Near-field broadband non-stationary channel estimation method for super-large scale MIMO

Abstract

The invention relates to the technical field of wireless communication, in particular to a near-field broadband non-stationary channel estimation method for super-large-scale MIMO (Multiple Input Multiple Output), which comprises the following steps of: based on a near-field broadband non-stationary channel model, analyzing a linear relationship between position parameters and frequency by utilizing matching of a received signal and a polar domain sample, and acquiring a position parameter most relevant to each physical path; according to the linear relation between the position parameter and the frequency, the deviation relation of the frequency-related visible area is analyzed; based on the deviation relation of the visible area, rough estimation and accurate estimation are respectively designed by using antenna receiving signal power, the frequency-related visible area range is analyzed, and the accurate estimation value of the channel gain is confirmed; and accumulating the channel reconstruction result according to the position parameter, the visual area and the channel gain in each physical path under the same frequency so as to recover the whole broadband channel, and optimizing the channel estimation process by using the offset relation of the visual area to realize efficient reconstruction.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless communications, and in particular to a near-field broadband non-stationary channel estimation method for ultra-large-scale MIMO. Background Art

[0002] Ultra-large-scale MIMO is one of the most promising research directions in future 6G wireless communication systems. The antenna array scale of hundreds or thousands of antennas provides a significant enhancement in the system in terms of multiplexing gain, spectrum efficiency and reliability. The XL-MIMO communication system for high-bandwidth such as millimeter waves and terahertz has also attracted extensive research. This is due to the fact that millimeter waves and terahertz provide rich spectrum resources for XL-MIMO, and XL-MIMO can provide high beamforming gain to compensate for the high path loss in high-bandwidth communications.

[0003] However, as the size of antenna arrays and spectrum resources increase, more users may be in broadband near-field scenarios. The use of large aperture arrays causes different channel conditions for the transmitted signals and causes significant near-field propagation problems. At the same time, when the antenna array reaches a larger size, different array parts will see different user terminals. This is because the signal scattering cannot cover the entire antenna array, and the energy of each terminal can only be concentrated on a part of the array, called the visible area. At this time, the channel exhibits beam splitting effects and spatial non-stationarity. The angles and distances at different subcarrier frequencies will deviate, and the visible area will also change, making channel estimation more complicated. In order to give full play to the performance advantages of XL-MIMO in complex scenarios, accurate channel state information needs to be obtained.

[0004] At present, relevant work has explored the problem of wideband channel and spatial non-stationary channel estimation in XL-MIMO system in near-field environment. In the existing literature on wideband channel estimation, in order to reduce pilot overhead, channel sparsity is usually used to achieve channel estimation using compressed sensing (CS) algorithms. They are often based on a common grid assumption and establish angle and distance sampling points to achieve this. Among them, the more common method is the orthogonal matching pursuit (OMP) algorithm. According to literature search, Cui et al. proved the sparsity of near-field channels in the polarization domain, designed a polar domain transformation matrix to represent the sampling of spatial angles and distances, and proposed a grid-based polar domain synchronous orthogonal matching pursuit algorithm (P-SOMP) to achieve channel estimation. However, the proposed algorithm is based on the fact that the channel is spatially non-stationary, that is, the signal components of the same scatterer and user in the entire broadband are visible to all BS antennas. Using this algorithm to estimate spatial non-stationary channels will have a significant impact on the estimation accuracy. The main reason is that different areas of the array will receive different levels of power of user signals. Therefore, when considering non-stationary channels, the most important thing is to be able to accurately estimate the VR of the channel. The search found that Han et al. estimated VR from the perspectives of sub-arrays and scatterers based on the mapping relationship between the two, and used the power relationship and OMP idea to achieve accurate positioning of the physical position of the path and channel estimation. However, when the array is divided, although the entire channel is spatially non-stationary, when considering a single sub-array, its corresponding channel is considered to be spatially stationary. When there is a broadband beam splitting effect, the scattering of the signal becomes complicated, and the method used cannot accurately describe this influence, making it difficult to achieve accurate channel estimation. Summary of the invention

[0005] The object of the present invention is to provide a near-field broadband non-stationary channel estimation method based on ultra-large-scale MIMO to solve the problems raised in the above background technology.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] A near-field broadband non-stationary channel estimation method for ultra-large-scale MIMO, the method comprising:

[0008] Based on the near-field broadband non-stationary channel model, the linear relationship between the position parameter and the frequency is analyzed by matching the received signal with the polar domain samples to obtain the position parameter most relevant to each physical path.

[0009] According to the linear relationship between position parameters and frequency, the offset relationship of the frequency-dependent visual area is analyzed;

[0010] Based on the offset relationship of the visible area, the rough estimation and precise estimation are designed using the antenna received signal power, the frequency-related visible area range is analyzed, and the precise estimation value of the channel gain is confirmed;

[0011] The results of channel reconstruction based on position parameters, visible area and channel gain in each physical path at the same frequency are accumulated to restore the entire wideband channel, and the offset relationship of the visible area is used to optimize the channel estimation process to achieve efficient reconstruction.

[0012] Preferably, the construction of the near-field broadband non-stationary channel model includes:

[0013] Assume that the base station uses a uniform linear array with N antennas to provide services to multiple single-antenna users. Then the near-field broadband non-stationary channel model h of a user at the mth subcarrier is m for:

[0014]

[0015] Among them, l represents the number of paths, L represents the total number of paths, represents the gain on the mth subcarrier of the lth path, a(θ l , r l , f m ) is the scatterer at (θ l , r l ) direction corresponds to the subcarrier frequency f m The array response vector is used to describe the spherical wavefront between the last hop scatterer and the base station, Φ l，m represents the visible area corresponding to the signal at the mth subcarrier of the lth path, P(Φ l，m ) represents the antenna range of receiving signals at different subcarrier frequencies.

[0016] Preferably, the linear relationship between the position parameter and the frequency is analyzed by matching the received signal with the polar domain sample, including:

[0017] Assume that a single-antenna user sends a pilot signal of all 1s to the base station, and then use the polar domain sampling matrix to transform the channel to the polar domain to match the relevant position parameters;

[0018] Among them, the polar domain sampling matrix F c Each sample in is the carrier frequency f c Corresponding position An array of response vectors:

[0019]

[0020] in, Indicates the nth a Angle sampling, satisfying Indicates that in the nth a The nth angle d distance sampling, satisfying δ represents a predefined parameter; where na ∈[1, 2, …, N a ],n d ∈[1, 2, …, N d ];

[0021] According to the polar sampling matrix F c Get the received signal ym of the mth subcarrier:

[0022] in, P is the transmission power, w m The mean is 0 and the variance is σ 2 The additive Gaussian white noise, then the entire antenna receiving signal can be expressed as Y = [y 1 ,y 2 ,…y M ], m≤M;

[0023] Using signal power to represent the polar domain sampling matrix F c The correlation between each sample in and the physical path parameter, then the mth subcarrier frequency is f m When (θ l , r l ) direction of power

[0024]

[0025] Confirm signal power When the maximum value is taken, the linear relationship between the position parameter and the frequency in the broadband non-stationary channel is obtained:

[0026] in

[0027] Preferably, the location parameters most relevant to each physical path are obtained, including:

[0028] Get the frequency f m The polar domain sampling matrix at :

[0029]

[0030]

[0031] in

[0032] For a detected path According to the polar sampling matrix F c The correlation between each sample in and the physical path parameters is used to confirm the maximum coherence at each frequency to obtain the best angle and distance parameters;

[0033] According to the linear relationship between position parameters and frequency in broadband non-stationary channels, the corresponding detected path is confirmed Position parameters at each frequency.

[0034] Preferably, analyzing the offset relationship of the frequency-dependent visual area according to the linear relationship between the position parameter and the frequency includes:

[0035] Get the subcarrier frequency f m The position parameter (θ l，m， r l,m );

[0036] According to the antenna spatial geometry, confirm that at the subcarrier frequency f m When (θ l，m , r l,m The range of the antenna reached by the scattered component in the direction of ) is similar to that of the carrier frequency f c The relationship between the corresponding visible area is:

[0037]

[0038] Where Δk represents the frequency-dependent visible area offset, represents rounding up, and d represents the antenna spacing;

[0039] Indicates that at the carrier frequency f c The position of the antenna that should have received the signal is affected by the beam splitting effect. The components of the signal scattering at different frequencies reach different antenna ranges, and the frequency-dependent visible area is estimated.

[0040] A preferred, rough estimate includes:

[0041] Obtain a uniform linear array with N antennas, and then divide the entire antenna array into multiple sub-intervals according to the power of the received signal;

[0042] The corresponding path to the lth path is (θ l，m , r l,m The actual power scattered on a certain antenna in the direction of l，n (θ l，m , r l,m ), then the power S of each antenna is obtained l =[S l，0 , S l,1 , …, S l,N-1 ];

[0043] Set the number of sub-intervals to be divided to Z, then each sub-interval contains antennas, z = [0, 1, ..., Z] represents the interval number, according to the power S of each antenna l The antenna power in the zth subinterval is obtained as S l，z(θ l，m， r l,m );

[0044] Assuming that the sub-interval outside the visible area only receives Gaussian white noise, the average noise power in the sub-interval is σ 2 ;

[0045] pass With the received signal y m Perform LS calculation to obtain the global channel gain in represents the array response vector, and at this time It is assumed that the scattering is to the entire antenna array, so the roughly estimated visible area decision satisfies:

[0046] Where 0<ε<1;

[0047] When any number of antennas in a subinterval are judged to be in the visible area, the subinterval is regarded as a subset of the visible area. By this method, continuous u subintervals are obtained as a rough estimate of the visible area, which is expressed as At this time, the rough value of the channel gain is obtained through the LS algorithm

[0048] Preferably, an accurate estimate, including:

[0049] Get a rough estimate of the visible area The window detection method is used to accurately estimate the visible area:

[0050] Set the exact estimate result to The window size is [k start , k end ], For the signal received by the antenna, the antenna power is obtained based on the roughly estimated channel gain: Then the average power of the antenna in the visible area is set to S in , the average power of the non-visible area is set to S out , where S in >S out ;

[0051] For continuous visible areas, reduce the window detection range to The scaling factor γ represents the window size and Φ l，m，z The ratio of

[0052] In order to obtain accurate estimation results, Estimate S by sorting the values ​​in the nth subinterval in ascending order out and S inThe value of satisfies:

[0053]

[0054] where Φ l，m，z (a) <N β <Φ l，m，z (b)

[0055] Then the power decision function is obtained: And by adjusting γ, different windows are obtained, corresponding to the frequency visible area Satisfies the following objective function:

[0056]

[0057] At this time, according to the array response vector and the estimated and visible areas Using LS algorithm to obtain accurate path gain estimation

[0058] Preferably, the results of reconstructing the channel according to the position parameter, the visible area and the channel gain in each physical path at the same frequency are accumulated to restore the entire broadband channel, and the channel estimation process is optimized by using the offset relationship of the visible area to achieve efficient reconstruction, including:

[0059] For each subcarrier frequency f of a certain path m The channel is estimated to obtain the reconstructed polar domain channel Then superimpose the channels corresponding to each path at this frequency to obtain After updating the residual, the channel is restored using the polar domain sampling matrix, i.e.

[0060] After obtaining the angle and distance of the lth path, the parameters of each subcarrier frequency are no longer estimated separately, but the m=M / 2th subcarrier, i.e., the center subcarrier frequency f is taken. M / 2 Corresponding to The estimated As a reference value, according to the relationship between the visible area and the frequency offset shown by Δk, the visible area at other subcarrier frequencies is obtained. Further reduce the computational complexity and achieve efficient channel recovery.

[0061] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0062] The present invention provides a near-field broadband non-stationary channel estimation method for ultra-large-scale XL-MIMO, which fully considers the characteristics of broadband spatial non-stationary channels and restores channels by utilizing the relationship between position parameters, visible area and frequency; at the same time, when estimating the range of the visible area, the estimated object is no longer a subarray, but is accurate to a single antenna, accurately describing the influence of the beam splitting effect on the visible area, pointing out the phenomenon of frequency-related visible area offset, and then using the offset relationship for estimation to effectively improve the channel reconstruction efficiency; in addition, the proposed method can be applied to XL-MIMO systems of different bandwidths, and provide better estimation performance in future high-frequency broadband wireless communication scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0064] Figure 1 It is a flow chart of a near-field broadband non-stationary channel estimation method for ultra-large-scale MIMO of the present invention;

[0065] Figure 2 is an estimated performance diagram under different partitioning schemes in an embodiment of the present invention.

[0066] Figure 3 It is a comparison chart of the estimation performance of various methods under different signal-to-noise ratios in the embodiments of the present invention.

[0067] Figure 4 It is an estimated performance diagram when different subcarrier frequencies are selected as references in an embodiment of the present invention.

[0068] Figure 5 It is a comparison chart of the estimation performance of various methods under different bandwidths in the embodiments of the present invention. DETAILED DESCRIPTION

[0069] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0070] See also Figure 1-Figure 5 , the present invention provides a technical solution:

[0071] (1) Based on the near-field broadband non-stationary channel model, the linear relationship between the position parameter and the frequency is analyzed by matching the received signal with the polar domain samples to obtain the position parameter most relevant to each physical path. The specific contents are as follows:

[0072] The base station uses a uniform linear array with N antennas to serve multiple single-antenna users. The user signal is reflected along the line-of-sight path to the base station antenna. For the line-of-sight path, the user antenna is regarded as an equivalent scatterer. The near-field channel is modeled in this way, in which a spherical wavefront is used to simulate the near-field channel, and a selection vector is used to describe the correspondence between the scatterer and the antenna to describe the non-stationarity of the channel. Then the channel of a user at the mth subcarrier is:

[0073]

[0074] Where ⊙ represents the Hadamard inner product, L represents the number of paths, and Represents the gain on the mth subcarrier of the lth path. is the mth subcarrier frequency, B and f c are bandwidth and carrier frequency respectively, f m The wavelength at the point where the light is emitted is c, which is the speed of light. l represents the arrival angle of the lth path, r l is the distance from the last hop scatterer to the antenna origin. a(θ l , r l , f m ) is the scatterer at (θ l , r 1 ) direction corresponds to the subcarrier frequency f m The array response vector is used to describe the spherical wavefront between the last hop scatterer and the base station and is expressed as:

[0075]

[0076] in According to the Fresnel approximation, is the distance from the last hop scatterer to the nth antenna of the base station, n∈{0, 1, ..., N-1}, is the antenna spacing, is the carrier wavelength; in addition, Φ l,m represents the visible area corresponding to the signal at the mth subcarrier of the lth path, then the selection vector P(Φ l,m ) describes the antenna range of receiving signals at different subcarrier frequencies, where the kth element is:

[0077]

[0078] In this embodiment, a broadband non-stationary XL-MIMO system with continuous VR is considered, where the number of channel paths L=6, the number of BS antennas N=256, the number of subcarriers M=128, and the carrier frequency f c=100GHz, corresponding to the carrier wavelength λ c =3mm, the antenna spacing is d=1.5mm, the Rayleigh distance is 98.3 meters, the distance range from the user to the BS antenna is [5, 80], and the angle range is According to the antenna coordinates of [0,255], the range of the initial visible area VR is set to [100,200].

[0079] The user sends a pilot signal of all 1s to the base station. In the channel estimation process, the channel is transformed to the polar domain using the polar domain sampling matrix to match the relevant path parameters, where the sampling matrix F c Each sample in is the carrier frequency f c Corresponding position An array response vector, namely:

[0080]

[0081] in For nth a Angle sampling, satisfying is the ndth distance sample at this angle, satisfying δ is a predefined parameter, n a ∈[1,2,…,N a ],n d ∈[1,2,…,N d ], then the received signal y of the mth subcarrier can be obtained m It is expressed as:

[0082]

[0083] in, P is the transmission power, w m The mean is 0 and the variance is σ 2 The additive Gaussian white noise. Therefore, the entire antenna receiving signal can be expressed as Y = [y 1 ,y 2 ,…yM];

[0084] In this embodiment, the number of angle sampling N a =512, the number of distance samples N d = 14. Transmit power P and variance σ 2 When it is 1, it corresponds to the received signal y at the m=1 subcarrier frequency 1 =[-0.3650+0.8603i, 1.0991+0.0863i,…,-1.1557+1.3548i] T , T represents transpose. Let δ = 0.8, then the polar samples obtained by sampling all antennas at the first angle and distance are Each element corresponds to

[0085] The signal power is used to represent the correlation between the polar domain samples and the physical path parameters, and the subcarrier frequency corresponding to the signal is f m When (θ l , r l ) direction is expressed as:

[0086]

[0087] Where H represents the conjugate transpose. When the maximum value is taken, the following relationship between the position parameter and the frequency in the broadband non-stationary channel under the beam splitting effect is obtained:

[0088]

[0089] in

[0090] The above relationship is used to further design the polar domain sampling matrix at other subcarrier frequencies so that the coherence between the sample and the physical path position in the entire broadband is as large as possible. Then the frequency is f m When , the corresponding sampling matrix, that is, the dictionary is:

[0091]

[0092] in So for each detected path The maximum coherence under all frequencies is accumulated to obtain the optimal angle and distance parameters, which are specifically expressed as:

[0093]

[0094] Where R = Y is the residual matrix, which is obtained by subtracting the estimated The result of the first path updates the residual matrix, and then The position parameters of the paths are matched and estimated; then the linear relationship between the position parameters and the frequency in the broadband non-stationary channel is used to obtain the position parameters of the corresponding path at other frequencies.

[0095] (2) According to the linear relationship between position parameters and frequency, the offset relationship of the frequency-related visible area is analyzed. The specific content is as follows:

[0096] Under the influence of beam splitting effect, the user signal is at the subcarrier frequency f m When (θ l，m , r l,m The range of the antenna reached by the scattered component in the direction of ) is similar to that of the carrier frequency f cThe relationship between the corresponding visible area is:

[0097]

[0098] Where Δk is the frequency-dependent visible area offset, means rounding up, and d is the antenna spacing. This shows that at the carrier frequency f c The degree of deviation of the position of the antenna that should have received the signal due to the beam splitting effect. For a certain path, when the number of subcarriers is constant, the size of the offset Δk depends on f m , and f m The upper and lower limits depend on the bandwidth B. As the bandwidth increases, the f at the boundary m With f c As the ratio of increases, the offset increases, that is, the signal scattered components at different frequencies reach different antenna ranges. Considering only the overall VR will lead to a large estimation error. Further, the visual area related to the estimated frequency is considered and estimated separately.

[0099] In this embodiment, when r = 15 meters, When the initial visible area is [100,200], the corresponding m 1 = 50, the visible area after the subcarrier frequency shift is approximately [116,216], corresponding to m 2 = 100, the visible area after the shift is approximately [75, 175], and the center subcarrier frequency corresponds to m = 64.

[0100] (3) Based on the offset relationship of the visible area, the antenna received signal power is used to design a rough estimate and an accurate estimate, analyze the frequency-related visible area range, and confirm the accurate estimate of the channel gain. The specific contents are as follows:

[0101] S1. Rough estimate:

[0102] The entire antenna array is divided into multiple sub-intervals, and the decision is made based on the received signal power, where the influence of noise power is mainly considered, and the interval where the entire visible area is included as much as possible while removing the non-visible area. The corresponding path for the lth path is (θ l，m , r l，m ) direction and the actual power scattered on the nth antenna is S l，n (θ l，m , r l，m ), expressed as:

[0103]

[0104] in is the estimated location parameter, ||ω m，n ||2 is the noise power, a(θ l，m , r l，m )=a(θ l , r l , f m )⊙P(Φ l，m ); the power of each antenna can be expressed as S l =[S l，0 , S l，1 , …, S l，N-1 ];

[0105] Set the number of sub-intervals to Z, then each sub-interval contains antenna, z = [0, 1, ..., Z] is the interval number. According to the defined power, the power in the zth subinterval is S l，z (θ l，m , r l，m );

[0106] Assume that the subinterval outside the visible area only receives Gaussian white noise, and the average noise power in the subinterval is σ 2 ,pass Perform LS calculation with the received signal ym to obtain the global channel gain And at this time It is assumed that the scattering is to the entire antenna array, so the roughly estimated visible area decision satisfies:

[0107]

[0108] Where 0<ε<1; when any number of antennas in a subinterval are judged to be in the visible area, the subinterval is regarded as a subset of the visible area. By this method, continuous u subintervals are obtained as a rough estimate of the visible area, which is expressed as At this time, the rough value of the channel gain is obtained through the LS algorithm

[0109] S2. Accurate estimation:

[0110] Based on the rough estimation of the approximate range of the corresponding frequency visible area, the window detection method is further used to accurately process it. At this time, only the boundary of the continuous visible area needs to be estimated;

[0111] Assume that the exact estimate is The window size is [k start , k end ], For the signal received by the antenna, we can get for:

[0112]

[0113] The average power in the actual visible area is greater than the average power in the non-visible area, which are respectively set as S in and S out ;

[0114] The precise estimation process is to find a value that is as small as possible but covers The area of ​​main power is used as an estimate of the visible area range, where is the received power obtained based on the rough estimation result;

[0115] Therefore, the scaling factor γ is introduced to represent the window size and Φ l，m，z For a continuous visible area, Φ l,m,z The first and last subintervals may contain antennas that are not in the VR, and the antennas in the middle subintervals are considered to be in the VR, thereby reducing the window detection range to Then, in order to obtain the ideal window, Estimate S by sorting the values ​​at different n in ascending order out and S in The value of satisfies:

[0116]

[0117]

[0118] where Φ l，m，z (a) <N β <Φ l，m，z (b), and let β be This results in the power decision function:

[0119]

[0120] By adjusting γ, different windows are obtained, corresponding to the frequency visible area Satisfies the following objective function:

[0121]

[0122]

[0123] k end =k start -1+γ(Φ l,m,z (b)-Φ l,m,z (a)+1),…,Φ l，m，z (b)

[0124] Minimizing the power difference ensures that the window corresponding range is highly similar to the frequency-dependent visible area; according to the estimated array response vector And the visible area Using LS algorithm to obtain accurate path gain estimation It is expressed as:

[0125]

[0126] in

[0127] (4) Accumulate the results of channel reconstruction based on position parameters, visible area and channel gain in each physical path at the same frequency, so as to restore the entire broadband channel, and optimize the channel estimation process by using the offset relationship of the visible area to achieve efficient reconstruction. The specific contents are as follows:

[0128] For each subcarrier frequency f of a certain path m The channel is estimated to obtain the reconstructed polar domain channel Then superimpose the channels corresponding to each path at this frequency to obtain After updating the residual, the channel is restored using the polar domain sampling matrix, i.e. Further, this process is improved. After obtaining the angle and distance of the lth path, the parameters of each subcarrier frequency are no longer estimated separately. Instead, the corresponding parameters of the m=M / 2th subcarrier, i.e., the center subcarrier frequency fM / 2, are taken. The estimated As a reference value, according to the relationship between the visible area and the frequency offset shown by Δk, the visible area at other subcarrier frequencies is obtained by compensating the offset Further reduce the computational complexity and achieve efficient channel recovery.

[0129] The estimated performance of the method proposed in the present invention in the actual system is as follows Figure 2 to Figure 5 As shown, the method of estimating the channel corresponding to each subcarrier frequency is recorded as Algorithm 1, and the method of taking the central subcarrier frequency as a reference is recorded as Algorithm 2.

[0130] like Figure 2 As shown in Figure 1, the performance of Algorithm 1 improves with the increase of SNR under different sub-interval partitioning schemes, where Z0=8, Z1=16, and Z2=32, and better estimation accuracy is achieved under the Z2 scheme.

[0131] The comparison of the estimation performance of the present invention and various methods under different signal-to-noise ratios is shown in Figure 2. Figure 3As shown, the least squares (LS) algorithm, the minimum mean square error (NMSE) algorithm, the near-field orthogonal matching pursuit (OMP) algorithm and the polar domain OMP (POMP) algorithm are used for channel estimation. The LS algorithm and the NMSE algorithm estimate separately according to different subcarrier frequencies without making any assumptions and processing on the channel, resulting in lower estimation accuracy. Although the near-field OMP algorithm takes into account the angle and distance factors of the near-field channel, it ignores the influence of beam splitting. The POMP algorithm accurately describes the relationship between the position parameter and the frequency, but does not involve the situation of the frequency-related visible area. The two methods in the present invention, on the basis of the above, also fully consider the relationship between the visible area and the frequency under broadband, thereby improving the channel estimation performance.

[0132] Figure 4 The estimation performance diagram of Algorithm 2 when different subcarrier frequencies are selected as reference is given. It can be seen that when the center subcarrier frequency is selected as the reference, better estimation accuracy can be achieved. This is because when the angle and distance of the path are constant, the farther the subcarrier frequency is from the center carrier frequency, the greater the corresponding position parameters and the degree of visible area offset will be. In addition, if only local scattering is performed on the antenna array or no scattering is performed on the antenna array, the accuracy of estimating other components using the reference offset will be affected.

[0133] Figure 5 This is a comparison chart of the estimation performance of the present invention and various methods under different bandwidths. As the bandwidth increases, the proposed Algorithms 1 and 2 always achieve better estimation accuracy than other algorithms, and their performances are similar. The computational complexity of Algorithm 2 is lower than that of Algorithm 1. This is due to the fact that Algorithm 2 uses the offset relationship of the frequency-related visible area to achieve efficient estimation.

[0134] In summary, the present invention studies the problem of near-field broadband non-stationary channel estimation in XL-MIMO systems, and proposes an accurate and efficient channel estimation method to solve the problem of difficulty in obtaining accurate channel state information in complex communication scenarios, which can achieve good estimation performance under low signal-to-noise ratio and high bandwidth conditions.

[0135] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A near-field broadband non-stationary channel estimation method for ultra-large-scale MIMO, characterized by: The method comprises: Based on the near-field broadband non-stationary channel model, the linear relationship between the position parameter and the frequency is analyzed by matching the received signal with the polar domain samples to obtain the position parameter most relevant to each physical path. According to the linear relationship between position parameters and frequency, the offset relationship of the frequency-dependent visual area is analyzed; Based on the offset relationship of the visible area, the rough estimation and precise estimation are designed using the antenna received signal power, the frequency-related visible area range is analyzed, and the precise estimation value of the channel gain is confirmed; The results of channel reconstruction based on position parameters, visible area and channel gain in each physical path at the same frequency are accumulated to restore the entire wideband channel, and the offset relationship of the visible area is used to optimize the channel estimation process to achieve efficient reconstruction.

2. The near-field broadband non-stationary channel estimation method according to claim 1, characterized in that: The construction of the near-field broadband non-stationary channel model includes: Assume that the base station uses a uniform linear array with N antennas to provide services to multiple single-antenna users. Then the near-field broadband non-stationary channel model h of a user at the mth subcarrier is m for: Among them, l represents the number of paths, L represents the total number of paths, represents the gain on the mth subcarrier of the lth path, a(θ l ,r l ,f m ) is the scatterer at (θ l ,r l ) direction corresponds to the subcarrier frequency f m The array response vector is used to describe the spherical wavefront between the last hop scatterer and the base station, Φ l,m represents the visible area corresponding to the signal at the mth subcarrier of the lth path, P(Φ l,m ) represents the antenna range of receiving signals at different subcarrier frequencies.

3. The near-field broadband non-stationary channel estimation method according to claim 1, characterized in that: The linear relationship between the position parameter and the frequency is analyzed by matching the received signal with the polar domain sample, including: Assume that a single-antenna user sends a pilot signal of all 1s to the base station, and then use the polar domain sampling matrix to transform the channel to the polar domain to match the relevant position parameters; Among them, the polar domain sampling matrix F c Each sample in is the carrier frequency f c Corresponding position An array of response vectors: in, Indicates the nth a Angle sampling, satisfying Indicates that in the nth a The nth angle d distance sampling, satisfying δ represents a predefined parameter; where n a ∈[1,2,…,N a ],n d ∈[1,2,…,N d ]; According to the polar sampling matrix F c Get the received signal y of the mth subcarrier m : in, P is the transmission power, w m The mean is 0 and the variance is σ 2 The additive Gaussian white noise of the antenna can be expressed as Y = [y1, y2, ... y M ], m≤M; Using signal power to represent the polar domain sampling matrix F c The correlation between each sample in and the physical path parameter, then the mth subcarrier frequency is f m When (θ l ,r l ) direction of power Confirm signal power When the maximum value is taken, the linear relationship between the position parameter and the frequency in the broadband non-stationary channel is obtained: in 4. The near-field broadband non-stationary channel estimation method according to claim 1, characterized in that: The obtaining of the location parameters most relevant to each physical path includes: Get the frequency f m The polar domain sampling matrix at : in For a detected path According to the polar sampling matrix F c The correlation between each sample in and the physical path parameters is used to confirm the maximum coherence at each frequency to obtain the best angle and distance parameters; According to the linear relationship between position parameters and frequency in broadband non-stationary channels, the corresponding detected path is confirmed Position parameters at each frequency.

5. The near-field broadband non-stationary channel estimation method according to claim 1, characterized in that: The analyzing the offset relationship of the frequency-dependent visual area according to the linear relationship between the position parameter and the frequency includes: Get the subcarrier frequency f m The position parameter (θ l,m ,r l,m ); According to the antenna spatial geometry, confirm that at the subcarrier frequency f m When (θ l,m ,r l,m The range of the antenna reached by the scattered component in the direction of ) is similar to that of the carrier frequency f c The relationship between the corresponding visible area is: Where Δk represents the frequency-dependent visible area offset, represents rounding up, and d represents the antenna spacing; Indicates that at the carrier frequency f c The position of the antenna that should have received the signal is affected by the beam splitting effect. The components of the signal scattering at different frequencies reach different antenna ranges, and the frequency-dependent visible area is estimated.

6. The near-field broadband non-stationary channel estimation method according to claim 1, characterized in that: The rough estimate includes: Obtain a uniform linear array with N antennas, and then divide the entire antenna array into multiple sub-intervals according to the power of the received signal; The corresponding path to the lth path is (θ l,m ,r l,m The actual power scattered on a certain antenna in the direction of l,n (θ l,m ,r l,m ), then the power S of each antenna is obtained l =[S l,0 ,S l,1 ,…,S l,N-1 ]; Set the number of sub-intervals to be divided to Z, then each sub-interval contains antennas, z = [0,1,…,Z] represents the interval number, according to the power S of each antenna l The antenna power in the zth subinterval is obtained as S l,z (θ l,m ,r l,m ); Assuming that the sub-interval outside the visible area only receives Gaussian white noise, the average noise power in the sub-interval is σ 2 ; pass With the received signal y m Perform LS calculation to obtain the global channel gain in represents the array response vector, and at this time It is assumed that the scattering is to the entire antenna array, so the roughly estimated visible area decision satisfies: Where 0<ε<1; When any number of antennas in a subinterval are judged to be in the visible area, the subinterval is regarded as a subset of the visible area. By this method, continuous u subintervals are obtained as a rough estimate of the visible area, which is expressed as At this time, the rough value of the channel gain is obtained through the LS algorithm 7. The near-field broadband non-stationary channel estimation method according to claim 1, characterized in that: The precise estimation includes: Get a rough estimate of the visible area The window detection method is used to accurately estimate the visible area: Set the exact estimate result to The window size is [k start ,k end ], For the signal received by the antenna, the antenna power is obtained based on the roughly estimated channel gain: Then the average power of the antenna in the visible area is set to S in , the average power of the non-visible area is set to S out , where S in >S out ; For continuous visible areas, reduce the window detection range to The scaling factor γ represents the window size and Φ l,m,z The ratio of In order to obtain accurate estimation results, Estimate S by sorting the values ​​in the nth subinterval in ascending order out and S in The value of satisfies: among themF l,m,z (a) <N β <F l,m,z (b), Then the power decision function is obtained: And by adjusting γ, different windows are obtained, corresponding to the frequency visible area Satisfies the following objective function: st k start =Φ l，m，z (a),…,Φ l，m，z (a)+(1-c)(Φ l，m，z (b)-F l，m，z (a)+1) k end =k start -1+γ(Φ l，m，z (b)-F l，m，z (a)+1),…,Φ l，m,z (b) At this time, according to the array response vector and the estimated and visible areas Using LS algorithm to obtain accurate path gain estimation 8. The near-field broadband non-stationary channel estimation method according to claim 1, characterized in that: The method accumulates the results of reconstructing the channel according to the position parameters, the visible area and the channel gain in each physical path at the same frequency, thereby restoring the entire broadband channel, and optimizing the channel estimation process by using the offset relationship of the visible area to achieve efficient reconstruction, including: For each subcarrier frequency f of a certain path m The channel is estimated to obtain the reconstructed polar domain channel Then superimpose the channels corresponding to each path at this frequency to obtain After updating the residual, the channel is restored using the polar domain sampling matrix, i.e. After obtaining the angle and distance of the lth path, the parameters of each subcarrier frequency are no longer estimated separately, but the m=M / 2th subcarrier, i.e., the center subcarrier frequency f is taken. M / 2 Corresponding to The estimated As a reference value, according to the relationship between the visible area and the frequency offset shown by Δk, the visible area at other subcarrier frequencies is obtained. Further reduce the computational complexity and achieve efficient channel recovery.

Images