CN113141202B - MIMO spatial non-stationary channel estimation method based on image contour extraction - Google Patents

Abstract

A method for spatially non-stationary channel estimation based on image contour extraction, which is characterized in that, in a massive MIMO spatially non-stationary channel moving scenario, the received signal with sparse angle delay domain is represented in the form of an image, and then the image contour is used. The extracted estimation algorithm obtains the estimated number of paths, the angle of each path, and the time delay. After the received signal with sparse spatial delay domain is represented in the form of an image, the estimation algorithm of image contour extraction is used to obtain the corresponding effective visual effect of each channel path. area estimation, so as to realize path gain and channel reconstruction; the present invention utilizes image contour extraction technology and the sparseness of the channel in the angular delay domain and the spatial delay domain to estimate the angle and delay of each path of the channel and do not depend on the sub-array. The divided visible area replaces the traditional iterative optimization method to solve the problem of spatially non-stationary channel estimation.

Description

MIMO spatial non-stationary channel estimation method based on image contour extraction

technical field

The present invention relates to a technology in the field of wireless communication, in particular to a massive MIMO space non-stationary channel estimation based on an image contour extraction technology.

Background technique

Most of the existing channel estimation techniques proposed for spatial non-stationary characteristics only involve the situation where both ends of the receiver and transmitter are stationary, but little attention is paid to the computational complexity of the algorithm and the scenario where the terminal is moving. A method design that achieves a good balance between computational complexity. At the same time, regarding the spatial non-stationary characteristics, that is, the consideration of the visual area in the prior art mostly depends on the sub-array division, and inappropriate division settings will inevitably lead to matching errors, thereby causing the performance loss of the reconstructed channel.

SUMMARY OF THE INVENTION

Aiming at the problems that the performance and complexity of the prior art are unbalanced, it is difficult to meet the calculation delay required by the actual mobile scene, and the spatial non-stationary characteristics and matching errors are ignored, the invention proposes a MIMO spatial non-stationary channel estimation method based on image contour extraction, In the massive MIMO OFDM system, the image contour extraction technology and the sparseness of the channel in the angular delay domain and the spatial delay domain are used to estimate the angle and delay of each path of the channel and the visible area independent of the sub-array division. It replaces the traditional iterative optimization method to solve the spatially non-stationary channel estimation problem.

The present invention is achieved through the following technical solutions:

The invention relates to a space non-stationary channel estimation method based on image contour extraction. In the massive MIMO space non-stationary channel moving scenario, the received signal with sparse angle delay domain is represented in the form of an image, and then the image contour extraction method is used to extract the received signal. The estimated number of paths, the angle of each path, and the time delay are obtained by using the estimation algorithm of . After the received signal with sparse spatial delay domain is represented in the form of an image, the estimation algorithm of image contour extraction is used to obtain the effective visible area corresponding to each channel path. estimation, so as to achieve path gain and channel reconstruction.

The described estimation algorithm for image contour extraction specifically includes:

1) When the two-dimensional coordinate (i, j) of any pixel on the image is the starting point of the outer boundary, set the marker NBD=NBD+1, record (i, j) and the updated NBD value, and set (i , j) the left point (i, j-1) is recorded as (i ₂ , j ₂ ); otherwise, skip to step 6).

并跳至步骤6)。2) Take (i ₂ , j ₂ ) as the starting point, take (i, j) as the center, and detect clockwise: when there are non-zero pixels in its upper, lower, left, and right neighborhoods, it is recorded as (i ₁ , j ₁ ), And update (i ₂ , j ₂ )=(i ₁ , j ₁ ), denote (i, j) as (i ₃ , j ₃ ); otherwise, the point binarizes the pixel value

and skip to step 6).

3) Take (i ₃ , j ₃ ) as the center of the circle, and detect the previous point counterclockwise with (i ₂ , j ₂ ) as the starting point: when there are non-zero pixels on the top, bottom, left and right of the center point (i ₃ , j ₃ ), Denoted as (i ₄ , j ₄ ).

当(i₃，j₃+1)并非步骤3)中已检测过的零像素点且满足

则

否则

的值不改变。4) When (i ₃ , j ₃ +1) is the zero pixel detected in step 3), the pixel value of this point is binarized

When (i ₃ , j ₃ +1) is not the zero pixel detected in step 3) and satisfies

but

otherwise

The value does not change.

5) When the starting point of the outer boundary of the current contour has been retrieved, that is, when (i ₄ , j ₄ )=(i, j) and (i ₃ , j ₃ )=(i ₁ , j ₁ ) are satisfied, then jump to Step 6); otherwise update (i ₂ , j ₂ )=(i ₃ , j ₃ ), (i ₃ , j ₃ )=(i ₄ , j ₄ ), and skip to step 3).

时，更新标记符

并从像素点(i，j+1)开始 继续光栅扫描检测，直至扫描至图像右下角顶点。6) When the point binarizes the pixel value

, update the marker

And continue raster scanning detection from pixel point (i, j+1) until scanning to the lower right corner vertex of the image.

technical effect

The invention as a whole solves the problem that the calculation complexity of the prior art is too high and cannot be applied to moving scenes and the estimation of the visible area has a matching error due to its dependence on sub-array division; compared with the prior art, the present invention utilizes the The contour extraction algorithm and its low-complexity pixel domain processing can achieve spatially non-stationary channel estimation and tracking in mobile scenarios with low computational complexity while ensuring the estimation performance.

Description of drawings

Figure 1 is a scene diagram of a massive MIMO spatially non-stationary channel model;

Figure 2 is a color image and a grayscale image of the received signal in the spatially non-stationary angle delay domain;

Figure 3 is a color image and a grayscale image of the received signal in the space non-stationary space delay domain;

Fig. 4 is the realization flow chart of spatial non-stationary channel estimation based on image contour extraction;

Fig. 5 is the channel estimation mean square error performance and processing delay comparison based on image contour extraction and baseline algorithm;

Figure 6 shows the comparison of channel tracking mean square error performance and processing delay based on image contour extraction and baseline algorithms.

Detailed ways

λ为发送信号波长，发送端为单天 线终端，则经过傅里叶变换后的空间频率域的接收信号Y＝H+N，其中：由N_r×N_s个元素组 成的复数矩阵H为大规模MIMO衰落相关系数，N_s为频域子载波数。As shown in Figure 4, this embodiment is based on a massive MIMO OFDM system with spatially non-stationary characteristics. When the base station receiving end is configured with N _r transmitting antennas to form a uniform linear array, the antenna spacing is

λ is the wavelength of the transmitted signal, and the transmitting end is a single-antenna terminal, then the received signal in the spatial frequency domain after Fourier transform Y=H+N, where: the complex matrix H composed of N _r ×N _s elements is large Massive MIMO fading correlation coefficient, N _s is the number of sub-carriers in the frequency domain.

其中：L为路径数，g_l为路径l的增益，

和

分别为

和μ_l＝ Δfτ_l对应的空域和频域舵矢量，θ_l和τ_l为路径l对应的信号出发角和时延，Δf为子载波间隔。Φ_l表示路径l的可视区域，

用于可视区域中的有效天线索引选择，如果 第m根天线属于集合Φ_l，则在向量p的第m个位置置1，否则置0。For ease of explanation, all pilots are transmitted in this embodiment, so the transmitted signal is omitted here, and N is additive white Gaussian noise with zero mean and unit variance, then the spatially non-stationary wireless channel model at the transmitting and receiving ends is:

Where: L is the number of paths, g _l is the gain of path l,

and

respectively

and μ _l = Δfτ _l corresponding to the airspace and frequency domain rudder vectors, θ _l and τ _l are the signal departure angle and time delay corresponding to path l, and Δf is the subcarrier spacing. Φ _l represents the visible area of path l,

Used for effective antenna index selection in the visible area, if the mth antenna belongs to the set Φ _l , it is set to 1 at the mth position of the vector p, otherwise it is set to 0.

This embodiment relates to a MIMO spatially non-stationary channel estimation method based on image contour extraction. In a massive MIMO spatially non-stationary channel moving scenario, after the received signal with sparse angle delay domain is represented in the form of an image, the image The estimation algorithm of contour extraction obtains the estimated number of paths, the signal departure angle of each path, and the propagation delay. After the received signal with sparse spatial delay domain is represented in the form of an image, the estimation algorithm of image contour extraction is used to obtain each channel path. Corresponding to the effective visible area estimation, so as to realize the path gain and channel reconstruction.

The estimated number of paths, the signal departure angle of each path, and the propagation delay are obtained in the following ways:

再在角度时延域稀疏的接收信号以图像的形式表现，即对

中每个元素

的模进一步缩 放以适应图像0-255的像素值范围：

基于缩放后的接收信号

与 MATLAB中的Image和Mat2gray函数，即可生成对应的角度时延域稀疏接收信号彩色图像和 灰度图像，如图2(a)与(b)所示，其中：D_a和D_S分别是前N_r行和前N_s行的ηN_r维与ηN_s维的离散 傅里叶变换矩阵，η为过采样率，函数max(·)用于获得输入矩阵中最大的元素模值。① In a massive MIMO channel, when the number of paths is much smaller than the number of antennas, the received signal corresponding to the channel and the all-one pilot sequence has a sparse characteristic in the angular delay domain, and the received signal is converted from the spatial frequency domain to the angular delay. area

Then the sparse received signal in the angular delay domain is represented in the form of an image, that is, the

each element in

The modulo is further scaled to fit the image's 0-255 pixel value range:

Based on the scaled received signal

With the Image and Mat2gray functions in MATLAB, the corresponding color image and grayscale image of the sparse received signal in the angle delay domain can be generated, as shown in Figure 2(a) and (b), where: D _a and D _S are respectively The discrete Fourier transform matrix of ηN _r and ηN _s dimensions of the first N _r rows and the first N _s rows, η is the oversampling rate, and the function max(·) is used to obtain the largest element modulus value in the input matrix.

再设置标记符NBD＝1与LNBD＝0，其中NBD 与LNBD分别为新边界(New BorDer)和前一个新边界(Last New BorDer)，以光栅扫描的方式遍 历图像各像素点，当扫描到新一行起始位置时，重置LNBD＝0，当且仅当检测到像素点(i，j) 是外边界开始点

且

)，并且当前LNBD≤0时，对图像进行轮廓提取。② On the basis of the grayscale image of the original received signal, the pixel value f _{i, j} of each pixel point (i, j) in the grayscale image is binarized by setting the threshold δ, namely:

Then set the markers NBD=1 and LNBD=0, where NBD and LNBD are the new boundary (New BorDer) and the previous new boundary (Last New BorDer) respectively, and traverse each pixel of the image in a raster scanning manner. When the starting position of a line, reset LNBD=0, if and only if the pixel point (i, j) is detected as the starting point of the outer boundary

and

), and when the current LNBD≤0, perform contour extraction on the image.

The described contour extraction specifically includes:

1) When (i, j) is the starting point of the outer boundary, NBD=NBD+1, record (i, j) and the updated NBD value, and record (i, j-1) at the left of (i, j) Do (i ₂ , j ₂ ); if (i, j) is not the starting point of the outer boundary, go to step 6).

并跳至步骤6)。2) Take (i ₂ , j ₂ ) as the starting point, and clockwise around (i, j) to detect whether there are non-zero pixels in the upper, lower, left and right sides, if so, record it as (i ₁ , j ₁ ), and update ( i ₂ , j ₂ )=(i ₁ , j ₁ ), denote (i, j) as (i ₃ , j ₃ ); otherwise, then

and skip to step 6).

3) Around (i ₃ , j ₃ ), take (i ₂ , j ₂ ) as the previous point of the starting point, counterclockwise to detect whether there are non-zero pixels on the top, bottom, left and right of the center point (i ₃ , j ₃ ), if so, Then denoted as (i ₄ , j ₄ ).

若不是，且满 足

则

否则，则执行下一步骤。4) Determine whether (i ₃ , j ₃ +1) is the zero pixel detected in step 3), then

If not, and satisfy

but

Otherwise, proceed to the next step.

5) Judging that when the algorithm has retrieved the starting point of the outer boundary of the current contour, that is, when (i ₄ , j ₄ )=(i, j) and (i ₃ , j ₃ )=(i ₁ , j ₁ ) are satisfied, then Go to step 6); otherwise, update (i ₂ , j ₂ )=(i ₃ , j ₃ ), (i ₃ , j ₃ )=(i ₄ , j ₄ ), and go to step 3).

则更新

并从像素点(i，j+1)开始继续光栅扫描检测，直至 扫描至图像右下角顶点。6) If

then update

And continue raster scanning detection from pixel point (i, j+1) until scanning to the lower right corner vertex of the image.

然 后以每个轮廓的外边界开始点(i，j)为初始点，计算当前第

个轮廓横向及纵向所占像素点 数，用以获得轮廓的高h_l和宽w_l，结合外边界开始点坐标即可得到当前轮廓包围的矩形的中心 点坐标

根据中心坐标获得第l个信道路径对应的信号出发角度

以及时延估计结果

③ When all contours are extracted, the pixel value of the starting point of the outer boundary of the last contour is the estimated number of paths.

Then take the starting point (i, j) of the outer boundary of each contour as the initial point, calculate the current

The number of pixels occupied by each contour horizontally and vertically is used to obtain the height h _l and width w _l of the contour, and the coordinates of the center point of the rectangle enclosed by the current contour can be obtained by combining the coordinates of the starting point of the outer boundary

Obtain the signal departure angle corresponding to the lth channel path according to the center coordinates

and delay estimation results

The effective visible area estimation corresponding to each channel path is obtained by the following methods:

并对

中每个元素

的模 进一步缩放以适应图像0-255的像素值范围：

基于缩放后的接收信号

与MATLAB中的Image和Mat2gray函数，即可生成对应的空间时延域稀疏接收信号彩色图像 和灰度图像，如图3(a)与(b)所示。i) Convert the received signal from the spatial frequency domain to the spatial delay domain:

and to

each element in

The modulo is further scaled to fit the image's 0-255 pixel value range:

Based on the scaled received signal

With the Image and Mat2gray functions in MATLAB, the corresponding color image and grayscale image of the sparse received signal in the space delay domain can be generated, as shown in Figure 3(a) and (b).

ii) Set the threshold δ to complete the binarization preprocessing on the pixel value of each pixel point of the image, and perform contour extraction on the image after satisfying the execution conditions of the contour extraction algorithm.

个轮廓横向及纵向所占像素点数，用以获得轮廓的高h_l和宽w_l；然后结合外边界开始点坐标得 到当前轮廓包围的矩形的横向中心点坐标

根据该很像中心坐标x_l与第l个外边界开 始点纵坐标j，得到第l个信道路径对应的时延信息

可视区域

iii) After extracting all the contours, take the starting point (i, j) of the outer boundary of each contour as the initial point, calculate the current No.

The number of pixels occupied by the horizontal and vertical directions of each contour is used to obtain the height h _l and width w _l of the contour; then the coordinates of the horizontal center point of the rectangle enclosed by the current contour are obtained by combining the coordinates of the starting point of the outer boundary

According to the coordinate xl of the likeness center and the ordinate j of the starting point of the lth outer boundary, the delay information corresponding to the _lth channel path is obtained

Visible area

即可将估计得到的角度与时延信息与可视区域进行匹配，获得各 信道路径对应参数信息

iv) According to the obtained delay

The estimated angle and delay information can be matched with the visible area, and the corresponding parameter information of each channel path can be obtained.

其中：

与vec(·)表示矩阵求伪逆以及将矩阵按列向量化的运算符。The path gain refers to: according to the angle, time delay of each path and the effective visible area corresponding to each channel path, the gain of each path is obtained by the least square method:

in:

And vec( ) represents an operator for pseudo-inverse and column-wise vectorization of matrices.

其中：

为估计路径数，

与

为路径l的增益、信号出发角度、可视区域与传播时延的估计值，a(·)与q(·)为对应角度 与时延的空域、频域舵矢量，p(·)用于对应可视区域中的有效天线索引选择。The channel estimation refers to: substituting the estimated total number of paths and the channel parameters of each path into the wireless channel model to reconstruct the spatially non-stationary channel:

in:

to estimate the number of paths,

and

is the estimated value of the gain of path l, the departure angle of the signal, the visible area and the propagation delay, a( ) and q( ) are the airspace and frequency domain rudder vectors corresponding to the angle and delay, p( ) is used for Corresponds to valid antenna index selections in the visible area.

The channel parameters of each path refer to: signal departure angle, signal propagation delay, effective visible area estimation and path gain.

成立时直接进行路径增益与信道估计，从而缩减信道估计所需的处理时延；否则进行完整的信 道空间参数与路径增益更新估计，从而重构时隙t的空间非平稳无线信道。Preferably, after the channel is reconstructed in the present invention, channel tracking is further implemented in the following manner: according to whether the visible area changes the channel space parameters corresponding to the current time slot t, that is, the angle, time delay and visible area, specifically: When reconstructing the channel of time slot t, the channel visible area estimation based on image contour extraction is performed, and when

When established, the path gain and channel estimation are directly performed, thereby reducing the processing delay required for channel estimation; otherwise, the complete channel space parameter and path gain update estimation is performed to reconstruct the spatially non-stationary wireless channel of time slot t.

The channel data used in verifying the algorithm performance in this embodiment is generated from the literature (Wu S, Wang C X, Aggoune E, et al. A General 3-D Non-Stationary 5G Wireless Channel Model [J]. IEEE Transactions on Communications, 2018 , 66 (7): 3065-3078.) in the mobile scene space non-stationary channel model, this embodiment will document (Han Y, Li M, Jin S, et al. Deep Learning Based FDDNon-Stationary Massive MIMO Downlink Channel Reconstruction IEEE Journal on Selected Areas in Communications, 2020.) The state-of-the-art channel estimation algorithm based on Newton's quadrature matching is used as the baseline for performance comparison. The specific parameters adopted in the present embodiment are as shown in Table 1:

Table 1 Relevant parameters used in this example parameter value parameter value The number of base station antennas N_r 64 CPU Intel i7 carrier frequency 2.6GHz Memory Capacity 16 GB Number of subcarriers N_s 56 Hard drive capacity 256GB Terminal moving speed 3m/s operating system Windows 10

This embodiment compares the spatially non-stationary channel estimation based on image contour extraction and the baseline method to show the superiority of the architecture proposed in this embodiment in terms of performance and computational complexity. As shown in Figure 5, the channel estimation algorithm based on contour extraction is significantly better than the baseline with 4 sub-arrays in terms of mean square error performance. This is mainly because the estimation of the visible area in this embodiment does not depend on sub-arrays, so It can eliminate the matching error between the sub-array division and the visible area, and has better visible area estimation and final channel estimation performance than the baseline. Compared with the baseline with the number of sub-arrays being 64, although the channel estimation method based on contour extraction still has room for improvement in terms of accuracy, in terms of computational complexity, this embodiment has a significant advantage, so that it can be applied to the processing time delay. Restricted mobile scenarios.

As shown in FIG. 6 , this embodiment extends the above experiment to a channel scenario after 2 seconds to demonstrate the performance of spatially non-stationary channel tracking in this embodiment. Due to the short processing time delay, the present embodiment can estimate the channel state after 2 seconds in real time, thus having reliable mean square error performance compared with the baseline method that cannot track the channel in real time due to high computational complexity.

Compared with the prior art, the method utilizes the sparse characteristics of the channel in the angular delay domain and the spatial delay domain and the image contour extraction algorithm on the problem of spatially non-stationary channel estimation, and replaces the pixel processing with low computational complexity with high complexity. The traditional iterative optimization method is adopted, so as to avoid the performance and computational complexity imbalance in the traditional method, and it is difficult to be used for the difficulty and defect of channel tracking in mobile scenarios.

Secondly, in an actual wireless system, the estimation of the visible area based on the division of the antenna sub-array usually has a matching error compared with the actual situation. Therefore, in order to avoid this error, the present invention uses each antenna as a unit for each path channel in the space domain. Perform visual area extraction to obtain superior estimation accuracy under the premise of low computational complexity.

This method provides a significant performance gain in terms of mean square error, and thanks to the low computational complexity of the algorithm, it can better achieve channel tracking in mobile scenarios under the premise of ensuring accuracy. High feasibility.

The above-mentioned specific implementation can be partially adjusted by those skilled in the art in different ways without departing from the principle and purpose of the present invention. The protection scope of the present invention is subject to the claims and is not limited by the above-mentioned specific implementation. Each implementation within the scope is bound by the present invention.

Claims (2)

1. A space non-stationary channel estimation method based on image contour extraction is characterized in that under a large-scale MIMO space non-stationary channel moving scene, after a received signal with a sparse angle delay domain is represented in an image form, the number of estimated paths, the angle and the time delay of each path are obtained by using an estimation algorithm of image contour extraction, after the received signal with the sparse space delay domain is represented in the image form, an effective visible region estimation corresponding to each channel path is obtained by using the estimation algorithm of image contour extraction, and therefore path gain and channel reconstruction are achieved;

the estimation algorithm for extracting the image contour specifically comprises the following steps:

1) when the two-dimensional coordinate (i, j) of any point pixel on the image is the outer boundary starting point, setting a marker NBD to be NBD +1, recording (i, j) and the updated NBD value, and recording a point (i, j-1) on the left of (i, j) as (i, j)₂，j₂) (ii) a Otherwise, jumping to the step 6);

2) to (i)₂，j₂) Taking (i, j) as a center of a circle as a starting point, and detecting clockwise: when there are non-zero pixels in the upper, lower, left and right four neighborhoods, it is recorded as (i)₁，j₁) And update (i)₂，j₂)＝(i₁，j₁) (i, j) is denoted as (i)₃，j₃) (ii) a Otherwise the point binarizes the pixel value

And jumping to step 6);

3) with (i)₃，j₃) As the center of circle, in (i)₂，j₂) The previous point which is the starting point is detected anticlockwise: center point (i)₃，j₃) When there are non-zero pixels above, below, left and right, it is recorded as (i)₄，j₄)；

4) When (i)₃，j₃+1) is the zero pixel point detected in step 3), then the binary pixel value of the point

When (i)₃，j₃+1) are not zero pixels detected in step 3) and satisfy

When it is, then

Otherwise

The value of (a) does not change;

5) when the outer boundary starting point of the current contour is searched, i₄，j₄) (ii) and (i, j)₃，j₃)＝(i₁，j₁) If yes, jumping to the step 6); otherwise update (i)₂，j₂)＝(i₃，j₃)，(i₃，j₃)＝(i₄，j₄) And jumping to the step 3);

6) when the pixel value is binarized at the point

When it is time, the marker is updated

And starting to continue raster scanning detection from the pixel point (i, j +1) until scanning to the vertex of the lower right corner of the image;

the estimated path number, the signal departure angle of each path and the propagation delay are obtained by the following modes:

firstly, in a large-scale MIMO channel, when the number of paths is far less than that of antennas, the received signals corresponding to the channel and the all-1 pilot frequency sequence have sparse characteristics in an angle time delay domain, and the received signals are converted from a space frequency domain to the angle time delay domain

Then the received signal with sparse angular time delay domain is represented in the form of image, namely, the received signal is represented in the form of image

Each element of

Is further scaled to

Based on scaled received signals

Generating corresponding angle time delay domain sparse received signal color images and gray level images, wherein: d_aAnd D_SRespectively being front N_rLine and first N_sN of a line_rDimension and η N_sA discrete Fourier transform matrix of dimension, η being the over-sampling rate, function max (·) being used to obtain the maximum element modulus value in the input matrix;

secondly, on the basis of the gray image of the original received signal, setting a threshold delta to the pixel value f of each pixel point (i, j) in the gray image_i，jPerforming binarization processing, namely:

setting markers NBD ═ 1 and LNBD ═ 0, wherein NBD and LNBD are respectively a new boundary and a previous new boundary, traversing each pixel point of the image in a raster scanning mode, resetting LNBD ═ 0 when a new line starting position is scanned, and only when the pixel point (i, j) is detected to be an outer boundary starting point, namely the pixel point (i, j) is detected to be an outer boundary starting point

And is

When the current LNBD is less than or equal to 0, extracting the outline of the image;

thirdly, after all the outlines are extracted, the pixel value of the starting point of the outer boundary of the last outline is the estimated path number

Then, taking the starting point (i, j) of the outer boundary of each contour as the initial point, calculating the current second point

The horizontal and vertical pixel points of the profile are used to obtain the height h of the profile_lAnd width w_lCombining the coordinates of the starting point of the outer boundary to obtain the coordinates of the central point of the rectangle surrounded by the current contour

Obtaining a signal starting angle corresponding to the ith channel path according to the central coordinate

And delay estimation results

The estimation of the effective visible area corresponding to each channel path is obtained by the following method:

i) converting the received signal from the spatial frequency domain to the spatial time delay domain:

and to

Each element of

Further scaling to:

based on scaled received signals

Generating a corresponding space time delay domain sparse received signal color image and a corresponding gray image;

ii) setting a threshold value delta to complete binarization preprocessing on the pixel value of each pixel point of the image, and performing contour extraction on the image after the execution condition of a contour extraction algorithm is met;

iii) after all the contours are extracted, taking the starting point (i, j) of the outer boundary of each contour as the initial point, calculating the current second contour

The horizontal and vertical pixel points of the profile are used to obtain the height h of the profile_lAnd width w_l(ii) a Then combining the coordinates of the starting point of the outer boundary to obtain the coordinates of the transverse center point of the rectangle surrounded by the current contour

From the close image center coordinate x_lObtaining the time delay information corresponding to the ith channel path according to the longitudinal coordinate j of the starting point of the ith outer boundary

Visual area

Further matching the estimated angle and time delay information with the visible area to obtain the corresponding parameter information of each channel path

The path gain refers to: and according to the angle and the time delay of each path and the effective visible area corresponding to each channel path, obtaining the gain of each path by a least square method:

wherein:

and vec (·) represents an operator for matrix pseudo-inversion and vectorization of the matrix by columns;

the channel estimation refers to: substituting the total path number obtained by estimation and channel parameters of each path into a wireless channel model to reconstruct a space non-stationary channel:

wherein:

in order to estimate the number of paths,

and

the estimated values of the gain, the signal departure angle, the visible area and the propagation delay of the path l are obtained, a (-) and q (-) are airspace and frequency domain rudder vectors corresponding to the angle and the delay, and p (-) is used for selecting an effective antenna index in the corresponding visible area;

the channel parameters of each path refer to: signal departure angle, signal propagation delay, effective visible area estimation and path gain.

2. The image contour extraction based spatial negation of claim 1The stationary channel estimation method is characterized in that after a channel is reconstructed, channel tracking is further realized by the following method: the method includes that whether channel space parameters, namely angles, time delays and visual areas, of a current time slot t are changed and updated correspondingly according to the visual areas, specifically: when channel reconstruction is carried out on the time slot t, channel visible region estimation based on image contour extraction is carried out, and when channel reconstruction is carried out on the time slot t

When the channel estimation is established, path gain and channel estimation are directly carried out, so that the processing time delay required by the channel estimation is shortened; otherwise, complete channel space parameters and path gain updating estimation are carried out, so that the spatial non-stationary wireless channel of the time slot t is reconstructed.

Images