A lightweight MIMO-OFDM underwater acoustic channel estimation method to mitigate the impact of spectrum leakage

Abstract

The present invention relates to a lightweight MIMO-OFDM underwater acoustic channel estimation method for mitigating the influence of spectrum leakage. An underwater user sends a pilot signal to a UWA base station, and the UWA base station converts the received signal to generate a corresponding sparse image in the angle-delay domain; the sparse image is binarized, and the center point of each bright spot is detected; due to the spectrum leakage phenomenon, there will be a point-like tailing phenomenon around the bright spot in the image, and the tailing noise of the central bright spot is deleted to ensure the estimation accuracy of the channel and the number of multipaths; the angle and delay of the channel are calculated according to the corresponding relationship between the channel parameters and the coordinates of the center point, and the complex fading factor is obtained, and then the channel matrix is ​​reconstructed through the channel model. The present invention adopts a lightweight detection method based on the characteristics of the sparse image after receiving the pilot domain conversion, and mitigates the problem of serious errors in the number of paths estimated due to spectrum leakage, and realizes relatively accurate MIMO-OFDM underwater acoustic channel estimation under the premise of low complexity.

Description

Technical Field

[0001] The present invention belongs to the technical field of underwater acoustic communication channel estimation, and in particular to a lightweight MIMO-OFDM underwater acoustic channel estimation method for alleviating the influence of spectrum leakage. Background Art

[0002] Multiple-input multiple-output (MIMO) technology can significantly improve spectrum efficiency through parallel transmission. It is very attractive for bandwidth-limited underwater acoustic channels, and the combination of MIMO and OFDM technology will further improve its communication performance. Underwater acoustic channel estimation is of great significance to both UWA (Underwater Acoustic) base stations and underwater users. With the rapid development of neural networks, the use of deep learning for MIMO underwater acoustic channel estimation has achieved good results. However, it usually has problems such as high training cost, poor generalization, and complex network structure.

[0003] After searching the existing literature, it was found that the Chinese patent application number is CN201911372155.7, and the name is "A method for fast reconstruction of downlink channels based on deep learning". This method uses the spatial reciprocity of the frequency division multiplexing system to estimate frequency-independent parameters in the uplink channel and use them as downlink channel parameters, which effectively reduces the burden and overhead of the feedback link and obtains better channel estimation results. However, in the resource-limited underwater acoustic communication environment, it is necessary to consider the storage of the model and the complexity of the algorithm. At the same time, the spectrum leakage problem is also worthy of attention. Therefore, it is necessary to study a lightweight MIMO-OFDM underwater acoustic channel estimation method that mitigates the impact of spectrum leakage. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and propose a lightweight MIMO-OFDM underwater acoustic channel estimation method that reduces the impact of spectrum leakage. The method can adopt a lightweight detection method based on the characteristics of the sparse image after the received pilot domain conversion, and reduce the problem of serious errors in path number estimation caused by spectrum leakage, thereby achieving relatively accurate MIMO-OFDM underwater acoustic channel estimation under the premise of low complexity.

[0005] The present invention solves the technical problem by adopting the following technical solutions:

[0006] A lightweight MIMO-OFDM underwater acoustic channel estimation method for mitigating the influence of spectrum leakage comprises the following steps:

[0007] Step 1: The underwater user sends a pilot signal to the UWA base station, and the UWA base station converts the received signal and generates a corresponding sparse image in the angle-delay domain;

[0008] Step 2: Binarize the sparse image to form a binary image γ, and search the range of 1 value according to the 0-1 value distribution to detect the center point of each bright spot. The point set formed is The corresponding bright spot set is in is the estimated preliminary channel multipath quantity;

[0009] Step 3: Delete the trailing noise at the center of the bright spot to ensure the accuracy of the estimation of the number of channels and multipaths;

[0010] Step 4: Calculate the angle and delay of the channel based on the corresponding relationship between the channel parameters and the coordinates of the bright spot center after removing the trailing noise, and then obtain the complex fading factor, and finally reconstruct the channel matrix through the channel model.

[0011] Furthermore, the step 1 comprises the following steps:

[0012] Step 1.1, the underwater user sends an all-one pilot signal to the multi-transducer UWA base station. After the pilot signal sent by the underwater user passes through the underwater acoustic channel, the signal is represented as:

[0013]

[0014] Where a(Θ)=[1,e j2πΘ ,...,e j2π(M-1)Θ ] T , q(Γ)=[1,e j2πΓ ,...,e j2π(N-1)Γ ] T , M, N and A are the number of transducers, the number of subcarriers and the transmission power respectively, Z is Gaussian white noise, L is the number of propagation paths, l is the path number, is the complex fading factor;

[0015] Step 1.2: Receive at the UWA base station and perform domain conversion and normalization.

[0016] Furthermore, the UWA base station is composed of transducers arranged uniformly and linearly.

[0017] Furthermore, the step 2 comprises the following steps:

[0018] Step 2.1, generate an angle delay domain sparse image in MATLAB software according to step 1, and use the im2bw function to generate a binary image γ;

[0019] In the second step, the regionprops function is used to find the connected components and their pixel indices in the binary image, and the connected components are counted. The resulting point set is expressed as The corresponding bright spot set is expressed as in To estimate the number of preliminary channel multipaths, χ is the area formed by the value of 1.

[0020] Furthermore, the step 3 comprises the following steps:

[0021] Step 3.1: Determine whether the area of ​​each element in the detected bright spot set P is greater than the threshold, that is, (L spot -6) 2 +ζ, where L spot represents the side length of the bright spot, ζ is evenly distributed in [-50,50], and the elements that meet the conditions in P are divided into the central bright spot set P c ={ k}, k∈[1,v], v represents the number of central bright spots;

[0022] Step 3.2: P c Each element performs the Euclidean distance operation with the elements in P;

[0023] Step 3.3: For the kth central bright spot, replace the item with a distance of 0 with a value greater than the maximum value, and transform the vector d formed by the Euclidean distance k Sort the elements from low to high, take the first four elements and record their corresponding indexes, which are recorded as

[0024] Step 3.4, judgment Whether all indexes in the same table satisfy the following conditions: the corresponding area is smaller than the area threshold κ area And the Euclidean distance between the bright spot and the center is lower than the distance threshold κ dist , those that meet the conditions are stored in the tailing noise set Ω of the kth central bright spot k ,in, κ dist =L ima / 10,L ima is the image side length, β is 0.5, M is the number of transducers;

[0025] Step 3.5, loop v times to get the final set of trailing noise points Ω = Ω1∪Ω2∪...∪Ω v , the binary image γ after removing the noise points is updated, and the bright spot set formed after the retained center point is judged as P′=P\Ω, and the point set is expressed as

[0026]

[0027] Furthermore, the step 4 comprises the following steps:

[0028] Step 4.1: Get the angles of each path of the final channel Latency in,

[0029] Step 4.2: The link complex fading factor is obtained in, Among them, y is the column vector formed by stacking the received signal Y, M and N are the number of transducers and subcarriers respectively, so the channel estimation result is

[0030]

[0031] The advantages and positive effects of the present invention are:

[0032] The present invention sends a pilot signal to a UWA base station through an underwater user, and the UWA base station converts the received signal and generates a corresponding sparse image in the angle-delay domain; the sparse image is binarized and the center point of each bright spot is detected; due to spectrum leakage, there will be a point-like tailing phenomenon around the bright spot in the image, and the tailing noise of the central bright spot is deleted to ensure the accuracy of the estimation of the channel and the number of multipaths; the angle and delay of the channel are calculated according to the corresponding relationship between the channel parameters and the coordinates of the center point, and the complex fading factor is obtained, and then the channel matrix is ​​reconstructed through the channel model. The present invention adopts a lightweight detection method based on the characteristics of the sparse image after the conversion of the received pilot domain, and alleviates the problem of serious errors in the number of paths estimated due to spectrum leakage, and realizes relatively accurate MIMO-OFDM underwater acoustic channel estimation under the premise of low complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a flow chart of underwater acoustic channel estimation of the present invention;

[0034] Figure 2 This is a result diagram of target detection using a pilot sparse image of an underwater acoustic channel in which the number of transducers and the number of subcarriers are both 32;

[0035] Figure 3 is a comparison diagram of the path estimation numbers of the underwater acoustic channel pilot sparse image before and after being processed by the method described in step 3 of the present invention;

[0036] Figure 4 It is a simulation comparison chart of the estimation accuracy of the method of the present invention and the channel estimation algorithm based on YOLOv3. DETAILED DESCRIPTION

[0037] The present invention is further described in detail below with reference to the accompanying drawings.

[0038] A lightweight MIMO-OFDM underwater acoustic channel estimation method that mitigates the impact of spectrum leakage, such as Figure 1 As shown, the following steps are included:

[0039] Step 1: The underwater user sends a pilot signal to the UWA (Underwater Acoustic) base station. The UWA base station converts the received signal and generates a corresponding sparse image in the angle-delay domain.

[0040] Step 1.1: The underwater user sends an all-one pilot signal to the multi-transducer UWA base station. The UWA base station is composed of transducers arranged uniformly and linearly. After the pilot signal sent by the underwater user passes through the underwater acoustic channel, the signal received by the base station is:

[0041]

[0042] Where a(Θ)=[1,e j2πΘ ,...,e j2π(M-1)Θ ] T , q(Γ)=[1,e j2πΓ ,...,e j2π(N-1)Γ ] T , M, N and A are the number of transducers, the number of subcarriers and the transmission power respectively, Z is Gaussian white noise, L is the total number of propagation paths, l is the path number, is the complex fading factor;

[0043] Step 1.2: Receive at the UWA base station and perform domain conversion, normalization and other processing to facilitate subsequent channel estimation.

[0044] The specific process of domain conversion is as follows: in:

[0045]

[0046] And normalize it. Generate the corresponding angle-delay domain sparse image.

[0047] Step 2: Binarize the sparse image to form a binary graph γ, use the regionprops function to find the connected components and their pixel indexes in the binary image and count the connected components. And search the range of 1 value according to the 0-1 value distribution, detect the center point of each bright spot, and the point set formed is The corresponding bright spot set is in is the estimated preliminary channel multipath quantity.

[0048] Step 2.1, generate an angle delay domain sparse image in MATLAB software according to step 1, and use the im2bw function to generate a binary image γ, with the parameter set to 0.2;

[0049] In the second step, the regionprops function is used to find the connected components and their pixel indexes in the binary image and count the connected components, that is, to search the range of 1 values ​​and detect the center points of each bright spot. The point set formed is expressed as The corresponding bright spot set is expressed as in To estimate the number of preliminary channel multipaths, χ is the area formed by the value of 1.

[0050] Step 3: Delete the trailing noise at the center of the bright spot to ensure the accuracy of the estimation of the channel and multipath quantity.

[0051] There will be a dot-like tailing phenomenon around the bright spot in the image. l It may be the central bright spot or the trailing noise around the central bright spot. This will lead to inaccurate estimation of the number of paths, so the "trailing noise" of the central bright spot must be deleted, that is, the central bright spot is selected and the surrounding points of each central bright spot are judged.

[0052] Step 3.1: Determine whether the area of ​​each element in the detected bright spot set P is greater than the threshold, that is, (L spot -6) 2 +ζ, where L spot represents the side length of the bright spot, ζ is evenly distributed in [-50,50], and the elements that meet the conditions in P are divided into the central bright spot set P c ={ k}, k∈[1,v], v represents the number of central bright spots;

[0053] Step 3.2: P c Each element performs the Euclidean distance operation with the elements in P;

[0054] Step 3.3: For the kth central bright spot, replace the item with a distance of 0 with a value greater than the maximum value, and transform the vector d formed by the Euclidean distance k Sort the elements from low to high, take the first four elements and record their corresponding indexes, which are recorded as

[0055] Step 3.4, judgment Whether all indexes in the same table satisfy the following conditions: the corresponding area is smaller than the area threshold κ area And the Euclidean distance between the bright spot and the center is lower than the distance threshold κ dist , those that meet the conditions are stored in the tailing noise set Ω of the kth central bright spot k ,in, κ dist =L ima / 10,L ima is the image side length, β is 0.5, M is the number of transducers;

[0056] Step 3.5, loop v times to get the final set of trailing noise points Ω = Ω1∪Ω2∪...∪Ω v , the binary image γ after removing the noise points is updated, and the bright spot set formed after the retained center point is judged as P′=P\Ω, and the point set is expressed as

[0057] Step 4: Calculate the angle and delay of the channel based on the corresponding relationship between the channel parameters and the coordinates of the bright spot center after removing the trailing noise, and then obtain the complex fading factor, and finally reconstruct the channel matrix through the channel model.

[0058] Step 4.1: Obtain the angles of each path of the final channel from the corresponding relationship between the channel parameters and the center point coordinates Latency in, Here, according to the image pixel index direction, the coordinate values ​​increase from left to right and from top to bottom, with the upper left corner as the coordinate origin;

[0059] Step 4.2: The link complex fading factor is obtained in, Among them, y is the column vector formed by stacking the received signal Y, M and N are the number of transducers and subcarriers respectively, so the channel estimation result is Calculate the normalized mean square error of the channel estimate in

[0060] like Figure 2 The figure shows the target detection result of the underwater acoustic channel pilot sparse image based on the lightweight MIMO-OFDM underwater acoustic channel estimation method for mitigating the influence of spectrum leakage of the present invention, where the number of transducers and the number of subcarriers are both 32. The center points of the bright spots are marked in the first and third columns. The first column represents the sparse image after binarization, the second column represents the binary image after deleting the noise, and the third column is the final marking result. From this, a point set is generated. It can be seen that after processing, the trailing noise around is successfully removed. Figure 3Shown is a comparison chart of the path estimation numbers of the underwater acoustic channel pilot sparse image before and after being processed by the method described in step three based on a lightweight MIMO-OFDM underwater acoustic channel estimation method for mitigating the impact of spectrum leakage according to the present invention. Among them, (a) represents the case where the SNR is 10, the number of transducers and the number of subcarriers are both 32, and the number of paths is 3; (b) represents the case where the SNR is 10, the number of transducers and the number of subcarriers are both 32, and the number of paths is 6; (c) represents the case where the SNR is 10, the number of transducers and the number of subcarriers are both 32, and the number of paths is 9; (d) represents the case where the SNR is 10, the number of transducers and the number of subcarriers are both 64, and the number of paths is 3. It can be seen that the multipath number estimation is improved compared to before processing. Figure 4 As shown, (a) represents the case where the number of transducers and subcarriers are both 32 and the number of paths is 3; (b) represents the case where the number of transducers and subcarriers are both 32 and the number of paths is 6. It can be seen that the channel estimation result of the method of the present invention is good, and compared with the channel estimation algorithm based on YOLOv3, there is no need to prepare a data set or train a model.

[0061] It should be emphasized that the embodiments described in the present invention are illustrative rather than restrictive. Therefore, the present invention includes but is not limited to the embodiments described in the specific implementation manner. Any other implementation manners derived by those skilled in the art based on the technical solution of the present invention also fall within the scope of protection of the present invention.

Claims

1. A lightweight MIMO-OFDM underwater acoustic channel estimation method for mitigating the impact of spectrum leakage, characterized by: The following steps are involved: Step 1: The underwater user sends a pilot signal to the UWA base station, and the UWA base station converts the received signal and generates a corresponding sparse image in the angle-delay domain; Step 1.1: The underwater user sends an all-one pilot signal to the multi-transducer UWA base station. The UWA base station is composed of transducers arranged uniformly and linearly. After the pilot signal sent by the underwater user passes through the underwater acoustic channel, the signal received by the base station is: Where a(Θ)=[1,e j2πΘ ,...,e j2π(M-1)Θ ] T , q(Γ)=[1,e j2πΓ ,...,e j2π(N-1)Γ ] T , M, N and A are the number of transducers, the number of subcarriers and the transmission power respectively, Z is Gaussian white noise, L is the total number of propagation paths, l is the path number, is the complex fading factor; Step 1.2: Receive at the UWA base station and perform domain conversion and normalization processing to facilitate subsequent channel estimation; The specific process of domain conversion is as follows: in: And normalize it. Generate a corresponding angle-delay domain sparse image; Step 2: Binarize the sparse image to form a binary image Υ, and search the range of 1 value according to the 0-1 value distribution to detect the center point of each bright spot. The point set formed is Π={(x l ,y l )}, The corresponding bright spot set is P = {χ l }, in is the estimated preliminary channel multipath quantity; Step 3: Delete the trailing noise at the center of the bright spot to ensure the accuracy of the estimation of the number of channels and multipaths; There will be a dot-like tailing phenomenon around the bright spot in the image. l It may be the central bright spot or the trailing noise around the central bright spot. This will lead to inaccurate estimation of the number of paths. It is necessary to delete the trailing noise of the central bright spot, select the central bright spot, and make judgments on the surrounding points of each central bright spot. Step 3.1: Determine whether the area of ​​each element in the detected bright spot set P is greater than the threshold, (L spot -6) 2 +ζ, where L spot Represents the side length of the bright spot, ζ is evenly distributed in [-50,50], and the elements that meet the conditions in P are divided into the central bright spot set P c ={ k }, k∈[1,v], v represents the number of central bright spots; Step 3.2: c Each element performs the Euclidean distance operation with the elements in P; Step 3.3: For the kth central bright spot, replace the item with a distance of 0 with a value greater than the maximum value, and transform the vector d formed by the Euclidean distance k Sort the elements from low to high, take the first four elements and record their corresponding indexes, which are recorded as j=1,2,..,4; Step 3.4, judgment Whether all indexes in the same table satisfy the following conditions: the corresponding area is smaller than the area threshold κ area And the Euclidean distance between the bright spot and the center is lower than the distance threshold κ dist , those that meet the conditions are stored in the tailing noise set Ω of the kth central bright spot k ,in, κ dist =L ima / 10,L ima is the image side length, β is 0.5, M is the number of transducers; Step 3.5, loop v times to get the final set of trailing noise points Ω = Ω1∪Ω2∪...∪Ω v , the binary image Υ after removing the noise points is updated, and the bright spot set formed after the retained center point is judged as Ρ′=Ρ\Ω, and the point set is expressed as Π′={(x l ′,y l ′)}, Step 4: Calculate the angle and delay of the channel based on the corresponding relationship between the channel parameters and the coordinates of the bright spot center after removing the trailing noise, and then obtain the complex fading factor, and finally reconstruct the channel matrix through the channel model. Step 4.1: Obtain the angles of each path of the final channel from the corresponding relationship between the channel parameters and the center point coordinates Latency in, Here, according to the image pixel index direction, the upper left corner is taken as the coordinate origin, and the coordinate values ​​increase from left to right and from top to bottom; Step 4.2: The link complex fading factor is obtained in, Among them, y is the column vector formed by stacking the received signal Y, M and N are the number of transducers and subcarriers respectively, so the channel estimation result is Calculate the normalized mean square error of the channel estimate in 2. According to claim 1, a lightweight MIMO-OFDM underwater acoustic channel estimation method for mitigating the influence of spectrum leakage is characterized by: The step 2 comprises the following steps: Step 2.1, generate an angle delay domain sparse image in MATLAB software according to step 1, and use the im2bw function to generate a binary image Y; In the second step, the regionprops function is used to find the connected components and their pixel indices in the binary image, and the connected components are counted. The point set formed is represented as Π={(x l ,y l )}, The corresponding bright spot set is expressed as P = {χ l }, in To estimate the number of preliminary channel multipaths, χ is the area formed by the value of 1.

Images