An orthogonal matching pursuit channel estimation method based on correlation peak search
By using relevant peak search in water acoustic communication, the problem of high computational complexity of orthogonal matching tracking algorithm is solved, efficient channel estimation is achieved, CPU computing time is reduced and estimation accuracy is maintained.
Patent Information
- Application Number
- CN202310421156.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-04-19
- Publication Date
- 2024-12-13
- Estimated Expiration
- 2043-04-19
AI Technical Summary
In the water acoustic communication environment, the calculation complexity of the orthogonal matching tracking algorithm is high, which is difficult to effectively reduce, affecting the efficiency and accuracy of channel estimation.
Using the method based on correlation peak search, the dictionary set is simplified by using the autocorrelation characteristics of linear frequency modulation signals. By searching and filtering the correlation peaks whose correlation peaks are greater than the threshold and their time delays, the number of calculations in the matching process is reduced, and the sparseness is estimated as the iterative round. The prior information of the time reduces the computational complexity.
While ensuring channel estimation accuracy, the calculation complexity of the orthogonal matching tracking algorithm is significantly reduced, the efficiency of channel estimation is improved, and the CPU computing time is reduced.
Smart Images
Figure CN116405349B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of communication technology and relates to an orthogonal matching tracking channel estimation method based on correlation peak search. Specifically, in an underwater acoustic communication environment, a linear frequency modulation signal is used as a preamble code for a transmitted signal, and a receiving end uses the good autocorrelation characteristics of the linear frequency modulation signal to estimate sparsity and simplify a dictionary set, thereby improving an orthogonal matching tracking algorithm to complete channel estimation and reduce the computational complexity of the orthogonal matching tracking algorithm. Background Art
[0002] Marine resources have extremely broad development prospects and application value. Countries around the world are constantly strengthening research on marine-related technologies and transforming the scientific research value of the ocean into their own comprehensive national strength. Underwater communication is one of the key technologies to promote the development of marine resources. Compared with light and electromagnetic waves, sound waves have lower propagation loss underwater and can be transmitted over longer distances, which can achieve communication transmission within a range of hundreds of meters or even thousands of meters. Therefore, it is used as an important carrier for underwater data transmission, and then underwater acoustic communication technology is born. However, the multipath effect of the underwater acoustic channel will cause amplitude fading, frequency expansion and inter-code interference of the received signal. Eliminating inter-code interference is a major problem faced in establishing a reliable high-speed underwater acoustic communication system. At present, one of the main means to solve this problem is equalization technology, and equalization requires sensing channel characteristics through channel estimation. Therefore, accurate channel estimation can provide reliable channel characteristics for the equalizer to better eliminate multipath delay.
[0003] Compressed sensing (CS) technology is a new theory developed on the basis of many mathematical theories such as matrix theory, probability statistics theory and optimization theory. CS theory can realize signal sampling and processing through extremely low sampling rate, while reducing the cost of data storage and transmission and reducing signal processing time and computing cost. Since it was proposed in 2006, CS theory has developed rapidly in many fields. Shallow sea water acoustic channel is usually regarded as a slowly time-varying coherent multipath channel. Under the premise of limited observation or processing time, the channel can be described by a time-invariant filter with sparse tap distribution. Therefore, compressed sensing technology can be used to reconstruct the channel impulse response. The orthogonal matching pursuit algorithm finds the support set of sparse vectors in an iterative way, and uses the least squares estimation on the support set to reconstruct the signal. However, the orthogonal matching pursuit algorithm requires the prior information of sparsity, and determines the number of matches between the received signal and the dictionary set according to the sparsity, which has large restrictions and high computational complexity.
[0004] The invention patent with application number 201110194559.9 discloses a coefficient channel estimation method based on orthogonal matching tracking, which includes the following steps: the input vector signal is a channel impulse response sequence, a measurement matrix is constructed using a known input training sequence, and a sparse channel impulse response model is constructed to restore and reconstruct the input vector signal by receiving the vector signal; the orthogonal matching pursuit algorithm is used to estimate the position and amplitude of each path to obtain the channel estimation value. This method only uses the classic OMP algorithm to perform channel estimation, and does not significantly reduce the computational complexity. The invention patent with application number 201610318947.6 discloses a channel estimation method based on the sparse characteristics of the underwater acoustic block structure. This method further considers the block distribution structure inside the underwater acoustic sparse signal on the basis of the sparse model, that is, the non-zero elements (or zero elements) of the sparse signal are not randomly distributed but presented in the form of blocks. Dividing the sparse signal into blocks can effectively reduce the signal sparsity, thereby reducing the number of pilots required for estimation, reducing the computational complexity, but cannot guarantee the estimation accuracy. Summary of the invention
[0005] The purpose of the present invention is to provide an orthogonal matching tracking channel estimation method based on correlation peak search for underwater acoustic communication scenarios, which reduces the computational complexity of the orthogonal matching tracking algorithm while ensuring the estimation accuracy.
[0006] The underwater acoustic channel is approximately represented by a linear time-invariant finite impulse response filter, and the impulse response is Where M is the number of multipaths, A m is the normalized amplitude of the mth multipath, τ m is the relative delay of the mth multipath relative to the direct path, m = 1, 2, ..., M; the real number form of the transmitted linear frequency modulation signal is s(t) = cos(2πf0t+πkt 2 ), FM slope T is the duration, signal bandwidth B = f H -f L , f H is the stop frequency, f L is the starting frequency; the corresponding discrete sequence is s(n), and its length is f s is the sampling frequency, the received signal y(t) is expressed as: Where ω(t) is zero-mean Gaussian white noise.
[0007] The estimation model after sampling and quantization is expressed as: y(n) = Ψh(n) + ω(n); h(n) is the channel impulse response sequence, y(n) is the received signal vector; ω(n) is the noise vector, Ψ is the dictionary matrix composed of s(n) and its delayed vector, h(n) = [h(0) h(1) … h(L-1)] T, L is the effective length of the channel, L<N, and the superscript T indicates transposition.
[0008] Ray theory approximates the propagation of sound waves as rays traveling along a specific spatial path from the sound source to the receiving point; the beam adds an intensity distribution perpendicular to the ray trajectory (such as a Gaussian distribution) to the ray, which can more accurately calculate the sound intensity at a given point in space; it is assumed that the speed of sound and the refractive index do not change in the horizontal direction and are only related to the depth of the sea water; it is assumed that the positions of the sound source and the receiving point are stationary; by setting the ocean environment parameters, the beam tracking model can be used to obtain the true impulse response of the underwater acoustic channel.
[0009] The specific estimation method of the present invention is:
[0010] Step (1) parameter initialization;
[0011] Initialize the residual vector r = y(n), sub-dictionary set Index Set Reconstructing atomic sets Set the correlation peak search threshold η and the estimation error tolerance σ, and initialize the number of iterations count = 1.
[0012] Step (2) Preliminary estimation of sparsity and multipath delay range, and simplification of sub-dictionary set;
[0013] The received signal vector y(n) is matched with the locally generated linear frequency modulation signal s(n) to obtain: in, is the number of sampling points of the delay corresponding to the m-th multipath, and ε(n) is the noise after matched filtering.
[0014] Search and filter the correlation peaks whose normalized amplitudes are greater than the threshold η. The delays corresponding to these correlation peaks include both the multipath relative delays relative to the direct path and some irrelevant path delays, which can reduce the number of columns in the dictionary matrix. Record the number of selected correlation peak positions L′ and the delay sampling position value c corresponding to each correlation peak. l , l=1,2,…,L′,L′<<L, after simplifying the dictionary set, the sub-dictionary set is obtained as follows:
[0015] Among them, i Represents the column vector i in the dictionary matrix Ψ.
[0016] Compared with the dictionary matrix Ψ, as the column index increases, the multipath delay also increases; the simplified sub-dictionary set eliminates some irrelevant paths, and the delay corresponding to each column vector no longer changes continuously, but will jump at the delay corresponding to the transmission path, that is, c1, c2, …, c L′As the index l increases, it changes in a step-like trend, and the number of statistical jumps is the sparsity κ.
[0017] Step (3) Calculate the correlation between the residual vector and the column vector in the sub-dictionary set, and select the atom with the largest correlation and its corresponding column index Among them, Γ i is the column vector i in the sub-dictionary set Γ, and the |<·>| operator takes the absolute value after calculating the inner product.
[0018] Step (4) updates the elements in the sub-dictionary set, index set and reconstruction atomic set, namely: is the column index λ count The set of sub-dictionaries of .
[0019] Step (5) estimates the channel impulse response based on the minimum mean square error criterion. The estimated value of the channel impulse response sequence is Using the least squares method to solve, the solution of the signal impulse response is: The superscript H indicates the conjugate transpose.
[0020] Step (6) calculates the approximate vector of y(n) as Update the residual vector:
[0021] If the residual is not lower than the estimated error tolerance, The operator |||| 2 represents the square of the vector modulus, or the number of iterations has not reached the maximum value, that is, count < κ, then count is increased by 1 and steps (3) to (6) are executed, otherwise step (7) is executed.
[0022] Step (7) outputs the final estimation result of the channel impulse response
[0023] The present invention takes into account the multipath effect of underwater acoustic communication channels. Aiming at the problem that orthogonal matching tracking needs sparsity as a priori information for channel estimation and the high computational complexity in the matching process, an orthogonal matching tracking channel estimation method based on correlation peak search is proposed. Before matching with a dictionary set, the linear frequency modulation signal has good autocorrelation characteristics, and the correlation peaks with correlation peaks greater than a threshold and their corresponding time delays are searched and screened to simplify the dictionary set. The sparsity can be estimated based on the time delay corresponding to the screened correlation peak as the priori information of the orthogonal matching tracking iteration round, and the number of calculations in the matching process is reduced, thereby reducing the computational complexity. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] Figure 1 is a flow chart of the method of the present invention;
[0025] Figure 2Schematic diagram of the frame structure sent by the source;
[0026] Figure 3 Schematic diagram of multipath delay screened after correlation peak search;
[0027] Figure 4 A schematic diagram of estimation results of an orthogonal matching tracking channel estimation method based on correlation peak search;
[0028] Figure 5 Schematic diagram of the CPU computing time required for channel estimation using orthogonal matching pursuit and orthogonal matching pursuit based on correlation peak search under different channel conditions
[0029] Figure 6 Schematic diagram of the normalized mean square error of channel estimation using least squares, orthogonal matching pursuit, and orthogonal matching pursuit based on correlation peak search under different signal-to-noise ratios. DETAILED DESCRIPTION
[0030] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.
[0031] The present invention proposes an orthogonal matching tracking channel estimation method based on correlation peak search, and its flow chart is as follows: Figure 1 As shown. With sampling frequency f s =8KHz as an example, the linear frequency modulation signal sent is s(t)=cos(2πf0t+πkt 2 ), the starting frequency is f L =23KHz, the end frequency is f H =27KHz, duration T=60ms; the corresponding digital sequence s(n) length N=480. This is achieved by the following steps:
[0032] According to the beam tracking model, the channel impulse response of the underwater acoustic channel in a 10-path multipath environment is: A m is the normalized amplitude of the mth multipath, τ m is the relative delay of the mth multipath relative to the direct path, m = 1, 2, …, 10, the effective channel length is L = 261, and the corresponding channel impulse response sequence is h(n) = [h(0) h(1) … h(260)] T The frame structure sent in actual application is as follows Figure 2 As shown. Use LFM2 close to the data frame to estimate the channel parameters. Ψ is the dictionary matrix composed of s(n) and its delayed vector:
[0033]
[0034] Step (1) parameter initialization;
[0035] Initialize the residual vector r = y(n), sub-dictionary set Index Set Reconstructing atomic sets Set the correlation peak search threshold to η = 0.1 and the estimation error tolerance to σ = 10 -3 , initialize the number of iterations count = 1.
[0036] Step (2) preliminarily estimates the sparsity and multipath delay range and simplifies the dictionary set:
[0037] The received signal is matched filtered with the linear frequency modulation signal, and the correlation peaks with normalized amplitude greater than η are selected. The delays corresponding to these correlation peaks include the multipath relative delay relative to the direct path, and exclude some irrelevant path delays, which can reduce the number of columns of the dictionary matrix. The number of selected correlation peak positions is recorded as 24 and the delay sampling position value c corresponding to each correlation peak l , l=1,2,…,24, then the sub-dictionary set obtained after simplifying the dictionary set is:
[0038] It can be observed that compared with the dictionary set Ψ, the number of matrix columns of the sub-dictionary set Γ is greatly reduced, which reduces the number of matches.
[0039] c1,c2,…,c L′ As the index l increases, it shows a step-like change trend. The number of statistical jumps is 10, and the update sparsity κ = 10. The multipath delay sampling position value after screening is as follows: Figure 3 As shown; the vertical axis is the time delay sampling position value corresponding to each correlation peak, and the horizontal axis is the sparsity. It can be seen that the cumulative jump is 10 times.
[0040] Step (3) calculates the correlation between the residual vector and the column vector in the sub-dictionary set, and selects the atom with the largest correlation and its corresponding column index; In the formula, Γ i is the column vector i in the sub-dictionary.
[0041] Step (4) updates the elements in the sub-dictionary set, index set and reconstruction atomic set, namely:
[0042] Step (5) estimates the channel impulse response based on the minimum mean square error criterion: Using the least squares method to solve, the solution of the signal impulse response is:
[0043] Step (6) calculates the approximate vector of y(n) as Update the residual vector:
[0044] If the residual is not lower than the estimated error tolerance, Or if the number of iterations does not reach the maximum value, that is, count < κ, then count is increased by 1 and steps (3) to (6) are executed, otherwise step (7) is executed;
[0045] After the iteration, the multipath delay set obtained is [0 5 2 142 24 30 57 98 199 260], and the corresponding support set Φ is:
[0046]
[0047] Based on the least mean square criterion, the estimated channel impulse response sequence can be obtained;
[0048]
[0049] After 10 iterations, step (7) outputs the final estimated channel impulse response:
[0050]
[0051] The channel impulse response sequence estimated by the improved orthogonal matching pursuit algorithm based on linear frequency modulation signal is as follows: Figure 4 shown.
[0052] Comparison of the computation time of orthogonal matching pursuit generated by BELLHOP and the method presented here under different channel parameters Figure 5 As shown in FIG. 1 , the channel parameters are three, seven, and ten multipaths, respectively. It can be seen that the orthogonal matching tracking channel estimation algorithm based on linear frequency modulation signal reduces the computational complexity.
[0053] The normalized estimation errors of the three estimation algorithms, least squares, orthogonal matching tracking, and orthogonal matching tracking based on linear frequency modulation signals under ten multipaths are shown in Figure 2. Figure 6 As shown in Figure 2, it can be seen that the estimation error of the orthogonal matching pursuit channel estimation algorithm based on linear frequency modulation signal is basically consistent with that of the orthogonal matching pursuit algorithm, and the estimation accuracy has not been reduced.
[0054] The above description of the embodiments is merely an enumeration of implementation forms of the present invention. The protection scope of the present invention should not be limited to the specific forms described in the embodiments. Improvements and modifications made to the present invention by those skilled in the art based on the disclosure of the present invention should be within the protection scope of the present invention.
Claims
1. An orthogonal matching tracking channel estimation method based on correlation peak search. The underwater acoustic channel is approximately represented by a linear time-invariant finite impulse response filter, and the impulse response is in, M is the number of multipaths, A m is the normalized amplitude of the mth multipath, τ m is the relative delay of the mth multipath relative to the direct path, m = 1, 2, ..., M; the real number form of the transmitted linear frequency modulation signal is s(t) = cos(2πf0t+πkt 2 ), FM slope T is the duration, signal bandwidth B = f H -f L , f H is the stop frequency, f L is the starting frequency; the corresponding discrete sequence is s(n), and its length is f s is the sampling frequency, the received signal y(t) is expressed as: Where, ω(t) is zero-mean Gaussian white noise; The estimation model after sampling and quantization is expressed as: y(n) = Ψh(n) + ω(n); h(n) is the channel impulse response sequence, y(n) is the received signal vector; ω(n) is the noise vector, Ψ is the dictionary matrix composed of s(n) and its delayed vector, h(n) = [h(0)h(1)…h(L-1)] T , L is the effective length of the channel, L<N, and the superscript T indicates transposition. It is characterized in that a linear frequency modulation signal is sent to perform channel estimation, and the specific estimation method is as follows: Step (1) parameter initialization; Initialize the residual vector r = y(n), sub-dictionary set Index Set Reconstructing atomic sets Set the correlation peak search threshold η and the estimation error tolerance σ, and initialize the number of iterations count = 1; Step (2) performs matched filtering on the received signal vector y(n) and the locally generated linear frequency modulation signal s(n), and obtains: in, is the number of sampling points of the delay corresponding to the mth multipath, ε(n) is the noise after matched filtering; Search and filter the correlation peaks whose normalized amplitudes are greater than the threshold η, and record the number of filtered correlation peak positions L′ and the time delay sampling position value c corresponding to each correlation peak l , l=1,2,…,L′,L′<<L, after simplifying the dictionary set, the sub-dictionary set is obtained as follows: Among them, i represents the column vector i in the dictionary matrix Ψ; c1,c2,…,c L′ As the index l increases, it changes in a step-like manner, and the number of statistical jumps is the sparsity κ; Step (3) calculates the correlation between the residual vector and the column vector in the sub-dictionary set, and selects the atom with the largest correlation and its corresponding column index λ count ; Step (4) updating the elements in the sub-dictionary set, the index set and the reconstruction atomic set; Step (5) estimates the channel impulse response based on the minimum mean square error criterion. The estimated value of the channel impulse response sequence is Using the least squares method to solve, the solution of the signal impulse response is: The superscript H indicates conjugate transpose; Step (6) calculates the approximate vector of y(n) as Update the residual vector: If the residual is not lower than the estimated error tolerance, The operator || || 2 represents the square of the vector modulus, or the number of iterations does not reach the maximum value, that is, count < κ, then count is increased by 1, and steps (3) to (6) are executed, otherwise step (7) is executed; Step (7) outputs the final estimation result of the channel impulse response 2. The orthogonal matching tracking channel estimation method based on correlation peak search according to claim 1, characterized in that: In step (3), Among them, Γ i is the column vector i in the sub-dictionary set Γ, and the |<·>| operator takes the absolute value after calculating the inner product.
3. The orthogonal matching tracking channel estimation method based on correlation peak search as claimed in claim 2, characterized in that: The update method of step (4) is: Γ λcount is the column index λ count The set of sub-dictionaries of .