain the target probability density distribution. In this algorithm, the multiplication operations involved in the spatial time-frequency beam matrix are implemented quickly using FFT. The expected value of the target probability density distribution is used as the estimate of the triple beam domain channel tensor.

[0067] Optionally, in one embodiment of the present invention, channel estimation and channel prediction utilize the mapping relationship between the space-frequency-time domain channel and the triple beam domain channel to perform a modulus product operation on the estimated value of the triple beam domain channel tensor and the space-time-frequency beam matrix to obtain the channel information of the pilot band and data segment in the current frame, which are respectively used as the results of channel estimation and channel prediction.

[0068] The following describes in detail the specific implementation process of the large-scale MIMO-OFDM channel information acquisition method based on time-frequency two-dimensional pilots involved in this invention, using specific communication system examples. It should be noted that the method of this invention is not only applicable to the specific system model mentioned in the example below, but also to other system models with different configurations.

[0069] I. System Configuration

[0070] Consider a single-cell time-division duplex (TDD) massive MIMO-OFDM system with carrier frequency f. c The system includes one base station, U single-antenna users, and a user set. The base station is equipped with an M-element uniform linear array, with an element spacing of half a wavelength. The system has N subcarriers. c The length of the cyclic prefix is ​​N g The subcarrier spacing is Δf, and the sampling interval is T. s The length of a single OFDM symbol is T. sym =(N c +N g )T s The set of indices of effective subcarriers is defined along the frequency domain as follows: Where K represents the number of effective subcarriers.

[0071] according to Figure 3 As shown, each data frame contains N. s =N b *N p There are N OFDM symbols, of which N are used as transmit pilots. p One. Assuming the current time slot is T, let... Then from the nth T The current frame is formed by the Tth time slot. The set of sequence numbers of all OFDM symbols within the current frame can be represented as: The set of all OFDM symbol indices in the pilot portion of the current frame can be represented as:

[0072] II. Triple-beambase channel tensor model

[0073] Because the channel model in this embodiment has a high dimensionality, the following tensor algebra is introduced for ease of subsequent description: For tensors When M = K and for m = 1, 2, ..., M, I m =J m At the time of its establishment, it was called Let i be a square tensor. When i1 = j1, i2 = j2, ..., i M =j M At that time, element This is called a pseudo-diagonal element. When When all elements except the pseudo-diagonal element are 0, This is a pseudo-diagonal tensor. Similar to matrix operations, the superscript (·) in a tensor... -1 and(·) * These also represent the inverse and conjugate operations, respectively. Definition M-transpose in The M-conjugate transpose is in For two quantities definition and The Einstein product is in:

[0074]

[0075] For tensors sum matrix definition The n-modulus product of X is in

[0076]

[0077] In addition, the following matrix is ​​defined in this embodiment:

[0078]

[0079] Assume the number of transmission paths between user u and the base station is P. u If the m-th array element of the base station is a given element, then the channel impulse response between the user and the m-th array element can be expressed as:

[0080]

[0081] Wherein, parameter α u,p , τ u,p ,ν u,p These represent the gain, direction cosine, time delay, and Doppler shift corresponding to the p-th path, respectively. With path incident angle θ u,p Between Assume that the channel information remains constant within one OFDM symbol time, but varies between symbols due to the Doppler effect; let x u,k,n The signal transmitted by user u on the k-th subcarrier of the n-th OFDM symbol, after OFDM modulation, can be represented as the received signal of the m-th array element on the k-th subcarrier of the n-th OFDM symbol at the base station:

[0082]

[0083] in, Let represent additive white Gaussian noise in the frequency domain. Let the signal response of user u and the m-th array element on the base station side be represented on the k-th subcarrier of the n-th OFDM symbol. Then:

[0084]

[0085] Since the antenna element spacing on the base station side in this embodiment is half a wavelength, the value of the direction cosine in this embodiment satisfies... Both the time delay domain range and the Doppler domain range are related to the symbol time length. According to the configuration requirements of the OFDM system, the time delay value satisfies τ. u,p ∈[0,1 / Δf]. Furthermore, assuming all users in the system are synchronized, v speed This represents the maximum user movement speed in the system, and defines parameters. The Doppler frequency shift value satisfies ν u,p ∈[-ν max / 2,ν max / 2]. Set N b <1 / (T) sym ν max If the range of values ​​for the Doppler frequency shift mentioned above is ν, then the range can be expanded to ν. u,p ∈[-1 / (2N b T sym ),1 / (2N b T sym )].definition and sets in N τ and N ν Let represent the number of elements in the three sets respectively, then the above and This can be achieved by applying the parameters [-0.5, 0.5], [0, 1 / Δf], and [-1 / (2N] respectively. b T sym ),1 / (2N b T sym Obtained by uniform segmentation sampling. When N τ and N ν When large enough, the discrete interval and It can be used in the approximate angular time delay domain, time delay domain, and Doppler frequency shift domain, respectively. In this case, let... And if this set is taken as the range of the triple beam (TB) domain, then the TB domain can be used to approximate the angle-time delay-Doppler domain. In summary, let... parameter Can be used for approximation τ u,p ν u,p Define a set Then there is Established.

[0086] according to Figure 3 Given the frame structure and the aforementioned channel model, assuming the current time slot is T, for any... τ∈S τ and ν∈S ν Define the spatial domain rudder vector, frequency domain rudder vector, and time domain rudder vector as follows:

[0087]

[0088]

[0089]

[0090] The three rudder vectors mentioned above correspond to beams in the spatial, frequency, and time domains, respectively, and point towards the direction cosine. The time delay τ and the Doppler frequency shift ν. Therefore, the quantized angle domain, time delay domain, and Doppler domain can be... and These are respectively referred to as the spatial beam domain, frequency beam domain, and time beam domain. Those skilled in the art will understand that the specific vector representation in the above model is only based on a uniform linear array as an example. For systems employing different antenna arrays such as uniform area arrays or uniform circular arrays, only the vector representation needs to be changed. Simply change it to the corresponding spatial domain rudder vector.

[0091] Using the rudder vectors in (7)-(9), we can define M×K×N as follows. s dimensional tensor:

[0092]

[0093] Among them, operators The outer product operation is represented by the (m,k,n)th element of the tensor. Therefore, the above tensor This can be used to represent the spatial-frequency-time (SFT) domain channel between user u and the base station in the current frame. Further, the following function is defined:

[0094]

[0095] Then tensor This can be further expressed as:

[0096]

[0097] Let matrix The SFT domain channel in the text can also be written as

[0098]

[0099] Among them, tensor The The elements are This can be called the TB-domain channel tensor, whose three dimensions correspond to the number of spatial beams, frequency beams, and time beams, respectively. Each element represents the channel gain on the corresponding triple beam, and the channel gains between different beams are independent of each other. Since V s 、V f and Each column corresponds to a physical beam in the spatial domain, frequency domain, or time domain. Therefore, in this embodiment, the above three matrices are called the spatial beam matrix, frequency beam matrix, and time beam matrix, or simply the space-time-frequency beam matrix. At this time, (13) can be called the triple beam base channel model, that is, the SFT domain channel tensor is expressed as the modulus product of the TB channel tensor and the space-time-frequency beam matrix.

[0100] Furthermore, the following tensor is defined: in:

[0101]

[0102] Then (13) can be further expressed as

[0103]

[0104] Among them, the operator * n Represents the Einstein product of tensors. It can be called a triple beam tensor.

[0105] like Figure 3 As shown, the current frame includes a total of N p There are 1 pilot band, and each pilot band consists of a single OFDM symbol. The SFT domain channel corresponding to the pilot band in the current frame can also be represented in tensor form:

[0106]

[0107] in, The tensor represents the time-domain beam matrix corresponding to the pilot band. The triple beam tensor corresponding to the pilot band satisfies Specifically, when the number of sampling points in the beam domain meets the following conditions... N τ =F τ K, N ν =F ν N p Among them, refinement factor Then the beam domain matrix V s 、V f and Both can be generated using a DFT matrix. Now, let F... N Represents an N×N dimensional DFT matrix. Representing the cyclic column shift of the DFT matrix, the beam matrix can be generated as follows:

[0108] Assuming that the channel gains on different beams all satisfy a cyclic symmetric complex Gaussian distribution with a mean of 0 and are mutually independent, the following tensor can be defined.

[0109]

[0110] No. The element is the TB field. The channel power on each beam. Furthermore, The covariance can be expressed using tensors This is represented as follows. Correspondingly, the TB-domain covariance can be expressed as... This tensor is a pseudo-diagonal tensor, and its diagonal elements According to (15)-(17), the covariance tensors of the TB domain and the SFT domain satisfy the following relationship:

[0111]

[0112] Based on the channel model described above, the TB domain channel tensor The three dimensions correspond to the spatial beam domain, frequency beam domain, and time beam domain, respectively. Since the channel delay spread is less than the length of the cyclic prefix, the following definition is used: Cyclic prefix length but The coordinates of the non-zero elements along the frequency beam domain are concentrated in [0, F]. τ N f Within the range. Meanwhile, due to Doppler extension... in Therefore The coordinates of the non-zero elements along the time beam domain are concentrated in [(N ν -N d F ν ) / 2:(N ν +N d F ν Within the range of ) / 2). In summary, It is a sparse tensor, and its non-zero elements are concentrated within a specific range along both the frequency and time beam domains. From (16), it can be seen that... and They share the same sparsity and, as a type of statistical channel information, It changes relatively slowly over time. (Utilizing...) The slow-varying and sparsity characteristics over time can be obtained using channel mapping methods or by acquiring online statistical channel information using probe signals. Therefore, in the pilot design and channel estimation stages, this embodiment assumes that for any user... It is known.

[0113] III. Pilot Design and Scheduling

[0114] When the current time slot is T, Figure 3 A schematic diagram is given of user u sending a two-dimensional time-frequency pilot signal to the base station in the current frame. This represents the frequency domain pilot sequence corresponding to user u, with coefficients on all pilot bands. It can form a continuous time-domain pilot sequence. The frequency domain pilot sequence and the time domain pilot sequence together constitute a two-dimensional time-frequency pilot.

[0115] In this embodiment of the invention, a time-frequency phase-shifted pilot (TFPSP) is used as a two-dimensional time-frequency pilot. Based on the channel model in (16), a frequency-domain pilot sequence is defined for user u. in This represents the pilot signal on the k-th effective subcarrier. Sequence It can be generated using phase-shifted sequences and Zadoff-Chu sequences as follows:

[0116]

[0117] Where, φ u ∈{0,1,…,K-1} represents the frequency domain phase shift factor. f Given a Zadoff-Chu sequence of length K. definition Then there is Established.

[0118] Coefficients along all pilot bands in the time domain It can form a continuous time-domain sequence. Similar to (17), this time-domain sequence can also be generated from Zadoff-Chu sequences and phase-shifted sequences, i.e.:

[0119]

[0120] Where γ represents N p Coprime positive integers, phase shift factor According to the pilot design method in (19), we have Established, that is The period length is Np A periodic sequence.

[0121] When the time slot is T, N in the current frame p Each pilot band corresponds to The time-domain pilot sequence that makes up the current frame according to Periodicity, sequence It can be generated in the following way:

[0122]

[0123] in, Phase shift factor x t To meet N p Point Zadoff-Chu sequence. (The sentence appears to be incomplete and requires further context.) definition Then there is Established.

[0124] sequence and Together, they form a TFPSP. Traditional frequency-domain phase-shift pilots can be considered as TFPSPs in... Right now This is a special case. By introducing a time-domain phase shift factor, TFPSP provides more available pilots for user terminal scheduling within a limited pilot overhead, further increasing the system's pilot capacity.

[0125] Assuming all users in the system simultaneously transmit TFPSPs to the base station on given time and frequency resources, the received signal at the base station can be represented in the following tensor form:

[0126]

[0127] Among them, tensor This indicates a TFPSP signal that satisfies... σ represents the received signal on the k-th subcarrier of the n-th pilot band in the current frame at the base station. p Indicates the transmitted signal power. This represents additive white Gaussian noise, where each element is independently and identically distributed.

[0128] Substituting (16) into (22), the received signal can be further expressed as:

[0129]

[0130] for definition:

[0131]

[0132] Under the minimum mean square error (MMSE) estimation criterion, the estimated value of the channel tensor in the TB domain for user u can be expressed as:

[0133]

[0134] in, The definition is as follows:

[0135]

[0136] Furthermore, define It can be represented as:

[0137]

[0138] According to (19) and (21), and The diagonal elements can all be generated from a single column of the DFT matrix, while the beam matrix V f and Both can be generated using DFT matrices, therefore the following relationship holds:

[0139] Define the following tensor transformation:

[0140]

[0141] Right now After all elements undergo the same position transformation, we get Analogous to the cyclic shift of a sequence, this embodiment refers to the above transformation (28) as a tensor cyclic shift, where a and b correspond to the cyclic shift lengths along its second and third dimensions, respectively. Similarly, the following cyclic shift transformation can be defined for the power distribution tensor:

[0142]

[0143] Using the definition in (28), Further writing is possible:

[0144]

[0145] in Substitute (30) into (27), It can be written as:

[0146]

[0147] in diagonal elements A pseudo-diagonal tensor with all other elements being 0. In summary, the transmission of TFPSP is equivalent to performing a cyclic shift transformation on the channel tensor in the TB domain. For ease of description, let's say... This is called the equivalent channel for user u in the TB domain. This is called the equivalent power distribution of user u in the TB domain.

[0148] Based on the mapping relationship between the SFT domain channel tensor and the TB domain channel tensor (16), the estimated... Subsequently, the SFT domain channel tensor estimate corresponding to the pilot band of the current frame can be expressed as:

[0149]

[0150] at this time With the actual channel tensor The mean square error between them is:

[0151]

[0152] According to the definition in (26), within the above error expression, pilot interference between different users is reflected in... In order to suppress interference between users and reduce the error of channel estimation, the following conclusions are given regarding ε. MSE The lower bound of the threshold, and the conditions that TFPSP must satisfy to reach that lower bound:

[0153] When M, K, N p →∞, if u≠u', therefore:

[0154]

[0155] If true, then ε MSE It can reach the minimum value ε MSE,min ,Right now:

[0156]

[0157] Compared to ε MSE , ε MSE,min It does not contain any cross-items between different users, and interference between users has been completely eliminated.

[0158] Equation (34) shows that the equivalent power distributions of different users do not overlap in the TB domain. Furthermore, the scheduling of TFPSPs can also follow the following criterion: by using appropriate pilot scheduling methods, the overlap of the equivalent power distributions between different users should be minimized, thereby reducing interference between users. According to this criterion, for two power distribution tensors of the same dimension... and Define its overlap:

[0159]

[0160] Then there is It is valid if and only if Sometimes, Therefore, condition (34) is equivalent to The aforementioned overlap can also be used to measure the linear correlation between two power distribution tensors in the triple beam domain. The overlap is zero when the non-zero elements in the two power distribution tensors are completely offset. Therefore, using overlap as the optimization objective, the scheduling of TFPSP can be expressed as the following optimization problem:

[0161]

[0162] The above problem (37) is a combinatorial optimization problem. In order to find an approximate optimal solution with low complexity, the scheduling of TFPSP can be performed according to the following algorithm:

[0163] (1) For all users in the system, all users are grouped according to the triple beam domain statistical channel information of different users. Users whose triple beam domain power distributions do not overlap or whose overlap is less than a certain threshold γ are grouped together. Users in the same group can reuse the same time-frequency two-dimensional pilot. The specific steps are as follows:

[0164] Step 1: Represent the overlapping graph of users and users in the system as an undirected graph. Its vertices correspond to each user, and its edges... That is, when the overlap between users is less than γ, the two users can be approximately considered as not overlapping in the TB domain; according to the undirected graph... The saturation algorithm is used to color the vertices of different users, so that adjacent users in the graph have different colors; the saturation of a vertex represents the sum of the color variations of all the colored vertices adjacent to that vertex; the specific steps of the saturation algorithm are as follows:

[0165] Step 2: For all Initialize c u =0 indicates that the vertices are not colored;

[0166] Step 3: Take the vertex u0 with the most neighboring users, and let...

[0167] Step 4: For all uncolored points, find the point u with the highest saturation, and let... Let the set of colors of the vertices surrounding the given vertex be denoted as .

[0168] Step 5: Repeat steps 2-4 until all users in the system have been colored; after coloring, the number of user groups is calculated. Group users with the same color into one group, and use each group to... This means that for i = 0, 1, ..., C-1, the power distribution is defined.

[0169] (2) After the user groups are completed, pilot signals are assigned to users in different groups so that the equivalent power distribution of the triple beam domain between different user groups does not overlap or the overlap is less than a certain threshold γ. The specific steps are as follows:

[0170] Step 1: Initialize the set of scheduled user groups Unscheduled user set as well as

[0171] Step 2: For If there exists φ∈{0,1,…,K-1}, Make Then there is otherwise

[0172] Step 3: For all And update the set

[0173] Step 4: Repeat steps 2 and 3 until all users have been scheduled.

[0174] IV. Channel Estimation and Prediction

[0175] Because tensor inversion is involved, in large-scale MIMO-OFDM systems, it can be directly calculated using equation (32). This will result in high computational complexity. To obtain [the desired result] with lower complexity... as well as This invention employs a tensor-based information geometry method.

[0176] Based on the received signal model in (23), the received signal at the base station can be further expressed as:

[0177]

[0178] in, It is a pseudo-diagonal tensor that satisfies The above The sum of the equivalent TB-domain channel tensors for all users, and their corresponding power distribution. According to the MMSE estimation criteria The estimated value can be expressed as:

[0179]

[0180] Among them, pseudo-diagonal tensor satisfy Comparing the estimation results in (39) and (25), we can establish and The relationship between them is as follows:

[0181]

[0182] Among them, superscript This represents the tensor obtained by taking the reciprocals of all non-zero elements in the tensor. According to (39)-(40), for All can be passed Therefore, the original channel estimation problem can be transformed into a computational problem involving the channel estimation function. The estimate.

[0183] Channel Tensor The power distribution can be viewed as Single-user TB domain channel. MMSE estimates and probability distribution Since their expectations are the same, therefore we have To simplify the representation, let b replace This represents the index of the TB domain. Based on the Gaussian distribution characteristics of additive white noise, the posterior probability density function can be expressed as:

[0184]

[0185] in, express The set of positions of non-zero elements, ψ p Let represent the normalization factor, and d represent the interaction term. Definition as well as Then there is in:

[0186]

[0187] Because d contains The addition and multiplication operations between different elements in a tensor are called cross terms. Define a tensor. for Sufficient statistics, of which as well as and tensors in If all other elements are 0, the posterior probability can be further expressed as:

[0188]

[0189] The existence of the cross term d makes The expectation operation is quite complex. To solve this problem, a target manifold is defined. Let be the set of probability density functions, and all probability density functions in this set can be represented in the following form:

[0190]

[0191] in, Represents the normalization factor, parameter NP represents a natural parameter. A dimensional tensor containing the statistical expectation, i.e., a first-order NP (FONP, First-Order Natural Parameter). And variance, i.e., second-order NP (SONP). The positions of non-zero elements in FONP and SONP must be consistent with... Keep it consistent. For ease of representation, let... compared to The probability density distribution in the target manifold contains no cross terms, making the calculation of its corresponding expectation relatively simple. According to information geometry theory, let:

[0192]

[0193] express Projecting onto m of the target manifold, then and They have the same expected value. Therefore, by solving... The expected value can reduce the computational complexity of MMSE estimation.

[0194] To solve with lower complexity Construct the following auxiliary manifold Each auxiliary manifold The probability density function in can be expressed as:

[0195]

[0196] Where, d n For intersecting terms, Represents the normalization factor. Represents NP, and similar, Include as well as For each Define its direction The m projection is Its NP is Include and Similarly, and They also have the same expectations. If at this point... View as If the approximation is d, then d n It can be approximated as

[0197]

[0198] in, And the cross term d can be approximated as Based on the above approximation, IGA uses a class of iterative algorithms to solve the problem. During the iteration of IGA, the NP of all auxiliary manifolds is first initialized, i.e. In the t-th iteration, NP is updated according to the following rules:

[0199]

[0200]

[0201] Where 0<α≤1 represents the oscillation factor during the iteration process.

[0202] To further simplify the iterative process of IGA, when N and When large enough, the NP values ​​of all auxiliary manifolds can be uniformly represented using... Approximately, the t-th iteration can be further simplified to:

[0203]

[0204]

[0205] Following the above iterative steps, when the initialization parameters satisfy... There is in each step of the iteration Established. Order The corresponding FONP is SONP is The detailed steps of the IGA algorithm are as follows:

[0206] Step 1: Set t = 0 to initialize

[0207] Step 2: Iterate as follows

[0208]

[0209]

[0210] in, Operators This indicates taking the reciprocal of all non-zero elements in the tensor. and All are pseudo-diagonal tensors, for all have besides and All other elements are 0;

[0211] Step 3: Update parameters t = t + 1;

[0212] Step 4: Repeat steps 2 and 3 until the algorithm converges, and obtain the result. The estimated value, and from this, we obtain The estimated value.

[0213] The implementation complexity of the above algorithm can be further reduced. According to (52) and (53), since most of the tensors involved in the algorithm are pseudo-diagonal, the complexity of the algorithm can be further reduced. and The Einstein product operation consumes most of the computation. To further reduce computational complexity, for arbitrary tensors... have:

[0214]

[0215] Based on the special structure of the sampling beam matrix, V s 、V f and All of these can be obtained from the DFT matrix; therefore, the Einstein product described above can be implemented using the FFT transformation. Similarly, for any tensor... have

[0216]

[0217] In addition, all other operations involved in IGA are related to pseudo-diagonal tensors. In summary, the time complexity of the IGA algorithm is O(n log n). in

[0218] For user u, obtained through IGA The estimated value Subsequently, based on the mapping relationship between the SFT domain channel and the TB domain channel, the pilot band channel estimate can be obtained by... Substituting into (15), we obtain the result. Simultaneously, channel prediction can also be performed using a similar mapping relationship; that is, in the data segment of the current time slot, the SFT domain channel can be obtained according to the following formula.

[0219]

[0220] V. Implementation Results

[0221] To enable those skilled in the art to better understand the present invention, the performance results of the channel information acquisition method in this embodiment under specific configurations are given below.

[0222] Consider a large-scale MIMO-OFDM system with the following system parameters: carrier frequency f c =5.8GHz, number of subcarriers N c = 2048, cyclic prefix length N g =144, subcarrier spacing Δf = 15kHz, effective subcarrier number K = 360, base station antenna element number M = 128, element spacing is half wavelength, number of users in the system U = 300, number of time slots N in one frame along the time domain p =8, N is the number of OFDM symbols in a single time slot. b =14, User terminal moving speed v speed =3km / h. Oversampling factor like Figure 4 As shown, based on the above parameter settings, the normalized mean square error performance of the following channel acquisition methods was compared under different signal-to-noise ratios: APSP-IGA: This method is based on the space-frequency beam domain channel model, and uses the IGA method for channel estimation to schedule frequency-domain adjustable phase-shifted pilots (APSP) for different users; TFPSP-IGA: This method is the proposed channel acquisition method, based on the triple beam base channel model, scheduling TFPSPs for different users, and using the IGA method for channel estimation; TFPSP-GAMP: This method adopts the same channel model and pilot scheduling algorithm as TFPSP-IGA, and uses the Generalized Approximate Message Passing (GAMP) algorithm for channel estimation; TFPSP-EPV: This method adopts the same channel model and pilot scheduling algorithm as TFPSP-IGA, and uses a variant of a type of expectation propagation algorithm for channel estimation. Figure 4 The number of iterations for all channel estimation algorithms was set to 300. As shown in the figure, the channel acquisition method based on time-frequency phase-shift pilot proposed in this invention has a significant performance improvement compared with other methods.

[0223] Figure 5The performance of the channel prediction algorithm proposed in this embodiment is given. The channel estimation algorithm adopts TFPSP-IGA. "with channel prediction" means that the SFT domain channel tensor of the current time slot data transmission segment is obtained using the method proposed in equation (56). "without channel prediction" means that the SFT domain channel tensor of the current time slot pilot segment is directly used as the channel tensor of the SFT domain of all OFDM symbols in the data transmission segment. Figure 5 This demonstrates the effectiveness of the channel prediction method proposed in this invention.

[0224] When selecting TFPSP as the time-frequency two-dimensional pilot signal Figure 6 The complexity comparison of the tensor-based IGA algorithm proposed in this invention with the GAMP algorithm, EPV algorithm, and traditional MMSE estimation algorithm is given. The IGA algorithm, GAMP algorithm, and EPV algorithm all require 300 iterations. Figure 6 As shown, by making full use of the special structure of the beam matrix, the IGA algorithm utilizes FFT operations, which significantly reduces computational complexity compared to other algorithms.

[0225] A second aspect of the present invention provides a large-scale MIMO-OFDM channel information acquisition system based on time-frequency two-dimensional pilot signals, including a base station and multiple user terminals; the base station is used to establish a triple-beambase channel tensor model of the space-frequency-time domain channel, representing the space-frequency-time domain channel as the modulus product of the space-time-frequency beam matrix and the triple-beam domain channel; using the statistical channel information of each user terminal in the triple-beam domain, corresponding time-frequency two-dimensional pilot signals are scheduled for each user; in the uplink, the triple-beam domain channel tensor is estimated based on the received signal, and the estimated triple-beam domain channel is mapped to the space-frequency-time domain to complete channel estimation and channel prediction;

[0226] The user terminal is used to transmit known time-frequency two-dimensional pilot signals to the base station on selected time and frequency resources in the uplink.

[0227] This invention proposes a method and system for acquiring channel information in large-scale MIMO-OFDM systems based on time-frequency two-dimensional pilot signals. Utilizing a triple-beambase channel tensor model for large-scale MIMO-OFDM systems, a method for acquiring channel information based on time-frequency two-dimensional pilot signals is proposed. Each user terminal transmits a known two-dimensional pilot signal to the base station on selected time and frequency resources. The base station uses the received time-frequency signal to perform channel estimation and prediction through an information geometry algorithm to obtain the channel information for each user. The proposed method can significantly improve the pilot capacity and channel information acquisition accuracy of large-scale MIMO-OFDM systems, exhibiting superior performance, especially in communication scenarios with a large number of users and high connection density.

[0228] Figure 7 This is a schematic diagram of an electronic device according to an embodiment of the present invention. The electronic device may include:

[0229] The memory 701, the processor 702, and the computer program stored on the memory 701 and executable on the processor 702.

[0230] When the processor 702 executes the program, it implements the large-scale MIMO-OFDM channel information acquisition method based on time-frequency two-dimensional pilot provided in the above embodiments.

[0231] Furthermore, electronic devices also include:

[0232] Communication interface 703 is used for communication between memory 701 and processor 702.

[0233] The memory 701 is used to store computer programs that can run on the processor 702.

[0234] The memory 701 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0235] If the memory 701, processor 702, and communication interface 703 are implemented independently, then the communication interface 703, memory 701, and processor 702 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 7Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0236] Optionally, in a specific implementation, if the memory 701, processor 702, and communication interface 703 are integrated on a single chip, then the memory 701, processor 702, and communication interface 703 can communicate with each other through an internal interface.

[0237] The processor 702 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0238] This embodiment also provides a computer-readable storage medium storing a computer program, characterized in that, when the program is executed by a processor, it implements the above-described method for acquiring large-scale MIMO-OFDM channel information based on time-frequency two-dimensional pilots.

[0239] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0240] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0241] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0242] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0243] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

Claims

1. A method for acquiring large-scale MIMO-OFDM channel information based on time-frequency two-dimensional pilots, characterized by utilizing a time-frequency two-dimensional pilot design method, The design method for time-frequency two-dimensional pilots includes the following steps: In a wireless communication system, a transmitting device sends a known two-dimensional signal to a receiving device on selected time and frequency resources; this two-dimensional signal is called a time-frequency two-dimensional pilot. The receiving device uses the received time-frequency two-dimensional pilot signal to perform channel estimation and obtain channel parameters. The time-frequency two-dimensional pilot consists of a frequency domain pilot sequence and a time domain pilot sequence. A time-frequency phase-shifted pilot is used as the time-frequency two-dimensional pilot, that is, both the frequency domain pilot sequence and the time domain pilot sequence are obtained by modulating the Zadoff-Chu sequence with a phase-shifted sequence. When the time-frequency two-dimensional pilot is a time-frequency phase-shifted pilot, different transmitting devices send time-frequency phase-shifted pilots with different or the same phase shifts to the receiving device. The method for acquiring large-scale MIMO-OFDM channel information based on time-frequency two-dimensional pilots includes the following steps: In large-scale MIMO-OFDM systems, the spatial-frequency-time domain channel is represented by a tensor, and the angle-delay-Doppler domain is sampled to establish a triple beam basis tensor model of the spatial-frequency-time domain channel tensor. The spatial-frequency-time domain channel tensor is characterized as the modulus product of the spatial-time-frequency beam matrix and the triple beam domain channel tensor. Using the statistical channel information of each user terminal in the triple beam domain, the base station schedules the corresponding time-frequency two-dimensional pilot signal for each user terminal. All user terminals send scheduled time-frequency two-dimensional pilot signals to the base station on the selected time and frequency resources. The base station estimates the triple beam domain channel tensor based on the received signals. The estimated triple beam domain channel tensor is mapped to the space-frequency-time domain to estimate and predict the space-frequency-time domain channel tensor.

2. The method according to claim 1, characterized in that, In the sampling of the angle-delay-Doppler domain, the angle domain sampling is performed in the direction cosine domain, which has a range of [-0.5, 0.5). The ranges of the delay domain and the Doppler domain are both related to the symbol time length. The direction cosine domain range, the delay domain range, and the Doppler domain range are uniformly segmented for sampling, and the corresponding number of uniform sampling points is greater than or equal to the number of antenna elements on the base station side, the number of effective subcarriers, and the number of time slots in one frame, respectively. The quantized angle-delay-Doppler domain is called the triple beam domain, and the quantized angle domain, time delay domain, and Doppler domain are called the spatial beam domain, frequency beam domain, and time beam domain, respectively.

3. The method according to claim 1, characterized in that, The space-frequency-time domain channel is a large-scale MIMO-OFDM transmission channel on the current frame, which is composed of the current time slot and multiple previous time slots. The current frame is composed of multiple time slots, each containing multiple OFDM symbols. One or more OFDM symbols are used to transmit pilot signals, and the remaining symbols are used for data transmission. Within the current frame, the space-frequency-time domain channel between each user terminal and the base station is represented as a three-dimensional tensor. The size of the three dimensions corresponds to the number of antenna elements on the base station side, the number of effective subcarriers, and the number of OFDM symbols in the frame, respectively.

4. The method according to claim 1, characterized in that, The triple beam domain channel tensor is a three-dimensional channel tensor. The three dimensions correspond to the number of spatial beams, the number of frequency beams, and the number of time beams, respectively. Each element represents the channel gain on the corresponding triple beam. The channel gains between different beams are independent of each other. The triple beamforming basis tensor model of the space-frequency-time domain channel tensor is a type of model representation of the space-frequency-time domain channel tensor. It represents the space-frequency-time domain channel tensor as the modulus product of the triple beamforming domain channel tensor and the space-time-frequency beamforming matrix. The space-time-frequency beamforming matrix consists of the space beamforming matrix, the frequency beamforming matrix, and the time beamforming matrix, respectively. These are composed of sampled space domain rudder vectors, sampled frequency domain rudder vectors, and sampled time domain rudder vectors, respectively. The rudder vectors correspond to the space beam, the frequency beam, and the time beam, pointing to the sampling angle, time delay, and Doppler frequency shift, respectively. The matrices formed by these vectors are the space beamforming matrix, the frequency beamforming matrix, and the time beamforming matrix, respectively.

5. The method according to claim 1, characterized in that, The statistical channel information in the triple beam domain is the power distribution tensor of the triple beam domain channel, represented as a three-dimensional tensor with the same dimension as the triple beam domain channel tensor and exhibiting sparsity. The statistical channel information in the triple beam domain is obtained through a channel map, which is a location-indexed channel knowledge database, or obtained using a probe signal through an online statistical channel information acquisition method.

6. The method according to claim 1, characterized in that, When the time-frequency two-dimensional pilot is a time-frequency phase-shift pilot, the steps for the base station to schedule the corresponding time-frequency two-dimensional pilot signal for each user terminal are as follows: All user terminals in a large-scale MIMO-OFDM system are grouped according to the triple beam domain statistical channel information of different user terminals. User terminals whose triple beam domain power distribution tensors do not overlap or whose overlap is less than a set threshold are grouped together. User terminals in the same group can reuse the same time-frequency two-dimensional pilot. After the user terminals are grouped, pilot signals are assigned to the user terminals in different groups so that the triple beam domain equivalent power distribution tensors between different user terminal groups do not overlap or the overlap is less than a set threshold. The triple beam domain equivalent power distribution tensor is the channel tensor obtained by shifting all elements of the triple beam domain channel power distribution tensor simultaneously along the frequency beam domain and the time beam domain. Its shift length and shift direction are determined by the phase shift factor of the time-frequency phase shift pilot. Overlap refers to the linear correlation between two power distribution tensors in the triple beam domain. When the positions of the non-zero elements in the two power distribution tensors are completely offset, the overlap is 0.

7. The method according to claim 1, characterized in that, Triple beam domain channel tensor estimation employs a tensor-based information geometry algorithm. By constructing a target manifold and an auxiliary manifold, the posterior probability density of the triple beam domain channel tensor is projected onto the target manifold, thereby obtaining the target probability density distribution. The multiplication operations involved in the spatial time-frequency beam matrix in the information geometry algorithm are quickly implemented using FFT. The expected value of the target probability density distribution is used as the estimate of the triple beam domain channel tensor.

8. The method according to claim 1, characterized in that, The estimation and prediction of the space-frequency-time domain channel tensor utilizes the mapping relationship between the space-frequency-time domain channel tensor and the triple beam domain channel tensor. The estimated value of the triple beam domain channel tensor is multiplied by the space-time-frequency beam matrix to obtain the channel information of the pilot band and data segment in the current frame, which are used as the estimation and prediction results of the space-frequency-time domain channel tensor, respectively.

9. A system for acquiring large-scale MIMO-OFDM channel information based on time-frequency two-dimensional pilots, utilizing the method for acquiring large-scale MIMO-OFDM channel information based on time-frequency two-dimensional pilots as described in any one of claims 1-8, characterized in that, The system includes a base station and multiple user terminals. The base station is used to establish a triple beam basis tensor model of the space-frequency-time domain channel tensor, and to characterize the space-frequency-time domain channel tensor as the modulus product of the space-time-frequency beam matrix and the triple beam domain channel tensor. The system uses the statistical channel information of each user terminal in the triple beam domain to schedule the corresponding time-frequency two-dimensional pilot signal for each user terminal. In the uplink, the triple beam domain channel tensor is estimated based on the received signal, and the estimated triple beam domain channel tensor is mapped to the space-frequency-time domain to complete the estimation and prediction of the space-frequency-time domain channel tensor. The user terminal is used to send known time-frequency two-dimensional pilot signals to the base station on selected time and frequency resources in the uplink.

Images