A radar ranging method based on fastica algorithm

Abstract

The application belongs to the technical field of signal processing, and particularly relates to a radar ranging method based on a FastICA algorithm; the method comprises the following steps: collecting observation signals of multiple radars, the multiple observation signals being generated after multiple radar source signals pass through an unknown linear mixing system; pre-processing multiple observation signal data, including centering and whitening processing, to obtain whitened data; inputting the whitened data into an improved FastICA algorithm module for processing to obtain separated signals; and inputting the separated signals into an LFM radar signal pulse compression system for processing to obtain distance information; the application effectively reduces consumption of hardware resources, improves the calculation speed of the algorithm in parallel, is conducive to reducing the size of a chip circuit, and reduces manufacturing costs.

Description

Technical Field

[0001] This invention belongs to the field of signal processing technology, specifically relating to a radar ranging method based on the FastICA algorithm. Background Technology

[0002] In communication signal processing, due to complex and dense electromagnetic interference, signals are affected by noise and strong signals in the environment, making signal sorting extremely difficult. Weak signals are hard to detect, thus hindering the accurate extraction of the desired target signal. Independent component analysis (ICA) in blind source separation methods is a method based on finding independent factors or components in statistical data. The FastICA algorithm is one of the more efficient blind source separation algorithms. When both variables and the mixed system are unknown, it decomposes the observed signal to obtain several statistically independent variables implicit within it. The traditional FastICA algorithm is an algorithm based on a reasonable approximation using the second-order Newton iteration method. To improve the stability of the FastICA algorithm and reduce the number of iterations, many scholars have proposed improved FastICA algorithms in recent years. Among them, the improved FastICA algorithm based on the third-order Newton iteration method has a faster convergence speed than the traditional algorithm.

[0003] Current FPGA implementations of the iterative formulas of the existing FastICA algorithm suffer from the following problems: 1. The algorithm structure is complex, involving a large amount of matrix processing, resulting in a huge computational load. Furthermore, the calculation results need to be written to and read from RAM multiple times, consuming significant time. 2. The implementation of polynomials in matrix operations in the algorithm mostly adopts a part-matrix approach, requiring subsequent calculations to wait for the completion of previous calculations, causing substantial waiting time and further hindering real-time performance. 3. The iterative polynomial structure required by the existing improved FastICA algorithm is even more complex, involving larger amounts of data and more complex calculations. Implementing the formula calculations in hardware circuits consumes substantial resources. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes a radar ranging method based on the FastICA algorithm, which includes:

[0005] S1: Acquire observation signal data from multiple radar signals;

[0006] S2: Preprocess the multi-channel observation signals to obtain whitened data; the preprocessing includes centering and whitening.

[0007] S3: Input the whitened data into the improved FastICA algorithm module for processing to obtain the separated signal;

[0008] S4: Input the separated signal into the LFM radar signal pulse compression system for processing to obtain range information.

[0009] Preferably, the improved FastICA algorithm module's process for processing whitened data includes:

[0010] S31: Perform iterative calculations on the whitened data according to the traditional iterative formula to obtain the intermediate variable matrix and iterative intermediate parameters;

[0011] S32: Iteratively calculate the intermediate variable matrix and intermediate parameters according to the improved iterative formula to obtain the initial unmixing matrix;

[0012] S33: Orthogonalize and normalize the initial unmixing matrix to obtain the updated unmixing matrix;

[0013] S34: Determine whether the unmixing matrix is ​​converged. If it is converged, output the unmixing matrix; otherwise, return to step S31.

[0014] S35: Multiply the demixing matrix with the whitened data to obtain the separated signal.

[0015] Furthermore, the process of iteratively calculating the whitened data according to the traditional iterative formula includes:

[0016] The column vectors of the unmixing matrix are multiplied by each group of data in the whitened data to obtain multiple initial separated signal data; the unmixing matrix is ​​randomly generated during the first calculation using the traditional iterative formula.

[0017] The initial separated signal data, unmixed matrix, and whitened data are substituted into the traditional iterative formula for calculation to obtain column vectors of multiple initial intermediate variable matrices; multiple initial iterative intermediate parameters are calculated based on the multiple initial separated signal data.

[0018] The updated intermediate variable matrix and iterative intermediate parameters are obtained by averaging the column vectors of multiple initial intermediate variable matrices and the multiple initial iterative intermediate parameters.

[0019] Furthermore, the traditional iterative formula is as follows:

[0020]

[0021] in, Z represents the column vector of the i-th intermediate variable matrix. i Let w represent the whitening data of the i-th group. i Let y represent the column vector of the i-th unmixing matrix. i This represents the i-th initial separated signal data.

[0022] Furthermore, the formula for calculating the initial iterative intermediate parameters based on the initial separation signal data is as follows:

[0023] β=yi 4

[0024] Where β represents the initial intermediate parameter of the iteration, y i This represents the i-th initial separated signal data.

[0025] Furthermore, the process of iteratively calculating the intermediate variable matrix and intermediate parameters according to the improved iterative formula includes:

[0026] The column vectors of multiple intermediate variable matrices and the whitened data are multiplied together to obtain multiple secondary separated signal data;

[0027] Substituting the secondary separated signal data, intermediate variable matrix, and whitening data into the improved iterative formula for calculation, we obtain column vectors of multiple initial unmixed matrices;

[0028] The average value is calculated for the corresponding elements of the column vectors of multiple initial unmixing matrices, and the calculated column vectors are combined to obtain the updated unmixing matrix.

[0029] Furthermore, the improved iterative formula is as follows:

[0030]

[0031] Among them, w i This represents the column vector of the i-th initial unmixing matrix. Z represents the column vector of the i-th intermediate variable matrix. i Let y represent the i-th whitened data set, β represent the initial iteration intermediate parameters, and y represent the y-th data set. i ′ represents the i-th secondary separated signal data.

[0032] Furthermore, the process of determining whether the unmixed matrix converges includes: setting a threshold for the difference between the unmixed matrix and a maximum number of iterations; when the absolute value of the difference between the current unmixed matrix and the unmixed matrix obtained in the previous iteration is less than the threshold for the difference between the unmixed matrix or the number of iterations reaches the maximum number of iterations, the unmixed matrix is ​​determined to converge; otherwise, the unmixed matrix is ​​determined not to converge.

[0033] The beneficial effects of this invention are as follows: This invention can meet the needs of general mixed signal separation applications. Experiments have verified that this invention can achieve ideal signal separation results, effectively separating the source signal from mixed signals with strong noise interference. The parallelized data processing structure doubles the data processing rate, and the reuse of functional modules effectively reduces resource consumption, which is beneficial for reducing chip circuit size and manufacturing costs. This invention has significant application potential in the development and production of portable chip devices for radio communication systems, audio and acoustics, and medical signal processing in industrial and military fields. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the radar ranging method based on the FastICA algorithm in this invention;

[0035] Figure 2 This is a schematic diagram of the functional structure of the improved FastICA algorithm module in this invention;

[0036] Figure 3 This is an iterative flowchart of the iterative formula calculation module in the improved FastICA algorithm of this invention;

[0037] Figure 4 This is the state transition diagram of the improved FastICA algorithm module in this invention. Detailed Implementation

[0038] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0039] Because the improved FastICA algorithm involves matrix operations of two polynomials, traditional implementations calculate the result of each matrix operation separately before proceeding with subsequent calculations. However, this increases the algorithm's computation time and consumes excessive resources to store the results, making hardware implementation more difficult. While the improved FastICA algorithm enhances stability and reduces the number of iterations, it uses the results of the original iterative polynomials as intermediate variables in the iterative process, substituting them into the new iterative formula. This significantly increases the number of computational steps and the computational load, further increasing computation time. Consequently, the algorithm's hardware circuitry loses its original timeliness, and the added iterative formula requires even more resources.

[0040] To mitigate the problems arising from the optimization of the iterative formula in the improved FastICA algorithm, this invention proposes a radar ranging method based on the FastICA algorithm, such as... Figure 1 As shown, the method includes the following:

[0041] S1: Collect observation signal data from multiple radar signals.

[0042] The observation signal data is obtained from multiple radar signals collected from the environmental system. That is, multiple radar signals are processed by an unknown linear hybrid system to generate multiple observation signal data.

[0043] S2: Preprocess the multi-channel observation signals to obtain whitened data. The preprocessing includes centering and whitening.

[0044] A hybrid signal separation module was constructed for a radar communication system. This module's function is to separate and reconstruct the multi-source signals from multiple observation signals obtained through an unknown linear mixture in a real-world system. The hybrid signal separation module is an FPGA logic circuit module. Within this module, FPGA logic circuits are used to implement preprocessing and signal separation of the multi-source observation signals. The FPGA logic circuit module includes a preprocessing module and an improved FastICA algorithm module.

[0045] The preprocessing module performs preprocessing on the multi-channel observation signals, including centering and whitening, to obtain whitened data and store it in the FPGA's RAM.

[0046] S3: Input the whitened data into the improved FastICA algorithm module for processing to obtain the separated signal.

[0047] The improved FastICA algorithm module includes an iterative operation module, an orthogonalization and normalization module, and a convergence judgment module.

[0048] like Figure 2 , Figure 3 As shown, the process of processing whitened data by the improved FastICA algorithm module includes:

[0049] S31: Perform iterative calculations on the whitened data according to the traditional iterative formula to obtain the intermediate variable matrix and iterative intermediate parameters.

[0050] The whitening data is divided into n groups. Each group of whitening data is read out and sequentially fed into the improved FastICA algorithm module for processing. In the iterative calculation module, the whitening data is first calculated in parallel using the traditional iterative formula. Specifically:

[0051] The calculation identifier is used to determine whether the iterative operation module is calculating the intermediate variable matrix and iterative intermediate parameters, or calculating the unmixing matrix. The corresponding formula is selected for calculation based on the calculation identifier, and the calculation identifier is set to low bit.

[0052] In the first calculation using the traditional iterative formula, the unmixing matrix is ​​randomly generated. The initial unmixing matrix is ​​denoted as W. This unmixing matrix W is then split into column vectors, w... i (i = 1, 2, ..., n) are the column vectors of the unmixing matrix W.

[0053] Multiply the column vector of the demixing matrix by each group of data in the whitened data to obtain multiple initial separated signal data:

[0054]

[0055] Among them, y i Z represents the i-th initial separated signal data. i This represents the i-th group of whitened data.

[0056] Substituting the initial separated signal data, unmixed matrix, and whitened data into the traditional iterative formula, we obtain column vectors of multiple initial intermediate variable matrices. Based on these initial separated signal data, we calculate multiple initial iterative intermediate parameters. The calculation formula is as follows:

[0057]

[0058] β=y i 4

[0059] in, Z represents the column vector of the i-th intermediate variable matrix. i Let w represent the whitening data of the i-th group. i Let y represent the column vector of the i-th unmixing matrix. i Let represent the i-th initial separated signal data, and β represent the initial iteration intermediate parameters.

[0060] The updated intermediate variable matrix and iterative intermediate parameters are obtained by averaging the column vectors of multiple initial intermediate variable matrices and the multiple initial iterative intermediate parameters.

[0061] S32: Iteratively calculate the intermediate variable matrix and intermediate parameters according to the improved iterative formula to obtain the initial unmixing matrix.

[0062] Set the computation identifier to the high bit. Multiply the column vectors of multiple intermediate variable matrices and the whitened data respectively to obtain multiple secondary separated signal data.

[0063]

[0064] Among them, y i ′ represents the i-th secondary separated signal data.

[0065] Substituting the secondary separated signal data, intermediate variable matrix, and whitening data into the improved iterative formula, we obtain column vectors of multiple initial unmixed matrices. The calculation formula is as follows:

[0066]

[0067] The average value of the corresponding elements of the column vectors of the multiple initial unmixed matrices is calculated to obtain the updated unmixed matrix.

[0068] S33: Orthogonalize and normalize the initial unmixing matrix to obtain the updated unmixing matrix.

[0069] The orthogonalization and normalization module orthogonalizes and normalizes the column vectors of all initial unmixed matrices, combining them to obtain the updated unmixed matrix W = {w1, w2, ..., w n}

[0070] S34: Determine whether the unmixing matrix is ​​converged. If it is converged, output the unmixing matrix; otherwise, return to step S31.

[0071] The convergence judgment module determines the unmixed matrix W = {w1, w2, ..., w} after the iterative operation. n If convergence is not achieved, W is re-entered into the iterative formula calculation module for the above iterative calculation; if convergence is achieved, the unmixed matrix W is output, completing the entire iterative process.

[0072] The process of determining whether the unmixed matrix is ​​convergent includes: setting a threshold for the difference between the unmixed matrix and a maximum number of iterations. When the absolute value of the difference between the current unmixed matrix and the unmixed matrix obtained in the previous iteration is less than the threshold for the difference between the unmixed matrix and the number of iterations has not reached the maximum number of iterations, the unmixed matrix is ​​determined to be convergent; otherwise, the unmixed matrix is ​​determined to be non-convergent.

[0073] S35: Multiply the demixing matrix by the whitened data to obtain the separated signal.

[0074] Multiplying the unmixing matrix by the whitening data matrix yields the separated signal Y = {y1, y2, ..., y}. n}

[0075] The state transition diagram of the state machine in the improved FastICA algorithm module of this invention is as follows: Figure 4 As shown in the figure, S1 is the waiting state, waiting for the iteration request signal to arrive; S2 is the processing state of Iteration Formula 1 (traditional iteration formula), and the calculation identifier is set to low bit; S3 is the processing state of Iteration Formula 2 (improved iteration formula), and the calculation identifier is set to high bit; S4 is the orthogonalization and normalization processing state; S5 is the convergence judgment state; S6 is the state transition, when the absolute value ε of the difference between the two unmixed matrices before and after iteration is greater than or equal to the highest difference ε. max That is, when the difference threshold of the unmixed matrices is met, and the number of iterations k is less than or equal to the maximum number of iterations M, it indicates that the iteration is not complete, and the state jumps to state S2; when the condition is met, the absolute value ε of the difference between the two unmixed matrices is less than the highest difference ε. max If the iteration number k is greater than the iteration number M, it indicates that the iteration has been completed, the state jumps to state S1, the result is output and the process ends.

[0076] The improved FastICA algorithm in this invention does not require waiting for a single polynomial operation. Instead, it uses multiple polynomials to process data simultaneously. The unmixed matrix is ​​split into multiple column vectors, and these column vectors are substituted into the formula for calculation at the same time. The whitened data is grouped and input into the iterative formula calculation module in the form of a data stream, forming a pipeline structure. By utilizing the parallel computing capabilities of FPGA, iterative calculations can be completed more efficiently.

[0077] S4: Input the separated signal into the LFM radar signal pulse compression system for processing to obtain range information.

[0078] The LFM radar signal pulse compression system employs a matched filter for pulse compression. It correlates the echo signal (separated signal) extracted by the improved FastICA algorithm module with the original signal, narrowing the wide pulse signal and deriving the correlation peak. The output signal of the matched filter incorporates the similarity between the radar echo signal and the locally generated fine pulse signal. When the echo signal and the fine pulse signal are perfectly matched in time and frequency, the output signal reaches its maximum value, reflecting the true distance of the target. This achieves accurate measurement of the target distance and obtains distance information.

[0079] In summary, this invention proposes a multi-stage pipelined parallel processing structure and modular function reuse, which significantly improves the real-time performance of the algorithm. Furthermore, the original two-step iterative formula is merged into a single module based on the similarities and differences between the formulas. Using calculation identifiers, the formula used in the current operation can be distinguished and the data substituted for calculation, greatly reducing hardware resource consumption and significantly improving the real-time performance of the algorithm. Simultaneously, the reuse of circuit modules greatly reduces the resource consumption of the computation module, accelerates the calculation process, meets the real-time requirements of signal separation applications, and helps to reduce the size of the chip circuit and lower manufacturing costs.

[0080] The above-described embodiments further illustrate the purpose, technical solution, and advantages of the present invention. It should be understood that the above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made to the present invention within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A radar ranging method based on the FastICA algorithm, characterized in that, include: S1: Collect observation signal data from multiple radars; S2: Preprocess the multi-channel observation signal data to obtain whitened data; the preprocessing includes centering and whitening. S3: The whitened data is input into the improved FastICA algorithm module for processing to obtain the separated signal; the process of the improved FastICA algorithm module processing the whitened data includes: S31: Perform iterative calculations on the whitened data according to the traditional iterative formula to obtain the intermediate variable matrix and iterative intermediate parameters; S32: Iteratively calculate the intermediate variable matrix and intermediate iteration parameters according to the improved iterative formula to obtain the initial unmixing matrix; the process of iteratively calculating the intermediate variable matrix and intermediate iteration parameters according to the improved iterative formula includes: The column vectors of multiple intermediate variable matrices and the whitened data are multiplied together to obtain multiple secondary separated signal data; Substituting the secondary separated signal data, intermediate variable matrix, and whitening data into the improved iterative formula, multiple column vectors of initial unmixed matrices are obtained; the improved iterative formula is: ； in, This represents the column vector of the i-th initial unmixing matrix. This represents the column vector of the i-th intermediate variable matrix. This represents the i-th group of whitened data. Indicates the intermediate parameters of the initial iteration. This represents the i-th secondary separated signal data; Calculate the average value of the corresponding elements of the column vectors of multiple initial unmixing matrices, and combine the calculated column vectors to obtain the updated unmixing matrix; S33: Orthogonalize and normalize the initial unmixing matrix to obtain the updated unmixing matrix; S34: Determine whether the unmixing matrix is ​​converged. If it is converged, output the unmixing matrix; otherwise, return to step S31. S35: Multiply the demixing matrix with the whitened data to obtain the separated signal; S4: Input the separated signal into the LFM radar signal pulse compression system for processing to obtain range information.

2. The radar ranging method based on the FastICA algorithm according to claim 1, characterized in that, The process of iteratively calculating whitened data according to traditional iterative formulas includes: The column vectors of the unmixing matrix are multiplied by each group of data in the whitened data to obtain multiple initial separated signal data; the unmixing matrix is ​​randomly generated during the first calculation using the traditional iterative formula. The initial separated signal data, unmixed matrix, and whitened data are substituted into the traditional iterative formula for calculation to obtain column vectors of multiple initial intermediate variable matrices; multiple initial iterative intermediate parameters are calculated based on the multiple initial separated signal data. The updated intermediate variable matrix and iterative intermediate parameters are obtained by averaging the column vectors of multiple initial intermediate variable matrices and the multiple initial iterative intermediate parameters.

3. The radar ranging method based on the FastICA algorithm according to claim 2, characterized in that, The traditional iterative formula is as follows: ； in, This represents the column vector of the i-th intermediate variable matrix. This represents the i-th group of whitened data. This represents the column vector of the i-th unmixing matrix. This represents the i-th initial separated signal data.

4. The radar ranging method based on the FastICA algorithm according to claim 2, characterized in that, The formula for calculating the initial iterative intermediate parameters based on the initial separated signal data is as follows: ； in, Indicates the intermediate parameters of the initial iteration. This represents the i-th initial separated signal data.

5. The radar ranging method based on the FastICA algorithm according to claim 1, characterized in that, The process of determining whether the unmixed matrix is ​​convergent includes: setting a threshold for the difference between the unmixed matrix and a maximum number of iterations. When the absolute value of the difference between the current unmixed matrix and the unmixed matrix obtained in the previous iteration is less than the threshold for the difference between the unmixed matrix or the number of iterations reaches the maximum number of iterations, the unmixed matrix is ​​determined to be convergent; otherwise, the unmixed matrix is ​​determined to be non-convergent.

Images