Channel estimation method and system based on adaptive synchronous prediction orthogonal matching pursuit

Abstract

The invention provides a channel estimation method and system based on adaptive synchronous prediction orthogonal matching pursuit, and the method comprises the steps: inserting a pilot signal into a transmitted signal, and generating a baseband signal; up-modulating the baseband signal to a specific frequency band, and transmitting a frequency band signal; receiving the frequency band signal, performing synchronous operation, performing down-modulation to a baseband, and generating a demodulation baseband signal; decomposing and demodulating the baseband signal, and estimating to obtain a sparse channel; and carrying out time delay compensation on the demodulation baseband signal according to the sparse channel, and recovering the original signal. According to the method, the pre-selection set is selected first, then optimization calculation is carried out in sequence, and the situation that zero coefficient tap atoms are wrongly selected due to too large inner products caused by noise pollution can be effectively avoided; through synchronous operation, the influence caused by individual pilot signals with relatively large errors is avoided; and only one subtraction and one division are added in each iteration, so that the performance is improved, and the calculation efficiency is also maintained.

Description

Technical Field

[0001] The present invention belongs to the field of underwater acoustic channel estimation, and in particular, relates to a channel estimation method and system based on adaptive synchronous prediction orthogonal matching pursuit. Background Art

[0002] With the continuous advancement of science and technology, the exploration of the ocean is getting deeper and deeper, and marine technology has also been well developed, but it also faces more challenges and technical difficulties. In underwater acoustic communication, due to the severe absorption and attenuation of high-frequency signals, medium and low-frequency signals are often required in long-distance underwater acoustic communication, and the available bandwidth of the channel is very narrow, so the spectrum resources are very limited. How to effectively use spectrum resources for communication is the key to improving communication quality. Orthogonal frequency division multiplexing (OFDM) technology converts serial high-speed data into parallel data, and then maps the parallel data to multiple orthogonal and partially overlapping subcarriers, and the high-speed data stream becomes several low-speed sub-data streams. Compared with traditional multi-carrier technology, OFDM greatly improves spectrum utilization due to the overlapping of subcarriers. However, in underwater acoustic communication, due to the multipath characteristics of the channel, especially in shallow water communication, the multipath effect is more obvious. The delay spread caused by the multipath effect brings serious frequency selective fading problems to the OFDM system. Therefore, in order to ensure the effective communication of OFDM, it is very important to estimate and compensate the underwater acoustic channel.

[0003] Traditional channel estimation methods, such as LMS, MMSE and ML, insert known pilot signals into the transmitted signal and use these pilot signals for channel estimation at the receiving end. The advantage is low algorithm complexity and simple calculation. However, when there is severe delay spread and Doppler spread in the underwater acoustic channel, more pilot signals need to be added to ensure good estimation performance, but inserting more pilot signals will reduce transmission efficiency. In addition, traditional channel estimation methods assume that the paths of the channel are densely distributed, so it is necessary to calculate the tap coefficients of each time domain impulse response corresponding to the channel. However, in the actual underwater acoustic channel, the channel tap energy is mainly concentrated on a small number of paths, and most of the tap coefficients are close to or equal to zero, so calculating each tap coefficient wastes computing resources. In fact, only a small number of tap coefficients are not zero, which can be understood as the sparseness of the channel, and the sparse theory can be used to solve the channel.

[0004] Based on the sparse characteristics of the underwater acoustic channel, a dictionary matrix is ​​constructed according to the time delay, and the received symbols of the pilot subcarriers are used as observation values, and the sparse channel response is reconstructed using a small number of atoms in the dictionary. At present, many sparse channel estimation methods have been proposed. The existing channel estimation methods based on sparse theory lack adaptability to different pilot signals and channels due to the limitations of hyperparameters in conventional sparse optimization models. They need to manually set parameters to obtain channels that meet the requirements, and are difficult to adapt to different channels in actual applications.

[0005] The patent document "An Orthogonal Matching Pursuit Channel Estimation Method for Underwater Acoustic OFDM System" (CN113055317A) discloses that based on the multipath sparse characteristics of the underwater acoustic channel, the sparsity of the channel is estimated in advance using wavelet transform technology as a condition for updating and stopping iterations in the OMP algorithm, but it still requires repeated estimation of taps.

[0006] The patent document "Sparse Underwater Acoustic Orthogonal Frequency Division Multiplexing Channel Estimation Method and Device Based on Basis Pursuit Denoising" (CN110138459A) discloses that the regularization parameter is corrected according to the signal-to-noise ratio and the sparse signal matrix to adapt to the noise change, and the sparse channel estimation result obtained is debiased to obtain the final channel estimation result. However, there is still the influence of individual pilot signals with large errors, and the calculation efficiency is low.

[0007] Sparse channel estimation methods based on convex optimization algorithms, such as the alternating direction multiplier method and the soft threshold iteration algorithm, essentially estimate all the taps in the channel, and mainly perform optimization calculations through a large number of iterations. As for greedy algorithms such as matching pursuit and orthogonal matching pursuit, although they do not require multiple iterations, they require prior information on the number of non-zero taps in the channel, and are more sensitive to noise and are easily disturbed by noise, resulting in incorrect tap estimation. Especially in shallow sea environments, the channel paths are densely distributed. Under low signal-to-noise ratio conditions, conventional methods are difficult to accurately estimate multiple paths with small delay differences.

[0008] Therefore, there is an urgent need for a channel estimation method and system that does not require a large number of iterative optimizations and can still accurately estimate under low signal-to-noise ratio conditions. Summary of the invention

[0009] In view of the defects in the prior art, the object of the present invention is to provide a channel estimation method and system based on adaptive synchronous prediction orthogonal matching pursuit.

[0010] According to the present invention, a channel estimation method based on adaptive synchronous prediction orthogonal matching pursuit is provided, comprising:

[0011] Step S1: inserting a pilot signal into a transmission signal to generate a baseband signal;

[0012] Step S2: up-modulate the baseband signal to a specific frequency band and transmit the frequency band signal;

[0013] Step S3: receiving the frequency band signal and performing synchronization operation, down-modulating the frequency to the baseband, and generating a demodulated baseband signal;

[0014] Step S4: Decompose the demodulated baseband signal and estimate a sparse channel;

[0015] Step S5: Delay compensation of the demodulated baseband signal is performed according to the sparse channel to restore the original signal.

[0016] Preferably, the step S1 comprises:

[0017] Step S1.1: Set the pilot signal to a complex number 3+i3 with the same elements, and determine that the pilot insertion mode is a block pilot;

[0018] Step S1.2: Input the transmit signal, sort it by column, insert pilot signals at equal intervals by column, and generate a complete baseband signal.

[0019] The step S2 comprises:

[0020] Step S2.1: The baseband signal is loaded onto mutually orthogonal and overlapping subcarriers through orthogonal frequency division multiplexing technology, and each element of each column of the signal corresponds to a subcarrier on the baseband;

[0021] Step S2.2: modulate each subcarrier from a baseband signal to a corresponding frequency band signal by inverse discrete Fourier transform to generate a Rayleigh channel;

[0022] Step S2.3: The modulated frequency band signal is transmitted through the transmitting transducer via the Rayleigh channel.

[0023] The step S3 comprises:

[0024] Step S3.1: The hydrophone receives the frequency band signal transmitted through the Rayleigh channel, and performs time synchronization processing on the received frequency band signal through an autocorrelation operation;

[0025] Step S3.2: Down-modulate the received frequency band signal to baseband through discrete Fourier transform operation to generate a demodulated baseband signal.

[0026] Preferably, the step S4 comprises:

[0027] Step S4.1: Construct an overcomplete dictionary D=diag(s p )F;

[0028] Among them, s p ∈C L , represents the pilot signal;

[0029] diag(·) represents the symbol of vector to diagonal matrix;

[0030] F∈C N×N , represents the discrete Fourier transform matrix;

[0031] D∈C L×N , represents an overcomplete dictionary;

[0032] N represents the length of the channel;

[0033] L represents the inverse Fourier transform length;

[0034] C represents the complex field.

[0035] Step S4.2: Construct a sparse decomposition model for signal estimation based on the overcomplete dictionary

[0036] in, represents the square of the largest singular value;

[0037] ‖·‖0 represents the number of non-zero elements in the vector;

[0038] z p represents a received pilot time domain signal;

[0039] h represents the underwater acoustic channel.

[0040] Step S4.3: Use the algorithm to solve the sparse decomposition model to obtain the estimated sparse channel

[0041] Preferably, the algorithm in step S4.3 is an ASFOMP algorithm, comprising:

[0042] Step S4.3.1: Initialize parameters, R p,i =z p,i , A=[], D1=D′D, D2=D′z, S=zeros(N,M), Gr=[], k=0, Re=[].

[0043] Step S4.3.2: Let k = k + 1, calculate the atomic D j and R p,i The inner product of

[0044] Step S4.3.3: Calculate the final residuals of the first P atoms respectively,

[0045] Step S4.3.4: Select the atom with the smallest final residual and record its sequence

[0046] Step S4.3.4: Store I into A, A = A ∪ I, and calculate R p,i The projection x on A = D1(A,A)\D2(A,:), the current signal R p,i = z p,i - D(:,A)x.

[0047] Step S4.3.5: Store it in Re,

[0048] Step S4.3.5: Judge the value of k. If 2 > k > 1, then let If k > 2, then Gr(k - 1) < w*mean(Gr(1:k - 2)). If 1 > k, then execute Step S4.3.2.

[0049] Step S4.3.6: Assign x to S, S(A,:) = x, take the mean value of each row in S, S = mean(S), and output the estimated channel

[0050] Among them, R p,i represents the error of the i-th pilot signal in the iteration;

[0051] z p,i represents the i-th pilot signal;

[0052] A represents the set of selected atom numbers;

[0053] D represents the overcomplete dictionary;

[0054] D′ represents the conjugate of D;

[0055] N represents the length of the channel;

[0056] M represents the number of pilots;

[0057] zeros(·) represents the operation of creating a zero matrix;

[0058] S represents the coefficient matrix;

[0059] The subscript j represents the number of the atom in the overcomplete dictionary D;

[0060] R F represents the final residual of the atom;

[0061] The superscript F represents the final residual;

[0062] I represents the number of the atom with the smallest final residual;

[0063] Re represents the set of errors in each iteration;

[0064] Gr represents the error gradient of each iteration;

[0065] x represents the received frequency band signal;

[0066] mean(·) represents the mean operation;

[0067] k represents the number of cycles;

[0068] P represents the pre-selection step size;

[0069] sort() represents descending sort operation;

[0070] g i represents the i-th element in G;

[0071] Indicates that the current optimal atom g i The final residual obtained by synchronous orthogonal matching pursuit with sparsity K.

[0072] Preferably, the step S5 comprises:

[0073] Step S5.1: Sparse channel The frequency domain response is obtained by fast Fourier transform

[0074] Step S5.2: Perform channel compensation to restore the original signal Calculate the bit error rate by comparing with the transmitted frequency band signal;

[0075] Wherein, Z represents the frequency domain of the received frequency band signal;

[0076] represents the original signal after channel compensation;

[0077] FFT() stands for Fast Fourier Transform;

[0078] IFFT() stands for Inverse Fast Fourier Transform.

[0079] According to the present invention, a channel estimation system based on adaptive synchronous prediction orthogonal matching pursuit is provided, comprising:

[0080] Module M1: inserts a pilot signal into the transmitted signal to generate a baseband signal;

[0081] Module M2: up-modulates the baseband signal to a specific frequency band and transmits the frequency band signal;

[0082] Module M3: Receives the frequency band signal and performs synchronization operation, down-modulates the frequency to the baseband, and generates a demodulated baseband signal;

[0083] Module M4: decomposes and demodulates the baseband signal and estimates the sparse channel;

[0084] Module M5: Delay compensation of the demodulated baseband signal according to the sparse channel to restore the original signal.

[0085] Preferably, the module M1 comprises:

[0086] Module M1.1: Set the pilot signal to a complex number 3+i3 with the same elements, and determine the pilot insertion mode to be a block pilot;

[0087] Module M1.2: Input the transmit signal, sort it by column, insert pilot signals at equal intervals in the columns, and generate a complete baseband signal.

[0088] The module M2 comprises:

[0089] Module M2.1: The baseband signal is loaded onto mutually orthogonal and overlapping subcarriers through orthogonal frequency division multiplexing technology. Each element of each column of the signal corresponds to a subcarrier on the baseband.

[0090] Module M2.2: modulate each subcarrier from the baseband signal to the corresponding frequency band signal by inverse discrete Fourier transform to generate a Rayleigh channel;

[0091] Module M2.3: transmits the modulated frequency band signal through the Rayleigh channel through the transmitting transducer.

[0092] The module M3 comprises:

[0093] Module M3.1: The hydrophone receives the frequency band signal transmitted through the Rayleigh channel and performs time synchronization processing on the received frequency band signal through autocorrelation operation;

[0094] Module M3.2: Down-modulate the received frequency band signal to baseband through discrete Fourier transform operation to generate a demodulated baseband signal.

[0095] Preferably, the module M4 comprises:

[0096] Module M4.1: Construct an overcomplete dictionary D = diag(s) based on the pilot signal and the Fourier transform matrix p )F;

[0097] Among them, s p ∈C L , represents the pilot signal;

[0098] diag(·) represents the symbol of vector to diagonal matrix;

[0099] F∈C N×N , represents the discrete Fourier transform matrix;

[0100] D∈C L×N , represents an overcomplete dictionary;

[0101] N represents the length of the channel;

[0102] L represents the inverse Fourier transform length;

[0103] C represents the complex field.

[0104] Module M4.2: Sparse Decomposition Model for Signal Estimation Based on Overcomplete Dictionary

[0105] in, represents the square of the largest singular value;

[0106] ‖·‖0 represents the number of non-zero elements in the vector;

[0107] z p represents a received pilot time domain signal;

[0108] h represents the underwater acoustic channel.

[0109] Module M4.3: Use the algorithm to solve the sparse decomposition model and obtain the estimated sparse channel

[0110] Preferably, the algorithm in the module M4.3 is the ASFOMP algorithm, comprising:

[0111] Module M4.3.1: Initialization parameters, R p,i =z p,i , A=[], D1=D′D, D2=D′z, S=zeros(N,M), Gr=[], k=0, Re=[];

[0112] Module M4.3.2: Let k = k + 1, calculate the atomic D j and R p,i The inner product of

[0113] Module M4.3.3: Calculate the final residuals of the first P atoms respectively, G = G (1: P),

[0114] Module M4.3.4: Select the atom with the smallest final residual and record its sequence

[0115] Module M4.3.4: Store I in A, A = A∪I, calculate R p,i The projection x on A = D1(A,A)\D2(A,:), the current signal R p,i =z p,i -D(:,A)x;

[0116] Module M4.3.5: Put into Re,

[0117] Module M4.3.5: Judge the value of k. If 2 > k > 1, then let If k > 2, then Gr(k - 1) < w*mwan(Gr(1:k - 2)). If 1 > k, then execute Module M4.3.2;

[0118] Module M4.3.6: Assign x to S, S(A, :) = x. Take the mean value of each row in S, S = mean(S), and output the estimated channel

[0119] where, R p,i represents the error of the i-th pilot signal in the iteration;

[0120] z p,i represents the i-th pilot signal;

[0121] A represents the set of selected atom serial numbers;

[0122] D represents the overcomplete dictionary;

[0123] D′ represents the conjugate of D;

[0124] N represents the length of the channel;

[0125] M represents the number of pilots;

[0126] zeros(·) represents the operation of creating a zero matrix;

[0127] S represents the coefficient matrix;

[0128] The subscript j represents the serial number of the atom in the overcomplete dictionary D;

[0129] R F represents the final residual of the atom;

[0130] The superscript F represents the final residual;

[0131] I represents the serial number of the atom with the smallest final residual;

[0132] Re represents the set of errors in each iteration;

[0133] Gr represents the error gradient in each iteration;

[0134] x represents the received band signal;

[0135] mean(·) represents the operation of taking the mean value;

[0136] k represents the number of loops;

[0137] P represents the pre-selection step size;

[0138] sort() represents descending sort operation;

[0139] g i represents the i-th element in G;

[0140] Indicates that the current optimal atom g i The final residual obtained by synchronous orthogonal matching pursuit with sparsity K.

[0141] Preferably, the module M5 comprises:

[0142] Module M5.1: Sparse Channels The frequency domain response is obtained by fast Fourier transform

[0143] Module M5.2: Perform channel compensation to restore the original signal The bit error rate is calculated by comparing it with the transmitted frequency band signal.

[0144] Wherein, Z represents the frequency domain of the received frequency band signal;

[0145] represents the original signal after channel compensation;

[0146] FFT() stands for Fast Fourier Transform;

[0147] IFFT() stands for Inverse Fast Fourier Transform.

[0148] Compared with the prior art, the present invention has the following beneficial effects:

[0149] 1. The present invention uses multiple pilot signals to perform channel estimation simultaneously through synchronous operation, thereby effectively avoiding the influence of individual pilot signals with large errors.

[0150] 2. The present invention first selects a pre-selected set, and then optimizes the calculated residuals in turn to select the optimal solution, which can effectively avoid the zero coefficient tap atom being contaminated by noise, resulting in excessive inner product and misselection.

[0151] 3. The present invention does not need to repeatedly estimate new taps, and only adds one subtraction and one division in each iteration, thereby improving performance while maintaining computational efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0152] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0153] Figure 1A schematic diagram of the flow of a channel estimation method based on adaptive synchronous prediction orthogonal matching pursuit;

[0154] Figure 2 Schematic diagram of the Ruili channel with ten paths;

[0155] Figure 3 Schematic diagram of channel estimation results of different methods when SNR=0dB;

[0156] Figure 4 Schematic diagram of channel estimation results of different methods when SNR=10dB;

[0157] Figure 5 It is a schematic diagram of comparison between baseband data and data received by different methods when SNR=10dB;

[0158] Figure 6 A schematic diagram comparing the sparsity estimated by ASFOMP and the actual sparsity;

[0159] Figure 7 Schematic diagram of channel estimation error under different signal-to-noise ratios. DETAILED DESCRIPTION

[0160] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0161] In order to improve the estimation performance of the underwater acoustic channel, a channel estimation method based on adaptive synchronous prediction orthogonal matching pursuit is provided according to the present invention. Figure 1 For example, the specific steps include:

[0162] Step S1: Generation of baseband signal: Generate a transmission signal x and insert a pilot signal to obtain a baseband signal.

[0163] Specifically, the step S1 includes:

[0164] Step S1.1: Set the pilot signal and determine the pilot insertion method;

[0165] The pilot signal is set to a complex number 3+i3 with the same elements. There are usually two ways to insert pilots: block pilots and comb pilots. Block pilots are insensitive to time selectivity, while comb pilots are insensitive to Doppler shift. Block pilots are selected.

[0166] Step S1.2: Input the transmission signal and insert the pilot to generate a complete baseband signal.

[0167] Sort the signal by columns and insert pilots at equal intervals in the columns.

[0168] Step S2: Perform an up-frequency modulation operation: modulate the baseband signal to a specific frequency band, and then transmit the signal.

[0169] Specifically, step S2 includes:

[0170] Step S2.1: Loading the baseband signal onto a plurality of mutually orthogonal and overlapping subcarriers through orthogonal frequency division multiplexing technology;

[0171] Each element of each column of the signal corresponds to a subcarrier at the baseband.

[0172] Step S2.2: modulate each subcarrier from the baseband to a specific frequency band by means of inverse discrete Fourier transform;

[0173] Step S2.3: The modulated signal is transmitted through the transmitting transducer.

[0174] Step S3: Receive the signal, perform synchronization operation on the signal, and then down-modulate the signal to baseband.

[0175] Specifically, step S3 includes:

[0176] Step S3.1: Receive the signal transmitted through the channel through the hydrophone, and then use the autocorrelation operation to perform time synchronization processing on the received signal.

[0177] Step S3.2: Demodulate the signal to baseband by a discrete Fourier transform operation.

[0178] In more preferred embodiments, Figure 2 For example, a baseband signal is generated and modulated to generate a Rayleigh channel. Through synchronous operation, multiple pilot signals are used to simultaneously perform channel estimation, effectively avoiding the influence of individual pilot signals with large errors.

[0179] Step S4: construct an overcomplete dictionary based on the pilot signal and the Fourier transform matrix, decompose the processed received signal into the dictionary using the ASFOMP algorithm, and obtain sparse coefficients. According to the sparse coefficients, the signal delay and amplitude are obtained, and then the channel is estimated.

[0180] A sparsity adaptive algorithm based on residual gradient is proposed to search for sparsity, and SFOMP is improved to adaptive synchronous prediction orthogonal matching pursuit (ASFOMP). When the residual gradient is less than the threshold, since the residual decreases with the increase of sparsity, it indicates that the newly estimated taps cannot improve the residual well, and all taps with higher energy have been estimated. Therefore, there is no need to continue estimating new taps, and the algorithm terminates. Such a sparsity adaptive algorithm only adds one subtraction and one division per iteration, so the increased computational effort can be ignored. Specifically, the step S4 includes:

[0181] Step S4.1: Construct an over-complete dictionary based on the pilot signal and the Fourier transform matrix.

[0182] D=diag(s p )F;

[0183] Among them, s p ∈C L , represents the pilot signal, diag(·) represents the vector to diagonal matrix symbol, F∈C N×N , is the discrete Fourier transform matrix, D∈C L×N is a dictionary, N represents the length of the channel, L represents the inverse Fourier transform length, and C represents the complex domain.

[0184] Step S4.2: Construct a sparse decomposition model of signal estimation based on the dictionary of step S4.1.

[0185]

[0186] in, represents the square of the 2-norm, that is, the square of the largest singular value, ‖·‖0 represents the 0-norm, that is, the number of non-zero elements in the vector, and z p is the received pilot time domain signal, and h represents the underwater acoustic channel.

[0187] Step S4.3: Use the ASFOMP algorithm to solve the sparse decomposition model of step S4.2 to obtain the estimated sparse signal The specific process of the ASFOMP algorithm is as follows:

[0188] First, initialize the parameters, signal R p,i =z p,i , A = [], D1 = D′D, D2 = D′z, S = zeros(N,M), Gr = [], k = 0, Re = []; then execute the loop, k = k + 1, calculate each atom D j and R p,i The inner product of and sort, that is Then select the first P atoms: G = G(1:P); calculate the final residuals of these P atoms respectively Then select the atom with the smallest final residual and record its sequence Then store I in A, A = A ∪ I; Next, calculate R p,i Projection on A: x = D1(A, A)\D2(A, :); Current signal R p,i = z p,i - D(:, A)x; Put in Re, If k > 1, then If k > 2, then Gr(k - 1) < w * mean(Gr(1:k - 2)). End the loop. Where: P is the pre - selected step size, sort() is the descending sorting symbol, g i is the i - th element in G, is the final residual obtained by using simultaneous orthogonal matching pursuit under the current optimal atom g i and sparsity K. R p,i represents the error of the i - th pilot signal in the iteration; z p,i represents the i - th pilot signal; A represents the set of selected atom numbers; D represents the over - complete dictionary; D′ represents the conjugate of D; N represents the length of the channel; M represents the number of pilots; zeros(·) represents the operation of creating a zero matrix; S represents the coefficient matrix; The subscript j represents the serial number of the atom in the dictionary D; R F represents the final residual of the atom; The superscript F represents the final residual; I represents the serial number of the atom with the smallest current final residual; Re represents the error set of each iteration; Gr represents the error gradient of each iteration; x represents the received band signal; mean(·) represents the operation of taking the mean; k represents the number of loops.

[0189] Assign x to S, S(A, :) = x, then take the mean of each row in S: S = mean(S), and finally output the estimated signal:

[0190] In each iteration, first select the top P atoms with larger inner products of the current residuals as the pre - selection set, and then independently select each atom for subsequent optimization to obtain P final residuals. The atom corresponding to the smallest residual is the optimal atom in the current iteration. This can effectively avoid the zero - coefficient tap atoms being misselected due to excessive inner products caused by noise pollution, and can accurately estimate multiple dense paths distributed in the shallow - sea complex channel under low SNR conditions.

[0191] Step S5: Perform time - delay compensation on the received demodulated baseband signal according to the channel estimated in step S4 to restore the original signal.

[0192] Specifically, the step S5 includes:

[0193] Step S5.1: The underwater acoustic channel estimated in step S4 The frequency domain response is obtained through fast Fourier transform.

[0194]

[0195] Step S5.2: Perform channel compensation to restore the original signal.

[0196]

[0197] Where Z is the received transmit signal (frequency domain), It is the original signal after channel compensation. After restoring the received original signal, the bit error rate can be analyzed by comparing it with the transmitted original signal. The bit error rate directly reflects the accuracy of channel estimation.

[0198] In more preferred embodiments, main communication parameters in orthogonal frequency division multiplexing (OFDM) communication based on simulated Rayleigh channel are shown in Table 1.

[0199] Table 1. Parameter settings for OFDM communication:

[0200]

[0201]

[0202] In order to simulate channel noise, Gaussian noise is added to the received signal. When the signal-to-noise ratio is 0dB, the channels estimated by different methods are as follows Figure 3 As shown. Figure 3 In the first table of , the channel estimated by OMP cannot estimate the ninth path well, and the amplitudes of the first and fifth paths have obvious errors compared with the true channel. Figure 3 In the second table, the channel estimated by SOMP also does not estimate the ninth path. Figure 3 In the third table of , the proposed method SFOMP successfully estimates all paths, and the amplitude of each path has no significant error. SFOMP requires a priori sparsity K. When K is unknown, ASFOMP can adaptively solve the sparsity problem. The results are shown in Figure 3 As shown in the fourth table of , ASFOMP accurately estimates 9 paths, which means that the estimated sparsity is 9, which is close to the true sparsity K.

[0203] The estimated channel error is shown in Table 2. SFOMP has the smallest error, and ASFOMP has a smaller error than the three compared methods. Table 3 shows the bit error rates of various methods. SFOMP has the lowest bit error rate, while ASFOMP has a lower bit error rate than the other three methods. This proves that the proposed SFOMP and ASFOMP methods have excellent channel estimation capabilities. Figure 4,Tables 4 and 5 respectively show the channel estimation results, channel estimation error and bit error rate of different methods when the SNR is 10dB. The SFOMP and ASFOMP methods are still better than other methods. Figure 5 The comparison between baseband data and data recovered by different methods under SNR=10dB is shown, where the red circle indicates the location of the bit error. It can be found that the bit errors of SFOMP and ASFOMP methods are significantly less than those of other methods. In order to comprehensively demonstrate the sparsity adaptation capability of ASFOMP, Figure 6 is the difference between the sparsity estimated by ASFOMP and the actual sparsity under different signal-to-noise ratios. As the signal-to-noise ratio increases, the estimated sparsity and the actual sparsity remain consistent. The corresponding channel estimation error is as follows: Figure 7 shown.

[0204] Table 4. Error in estimated channel when SNR = 10dB:

[0205] Way LS OMP SOMP SFOMP ASFOMP Bit Error(dB) -0.19 -11.564 -11.501 -51.474 -51.474

[0206] Table 5. Bit error rate when SNR = 10dB:

[0207] Way LS OMP SOMP SFOMP ASFOMP Bit Error Rate (BER) 47.73e-3 44.95e-3 43.58e-3 42.18e-3 42.18e-3

[0208] The present invention also provides a channel estimation system based on adaptive synchronous prediction orthogonal matching tracking. The channel estimation system based on adaptive synchronous prediction orthogonal matching tracking can be realized by executing the process steps of the channel estimation method based on adaptive synchronous prediction orthogonal matching tracking, that is, those skilled in the art can understand the channel estimation method based on adaptive synchronous prediction orthogonal matching tracking as a preferred implementation of the channel estimation system based on adaptive synchronous prediction orthogonal matching tracking.

[0209] According to the present invention, a channel estimation system based on adaptive synchronous prediction orthogonal matching pursuit is provided, comprising:

[0210] Module M1: inserts a pilot signal into the transmitted signal to generate a baseband signal;

[0211] Module M2: up-modulates the baseband signal to a specific frequency band and transmits the frequency band signal;

[0212] Module M3: Receives the frequency band signal and performs synchronization operation, down-modulates the frequency to the baseband, and generates a demodulated baseband signal;

[0213] Module M4: decomposes and demodulates the baseband signal and estimates the sparse channel;

[0214] Module M5: Delay compensation of the demodulated baseband signal according to the sparse channel to restore the original signal.

[0215] In more preferred embodiments, the module M1 includes:

[0216] Module M1.1: Set the pilot signal to a complex number 3+i3 with the same elements, and determine the pilot insertion mode to be a block pilot;

[0217] Module M1.2: Input the transmit signal, sort it by column, insert pilot signals at equal intervals in the columns, and generate a complete baseband signal.

[0218] The module M2 comprises:

[0219] Module M2.1: The baseband signal is loaded onto mutually orthogonal and overlapping subcarriers through orthogonal frequency division multiplexing technology. Each element of each column of the signal corresponds to a subcarrier on the baseband.

[0220] Module M2.2: modulate each subcarrier from the baseband signal to the corresponding frequency band signal by inverse discrete Fourier transform to generate a Rayleigh channel;

[0221] Module M2.3: transmits the modulated frequency band signal through the Rayleigh channel through the transmitting transducer.

[0222] The module M3 comprises:

[0223] Module M3.1: The hydrophone receives the frequency band signal transmitted through the Rayleigh channel and performs time synchronization processing on the received frequency band signal through autocorrelation operation;

[0224] Module M3.2: Down-modulate the received frequency band signal to baseband through discrete Fourier transform operation to generate a demodulated baseband signal.

[0225] In more preferred embodiments, the module M4 includes:

[0226] Module M4.1: Construct an overcomplete dictionary D = diag(s) based on the pilot signal and the Fourier transform matrix p )F;

[0227] Among them, s p ∈C L , represents the pilot signal;

[0228] diag(·) represents the symbol of vector to diagonal matrix;

[0229] F∈C N×N , represents the discrete Fourier transform matrix;

[0230] D∈C L×N , represents an overcomplete dictionary;

[0231] N represents the length of the channel;

[0232] L represents the inverse Fourier transform length;

[0233] C represents the complex field.

[0234] Module M4.2: Sparse Decomposition Model for Signal Estimation Based on Overcomplete Dictionary

[0235] in, represents the square of the largest singular value;

[0236] ‖·‖0 represents the number of non-zero elements in the vector;

[0237] z p represents a received pilot time domain signal;

[0238] h represents the underwater acoustic channel.

[0239] Module M4.3: Use the algorithm to solve the sparse decomposition model and obtain the estimated sparse channel

[0240] In more preferred embodiments, the algorithm in the module M4.3 is the ASFOMP algorithm, including:

[0241] Module M4.3.1: Initialization parameters, R g,i =z p,i , A=[], D1=D′D, D2=D′z, S=zeros(N,M), Gr=[], k=0, Re=[];

[0242] Module M4.3.2: Let k = k + 1, calculate the atomic D j and R p,i The inner product of

[0243] Module M4.3.3: Calculate the final residuals of the first P atoms, G = G(1:P), R F =[res(D g1 ),res(D g2 ),…,res(D gP )[;

[0244] Module M4.3.4: Select the atom with the smallest final residual and record its sequence

[0245] Module M4.3.4: Store I in A, A = A∪I, calculate R p,i The projection x on A = D1(A,A)\D2(A,:), the current signal R p,i =z p,i -D(:,A)x;

[0246] Module M4.3.5: Put into Re,

[0247] Module M4.3.5: Judge the value of k. If 2 > k > 1, then let If k > 2, then Gr(k - 1) < w * mean(Gr(1:k - 2)). If 1 > k, then execute Module M4.3.2;

[0248] Module M4.3.6: Assign x to S, S(A,:) = x. Take the mean value of each row in S, S = mean(S), and output the estimated channel

[0249] where, R p,i represents the error of the i-th pilot signal in the iteration;

[0250] z p,i represents the i-th pilot signal;

[0251] A represents the set of selected atom serial numbers;

[0252] D represents the overcomplete dictionary;

[0253] D′ represents the conjugate of D;

[0254] N represents the length of the channel;

[0255] M represents the number of pilots;

[0256] zeros(·) represents the operation of creating a zero matrix;

[0257] S represents the coefficient matrix;

[0258] The subscript j represents the serial number of the atom in the overcomplete dictionary D;

[0259] R F represents the final residual of the atom;

[0260] The superscript F represents the final residual;

[0261] I represents the serial number of the atom with the minimum final residual;

[0262] Re represents the set of errors in each iteration;

[0263] Gr represents the error gradient in each iteration;

[0264] x represents the received band signal;

[0265] mean(·) represents the operation of taking the mean value;

[0266] k represents the number of loops;

[0267] P represents the pre-selection step size;

[0268] sort() represents descending sort operation;

[0269] g i represents the i-th element in G;

[0270] Indicates that the current optimal atom g i The final residual obtained by synchronous orthogonal matching pursuit with sparsity K.

[0271] In more preferred embodiments, the module M5 includes:

[0272] Module M5.1: Sparse Channels The frequency domain response is obtained by fast Fourier transform

[0273] Module M5.2: Perform channel compensation to restore the original signal The bit error rate is calculated by comparing it with the transmitted frequency band signal.

[0274] Wherein, Z represents the frequency domain of the received frequency band signal;

[0275] represents the original signal after channel compensation;

[0276] FFT() stands for Fast Fourier Transform;

[0277] IFFT() stands for Inverse Fast Fourier Transform.

[0278] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0279] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A channel estimation method based on adaptive synchronous prediction orthogonal matching pursuit, characterized in that: include: Step S1: inserting a pilot signal into a transmission signal to generate a baseband signal; Step S2: up-modulate the baseband signal to a specific frequency band and transmit the frequency band signal; Step S3: receiving the frequency band signal and performing synchronization operation, down-modulating the frequency to the baseband, and generating a demodulated baseband signal; Step S4: Decompose the demodulated baseband signal and estimate a sparse channel; Step S5: Delay compensation of the demodulated baseband signal is performed according to the sparse channel to restore the original signal.

2. The channel estimation method based on adaptive synchronous prediction orthogonal matching pursuit according to claim 1, characterized in that: The step S1 comprises: Step S1.1: Set the pilot signal to a complex number 3+i3 with the same elements, and determine that the pilot insertion mode is a block pilot; Step S1.2: input the transmission signal, sort it by column, insert pilot signals at equal intervals in the columns, and generate a complete baseband signal; The step S2 comprises: Step S2.1: The baseband signal is loaded onto mutually orthogonal and overlapping subcarriers through orthogonal frequency division multiplexing technology, and each element of each column of the signal corresponds to a subcarrier on the baseband; Step S2.2: modulate each subcarrier from a baseband signal to a corresponding frequency band signal by inverse discrete Fourier transform to generate a Rayleigh channel; Step S2.3: sending the modulated frequency band signal through the Rayleigh channel through the transmitting transducer; The step S3 comprises: Step S3.1: The hydrophone receives the frequency band signal transmitted through the Rayleigh channel, and performs time synchronization processing on the received frequency band signal through an autocorrelation operation; Step S3.2: Down-modulate the received frequency band signal to baseband through discrete Fourier transform operation to generate a demodulated baseband signal.

3. The channel estimation method based on adaptive synchronous prediction orthogonal matching pursuit according to claim 1, characterized in that: The step S4 comprises: Step S4.1: Construct an overcomplete dictionary D=diag(s p )F; Among them, s p ∈C L , represents the pilot signal; diag(·) represents the symbol of vector to diagonal matrix; F∈C N×N , represents the discrete Fourier transform matrix; D∈C L×N , represents an overcomplete dictionary; N represents the length of the channel; L represents the inverse Fourier transform length; C represents the complex domain; Step S4.2: Construct a sparse decomposition model for signal estimation based on the overcomplete dictionary in, represents the square of the largest singular value; ‖·‖0 represents the number of non-zero elements in the vector; z p represents a received pilot time domain signal; h represents the water acoustic channel; Step S4.3: Use the algorithm to solve the sparse decomposition model to obtain the estimated sparse channel 4. The channel estimation method based on adaptive synchronous prediction orthogonal matching pursuit according to claim 3, characterized in that: The algorithm in step S4.3 is the ASFOMP algorithm, including: Step S4.3.1: Initialize parameters, R p,i =z p,i , A=[],D1=D ′ D, D2 = D ′ z, S=zeros(N,M), Gr=[], k=0, Re=[]; Step S4.3.2: Let k = k + 1, calculate the atomic D j and R p,i The inner product of Step S4.3.3: Calculate the final residuals of the first P atoms respectively, G = G (1: P), Step S4.3.4: Select the atom with the smallest final residual and record its sequence Step S4.3.4: Store I in A, A = A∪I, calculate R p,i The projection x on A = D1(A,A)\D2(A,:), the current signal R p,i =z p,i -D(:,A)x; Step S4.3.5: Existence in Re, Step S4.3.5: Determine the value of k. If 2 > k > 1, then let If k > 2, then Gr(k - 1) < w * mean(Gr(1:k - 2)). If 1 > k, then execute Step S4.3.2; Step S4.3.6: Assign x to S, S(A,:) = x, take the mean S = mean(S) for each row in S, and output the estimated channel Among them, R p,i represents the error of the i-th pilot signal in the iteration; z p,i represents the i-th pilot signal; A represents the selected atomic number set; D represents an overcomplete dictionary; D ′ represents the conjugation of D; N represents the length of the channel; M represents the number of pilots; zeros(·) represents the operation of creating a zero matrix; S represents the coefficient matrix; The subscript j represents the index of the atom in the overcomplete dictionary D; R F represents the final residue of the atom; The superscript F indicates the final residual; I represents the atomic number with the smallest final residual; Re represents the error set of each iteration; Gr represents the error gradient of each iteration; x represents the received frequency band signal; mean(·) represents the mean operation; k represents the number of cycles; P represents the pre-selection step size; sort() represents descending sort operation; g i represents the i-th element in G; res(D gi ) indicates that the current optimal atom g i The final residual obtained by synchronous orthogonal matching pursuit with sparsity K.

5. The channel estimation method based on adaptive synchronous prediction orthogonal matching pursuit according to claim 1, characterized in that: The step S5 comprises: Step S5.1: Sparse channel The frequency domain response is obtained by fast Fourier transform Step S5.2: Perform channel compensation to restore the original signal Calculate the bit error rate by comparing with the transmitted frequency band signal; Wherein, Z represents the frequency domain of the received frequency band signal; represents the original signal after channel compensation; FFT() stands for Fast Fourier Transform; IFFT() stands for Inverse Fast Fourier Transform.

6. A channel estimation system based on adaptive synchronous prediction orthogonal matching pursuit, characterized in that: include: Module M1: inserts a pilot signal into the transmitted signal to generate a baseband signal; Module M2: up-modulates the baseband signal to a specific frequency band and transmits the frequency band signal; Module M3: Receives the frequency band signal and performs synchronization operation, down-modulates the frequency to the baseband, and generates a demodulated baseband signal; Module M4: decomposes and demodulates the baseband signal and estimates the sparse channel; Module M5: Delay compensation of the demodulated baseband signal according to the sparse channel to restore the original signal.

7. The channel estimation system based on adaptive synchronous prediction orthogonal matching pursuit according to claim 6, characterized in that: The module M1 comprises: Module M1.1: Set the pilot signal to a complex number 3+i3 with the same elements, and determine the pilot insertion mode to be a block pilot; Module M1.2: Input the transmit signal, sort it by column, insert pilots at equal intervals in the columns, and generate a complete baseband signal; The module M2 comprises: Module M2.1: The baseband signal is loaded onto mutually orthogonal and overlapping subcarriers through orthogonal frequency division multiplexing technology. Each element of each column of the signal corresponds to a subcarrier on the baseband. Module M2.2: modulate each subcarrier from the baseband signal to the corresponding frequency band signal by inverse discrete Fourier transform to generate a Rayleigh channel; Module M2.3: transmits the modulated frequency band signal through the Rayleigh channel through the transmitting transducer; The module M3 comprises: Module M3.1: The hydrophone receives the frequency band signal transmitted through the Rayleigh channel and performs time synchronization processing on the received frequency band signal through autocorrelation operation; Module M3.2: Down-modulate the received frequency band signal to baseband through discrete Fourier transform operation to generate a demodulated baseband signal.

8. The channel estimation system based on adaptive synchronous prediction orthogonal matching pursuit according to claim 6, characterized in that: The module M4 comprises: Module M4.1: Construct an overcomplete dictionary D = diag(s based on the pilot signal and the Fourier transform matrix p )F; Among them, s p ∈C L , represents the pilot signal; diag(·) represents the symbol of vector to diagonal matrix; F∈C N×N , represents the discrete Fourier transform matrix; D∈C L×N , represents an overcomplete dictionary; N represents the length of the channel; L represents the inverse Fourier transform length; C represents the complex domain; Module M4.2: Sparse Decomposition Model for Signal Estimation Based on Overcomplete Dictionary in, represents the square of the largest singular value; ‖·‖0 represents the number of non-zero elements in the vector; z p represents a received pilot time domain signal; h represents the water acoustic channel; Module M4.3: Use the algorithm to solve the sparse decomposition model and obtain the estimated sparse channel 9. The channel estimation system based on adaptive synchronous prediction orthogonal matching pursuit according to claim 8, characterized in that: The algorithm in the module M4.3 is the ASFOMP algorithm, including: Module M4.3.1: Initialization parameters, R g,i =z p,i , A=[],D1=D ′ D, D2 = D ′ z, S=zeros(N,M), Gr=[], k=0, Re=[]; Module M4.3.2: Let k = k + 1, calculate the atomic D i and R p,i The inner product of Module M4.3.3: Calculate the final residuals of the first P atoms respectively, G = G (1: P), Module M4.3.4: Select the atom with the smallest final residual and record its sequence Module M4.3.4: Store I in A, A = A∪I, calculate R p,i The projection x on A = D1(A,A)\D2(A,:), the current signal R p,i =z p,i -D(:,A)x; Module M4.3.5: Existence in Re, Module M4.3.5: Determine the value of k. If 2 > k > 1, then let If k > 2, then Gr(k - 1) < w * mean(Gr(1:k - 2)). If 1 > k, then execute Module M4.3.2; Module M4.3.6: Assign x to S, S(A,:) = x, take the mean of each row in S S = mean(S), and output the estimated channel Among them, R p,i represents the error of the i-th pilot signal in the iteration; z p,i represents the i-th pilot signal; A represents the selected atomic number set; D represents an overcomplete dictionary; D ′ represents the conjugation of D; N represents the length of the channel; M represents the number of pilots; zeros(·) represents the operation of creating a zero matrix; S represents the coefficient matrix; The subscript j represents the index of the atom in the overcomplete dictionary D; R F represents the final residue of the atom; The superscript F indicates the final residual; I represents the atomic number with the smallest final residual; Re represents the error set of each iteration; Gr represents the error gradient of each iteration; x represents the received frequency band signal; mean(·) represents the mean operation; k represents the number of cycles; P represents the pre-selection step size; sort() represents descending sort operation; g i represents the i-th element in G; res(D gi ) indicates that the current optimal atom g i The final residual obtained by synchronous orthogonal matching pursuit with sparsity K.

10. The channel estimation system based on adaptive synchronous prediction orthogonal matching pursuit according to claim 6, characterized in that: The module M5 comprises: Module M5.1: Sparse Channels The frequency domain response is obtained by fast Fourier transform Module M5.2: Perform channel compensation to restore the original signal Calculate the bit error rate by comparing with the transmitted frequency band signal; Wherein, Z represents the frequency domain of the received frequency band signal; represents the original signal after channel compensation; FFT() stands for Fast Fourier Transform; IFFT() stands for Inverse Fast Fourier Transform.

Images