A high-precision method for reducing noise of radar target signals

Abstract

The present application provides a high-precision radar target signal noise reduction method, which relates to the fields of radio frequency technology and wireless communication technology, and includes the following steps: Step 1: Perform FFT transformation on the time domain signal received from the millimeter wave radar to obtain a frequency domain signal; Step 2: Use the RML algorithm to process the frequency domain signal to obtain the target signal after preliminary filtering; Step 3: Calculate the gradient of the target signal; Step 4: Update the momentum variable and second-order moment variable of the target signal gradient; Step 5: Perform deviation correction on the momentum variable and second-order moment variable; Step 6: Calculate the update step size and determine the update direction; Step 7: Use the update step size to update the frequency domain estimation to obtain the frequency domain estimation result. The present application uses the RML algorithm combined with the Adam algorithm for estimation and optimization, which improves the accuracy of the frequency domain estimation, effectively removes various clutters, and improves the target signal extraction performance of the radar system.

Description

Technical Field

[0001] The present application relates to the fields of radio frequency technology and wireless communication technology, and in particular to a high-precision radar target signal noise reduction method. Background Art

[0002] When detecting target signals, FMCW (Frequency Modulated Continuous Wave) radar transmits a continuous wave with changing frequency within the frequency sweep period. The echo reflected by the object has a certain frequency difference with the transmitted signal. By measuring the frequency difference, the distance information between the target and the radar can be obtained. The difference frequency signal has a low frequency. Since the application environment is usually more complex, it will face the problem of noise interference. In the existing technology, Doppler FFT method or deep learning model is often used for noise reduction. The Doppler FFT method uses fast Fourier transform to perform spectrum analysis on the echo signal to extract the speed and distance information of the target. It has high computational efficiency and is suitable for real-time processing. However, it is sensitive to noise and requires a high signal-to-noise ratio to work effectively. It cannot adapt to noise reduction work at low signal-to-noise ratio. The deep learning model can automatically adjust parameters to achieve the best denoising effect by learning the noise pattern in the training data. It has good denoising effect and strong adaptability. However, it has high computational complexity, requires a large amount of training data and computing resources, and is difficult to quickly adapt to complex and changing environments. Summary of the invention

[0003] This application is made in view of the above-mentioned subject, and its purpose is to provide a high-precision radar target signal noise reduction method, firstly, the frequency domain signal is accurately estimated by using the RML algorithm, and then optimized in combination with the Adam algorithm to reduce the influence of noise and improve the accuracy of frequency domain estimation, which not only improves the accuracy and robustness of frequency domain estimation, but also reduces the computational complexity. It can effectively remove various clutters when processing target signals, improve the target signal extraction performance of the radar system, help reduce the influence of environmental clutter, and can also adapt to the processing requirements under different signal-to-noise ratio conditions, providing a more reliable solution for target detection of radar systems in complex scenes.

[0004] Specifically, the first aspect of the present application provides a high-precision radar target signal noise reduction method, comprising the following steps:

[0005] Step 1: Perform FFT transformation on the time domain signal received from the millimeter wave radar to obtain the frequency domain signal;

[0006] FFT (Fast Fourier Transform) is an efficient discrete Fourier transform algorithm used to convert time domain signals into frequency domain signals, which makes spectrum analysis and filter design more convenient.

[0007] Step 2: Use the RML algorithm to process the frequency domain signal to obtain the target signal after preliminary filtering;

[0008] Step 3: Calculate the gradient of the target signal;

[0009] Step 4: Use the Adam algorithm to update the momentum variable and second-order moment variable of the target signal gradient;

[0010] Step 5: bias correction of momentum variables and second-order moment variables;

[0011] Step 6: Combine the momentum variable and the second-order moment variable after bias correction to calculate the update step size and determine the update direction of the frequency domain estimation;

[0012] Step 7: Use the update step size to update the frequency domain estimation to obtain a more accurate frequency domain estimation result.

[0013] Furthermore, the formula of the FFT transformation is as follows:

[0014] ;

[0015] in : frequency domain signal; : Target millimeter wave radar signal;

[0016] : angular frequency; : Phase difference.

[0017] Furthermore, the RML algorithm is used to process the frequency domain signal, and the formula is as follows:

[0018] ;

[0019] in : Target output signal; : matched filter;

[0020] : Convolution operation; : Frequency domain signal.

[0021] Further, the calculating the gradient of the target signal specifically includes:

[0022] Step 1: Calculate the loss function of the target signal. The formula is as follows:

[0023] ;

[0024] in: : loss function of the target signal; : frequency domain signal;

[0025] : Target output signal;

[0026] Step 2: Calculate the derivative of the loss function to obtain the gradient of the frequency domain signal. The formula is as follows:

[0027] ;

[0028] in : The gradient of the loss function with respect to the frequency domain signal;

[0029] : Derivative operation of the loss function; : The starting state of the operation.

[0030] Furthermore, the Adam algorithm is used to update the momentum variable and the second-order moment variable of the target signal gradient, and the formula is as follows:

[0031] ;

[0032] ;

[0033] in : updated momentum variable; : Momentum variable before updating;

[0034] : momentum decay rate; : The gradient of the loss function with respect to the frequency domain signal;

[0035] : Updated second-order moment variable; : The second-order moment variable before updating;

[0036] : Second-order moment decay rate.

[0037] Used to adjust the influence of past gradients on the momentum variable. Used to adjust the influence of the square of the past gradient on the second-order moment variable. This will make the influence of past gradients greater, thereby increasing the inertia of momentum and making the update process smoother and more stable. Similar, larger This will make the influence of the square of the past gradient greater, thereby increasing the inertia of the second-order moment and making the update process smoother and more stable.

[0038] Furthermore, the momentum variable and the second-order moment variable are subjected to deviation correction, and the formula is as follows:

[0039] ;

[0040] ;

[0041] in : Corrected momentum variable; : updated momentum variable;

[0042] : the momentum decay rate to the power of t;

[0043] : Corrected second-order moment variable; : Updated second-order moment variable;

[0044] : The t-th power of the second-order moment decay rate.

[0045] Furthermore, the calculation update step size determines the update direction of the frequency domain estimation, and the formula is as follows:

[0046] ;

[0047] ;

[0048] in : Update step size; : learning rate; : Update direction;

[0049] : symbolic function; : Corrected momentum variable;

[0050] : Momentum variable before update.

[0051] Furthermore, the calculation rule of the symbol function is as follows:

[0052] .

[0053] Furthermore, the frequency domain estimation is updated using the update step size, and the formula is as follows:

[0054] ;

[0055] in : Updated frequency domain signal; : Estimation signal in frequency domain;

[0056] : Update step size.

[0057] Furthermore, the signal-to-noise ratio of the updated frequency domain signal is greater than the signal-to-noise ratio of the initial time domain signal.

[0058] In the second aspect, the present application also provides a computing device, which has the function of implementing the method described in the first aspect above. The beneficial effects can be found in the description of the first aspect, which will not be repeated here. The function can be implemented by hardware, or by executing the corresponding software implementation through hardware. The hardware or software includes one or more modules corresponding to the above functions. In one possible design, the structure of the device includes an acquisition module, a training module, and optionally, a construction module. These modules can implement the function of the training node in the method example of the first aspect above. Please refer to the detailed description in the method example for details, which will not be repeated here.

[0059] In the third aspect, the present application also provides a computing device, which is used to implement the functions of the method described in the first aspect above. The beneficial effects can be found in the description of the first aspect and will not be repeated here. The structure of the computing device includes a processor and a memory, and the memory is used to store instructions and / or data. The memory is coupled to the processor, and when the processor executes the program instructions stored in the memory, the function of the training node in the example of the first aspect above can be implemented. The structure of the computing device also includes a communication interface for communicating with other devices.

[0060] In a fourth aspect, the present application also provides a computer-readable storage medium, in which instructions are stored, and when the computer-readable storage medium is run on a computer, the computer executes the method in the above-mentioned first aspect and various possible designs of the first aspect.

[0061] In a fifth aspect, the present application also provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute the method in the above-mentioned first aspect and various possible designs of the first aspect.

[0062] In a sixth aspect, the present application also provides a computing chip, which is connected to a memory, and the chip is used to read and execute a software program stored in the memory, and to execute the methods in the above-mentioned first aspect and various possible implementation methods of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the embodiments of the present drawings or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present drawings. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying any creative work.

[0064] Figure 1 A flowchart of this application;

[0065] Figure 2This is a comparison chart of the results of this application in Experiment 1 with the traditional RML algorithm;

[0066] Figure 3 This is a comparison chart of the results of this application in Experiment 2 with the traditional RML algorithm.

[0067] The purpose, features and advantages of this figure will be further described in conjunction with the embodiments with reference to the accompanying drawings. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments provided in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0069] Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. In addition, it can also be understood that although the efforts made in this development process may be complicated and lengthy, for ordinary technicians in this field related to the content disclosed in this application, some changes in design, manufacturing or production based on the technical content disclosed in this application are just conventional technical means, and should not be understood as insufficient content disclosed in this application.

[0070] Unless otherwise specified, all embodiments and optional embodiments of the present application can be combined with each other to form a new technical solution.

[0071] Unless otherwise specified, all technical features and optional technical features of this application can be combined with each other to form a new technical solution.

[0072] If there is no special explanation, all steps of the present application can be performed sequentially or randomly, preferably sequentially. For example, the method includes steps (a) and (b), which means that the method may include steps (a) and (b) performed sequentially, or may include steps (b) and (a) performed sequentially. For example, the method may further include step (c), which means that step (c) may be added to the method in any order. For example, the method may include steps (a), (b) and (c), or may include steps (a), (c) and (b), or may include steps (c), (a) and (b), etc.

[0073] If there is no special explanation, the "include" and "comprising" mentioned in this application are open-ended or closed-ended. For example, the "include" and "comprising" may mean that other components not listed may also be included or only the listed components may be included or only the listed components may be included.

[0074] If not specifically stated, in this application, the term "or" is inclusive. For example, the phrase "A or B" means "A, B, or both A and B". More specifically, any of the following conditions satisfies the condition "A or B": A is true (or exists) and B is false (or does not exist); A is false (or does not exist) and B is true (or exists); or both A and B are true (or exist).

[0075] In order to better understand the solutions of the embodiments of the present application, some relevant terms and concepts that may be involved in the embodiments of the present application are first introduced below.

[0076] (1) FFT (Fast Fourier Transform) is an efficient algorithm for discrete Fourier transform. In signal processing, Fourier transform is a very important mathematical tool that can convert signals from the time domain to the frequency domain, thus helping us analyze the frequency components and characteristics of the signal. In digital signal processing, the discrete form of Fourier transform is called fast Fourier transform.

[0077] (2) RML (Recursive Maximum Likelihood) algorithm: RML is a dynamic parameter estimation method mainly used for time series models such as hidden Markov models (HMM) and autoregressive models. Its core idea is to estimate model parameters online or incrementally, rather than processing all data at once. This method continuously updates parameter values ​​as new data are added during the parameter estimation process.

[0078] (3) Adam (Adaptive Moment Estimation) algorithm is an optimization algorithm that combines the momentum method and adaptive learning rate. The core idea is to use the first-order momentum (mean) and second-order momentum (uncentered variance) of the gradient to update the network weights. Specifically, it calculates the first-order moment (mean) and second-order moment (uncentered variance) of the gradient and performs bias correction on them to ensure the accuracy of gradient estimation in the early stages of training.

[0079] In this embodiment, Figure 1 As shown, a high-precision radar target signal noise reduction method comprises the following steps:

[0080] Step 1: Perform FFT transformation on the time domain signal received from the millimeter wave radar to obtain the frequency domain signal;

[0081] FFT transform is an efficient algorithm of discrete Fourier transform, which is used to convert time domain signals into frequency domain signals, making spectrum analysis and filter design more convenient.

[0082] Step 2: Use the RML algorithm to process the frequency domain signal to obtain the target signal after preliminary filtering;

[0083] Step 3: Calculate the gradient of the target signal;

[0084] Step 4: Use the Adam algorithm to update the momentum variable and second-order moment variable of the target signal gradient;

[0085] Step 5: bias correction of momentum variables and second-order moment variables;

[0086] Step 6: Combine the momentum variable and the second-order moment variable after bias correction to calculate the update step size and determine the update direction of the frequency domain estimation;

[0087] Step 7: Use the update step size to update the frequency domain estimation to obtain a more accurate frequency domain estimation result.

[0088] Furthermore, the formula for FFT transformation is as follows:

[0089] .

[0090] Furthermore, the RML algorithm is used to process the frequency domain signal, and the formula is as follows:

[0091] .

[0092] Furthermore, the gradient of the target signal is calculated, specifically including:

[0093] Step 1: Calculate the loss function of the target signal. The formula is as follows:

[0094] ;

[0095] Step 2: Calculate the derivative of the loss function to obtain the gradient of the frequency domain signal. The formula is as follows:

[0096] .

[0097] Furthermore, the Adam algorithm is used to update the momentum variable and second-order moment variable of the target signal gradient. The formula is as follows:

[0098] ;

[0099] .

[0100] Used to adjust the influence of past gradients on the momentum variable. Used to adjust the influence of the square of the past gradient on the second-order moment variable. This will make the influence of past gradients greater, thereby increasing the inertia of momentum and making the update process smoother and more stable. Similar, larger This will make the influence of the square of the past gradient greater, thereby increasing the inertia of the second-order moment and making the update process smoother and more stable.

[0101] In this embodiment, set =0.9, =0.99.

[0102] Furthermore, the momentum variable and the second-order moment variable are corrected for deviations, and the formula is as follows:

[0103] ;

[0104] .

[0105] Furthermore, the update step size is calculated to determine the update direction of the frequency domain estimation. The formula is as follows:

[0106] ;

[0107] .

[0108] Furthermore, the calculation rules of the symbolic function are as follows:

[0109] .

[0110] Furthermore, the frequency domain estimation is updated using the update step size, and the formula is as follows:

[0111] .

[0112] Furthermore, the signal-to-noise ratio of the updated frequency domain signal is greater than the signal-to-noise ratio of the initial time domain signal.

[0113] In this embodiment, two groups of experiments were conducted to illustrate the practical effect of the present application, as follows:

[0114] Experiment 1: Fixed the radar position, selected a person as the experimental subject, and let the experimental subject lie still on the bed. Adjust the environment and the signal-to-noise ratio of the signal. The signal-to-noise ratio range is set between 5dB and 30dB. The experimental results are as follows Figure 2 As shown, Figure 2The traditional RML algorithm is compared in , and it can be seen that the root mean square error of the proposed method is smaller than that of the traditional RML algorithm under different signal-to-noise ratio environments, indicating that the proposed method has a smaller estimation error than the traditional RML method.

[0115] Experiment 2: Fixed the radar position, selected one person as the experimental subject, the experimental subject was in a moving state, adjusted the environment and the signal-to-noise ratio, and the signal-to-noise ratio range was set between 5dB and 30dB. The experimental results are as follows Figure 3 As shown, Figure 3 The traditional RML algorithm is compared in , and it can be seen that the root mean square error of the proposed method is much smaller than that of the traditional RML algorithm under different signal-to-noise ratio environments, indicating that the proposed method has a smaller estimation error than the traditional RML method in complex environments.

[0116] It should be noted that the present application is not limited to the above-mentioned embodiments. The above-mentioned embodiments are only examples, and the embodiments having the same structure as the technical idea and exerting the same effect within the scope of the technical solution of the present application are all included in the technical scope of the present application. In addition, without departing from the scope of the main purpose of the present application, various modifications that can be thought of by those skilled in the art to the embodiments and other methods of combining some of the constituent elements in the embodiments are also included in the scope of the present application.

Claims

1. A high-precision radar target signal noise reduction method, characterized in that: The following steps are involved: Step 1: Perform FFT transformation on the time domain signal received from the millimeter wave radar to obtain the frequency domain signal; Step 2: Use the RML algorithm to process the frequency domain signal to obtain the target signal after preliminary filtering; Step 3: Calculate the gradient of the target signal; Step 4: Use the Adam algorithm to update the momentum variable and second-order moment variable of the target signal gradient; Step 5: Perform bias correction on momentum variables and second-order moment variables. The formula is as follows: ； ； in : updated momentum variable; : Momentum variable before updating; : momentum decay rate; : The gradient of the loss function with respect to the frequency domain signal; : Updated second-order moment variable; : The second-order moment variable before updating; : second-order moment decay rate; Step 6: Combine the momentum variable and the second-order moment variable after bias correction to calculate the update step size and determine the update direction of the frequency domain estimate. The formula is as follows: ； ； in : Update step size; : learning rate; : Update direction; : symbolic function; : Corrected momentum variable; : Momentum variable before updating; Step 7: Use the update step size to update the frequency domain estimation to obtain a more accurate frequency domain estimation result.

2. A high-precision radar target signal noise reduction method according to claim 1, characterized in that: The formula for the FFT transform is as follows: ； in : frequency domain signal; : Target millimeter wave radar signal; : angular frequency; : Phase difference.

3. A high-precision radar target signal noise reduction method according to claim 1, characterized in that: The RML algorithm is used to process the frequency domain signal, and the formula is as follows: ； in : Target output signal; : matched filter; *: convolution operation; : Frequency domain signal.

4. A high-precision radar target signal noise reduction method according to claim 1, characterized in that: The calculating the gradient of the target signal specifically includes: Step 1: Calculate the loss function of the target signal. The formula is as follows: ； in: : loss function of the target signal; : frequency domain signal; : Target output signal; Step 2: Calculate the derivative of the loss function to obtain the gradient of the frequency domain signal. The formula is as follows: ； in : The gradient of the loss function with respect to the frequency domain signal; : Derivative operation of the loss function; : The starting state of the operation.

5. A high-precision radar target signal noise reduction method according to claim 1, characterized in that: The Adam algorithm is used to update the momentum variable and the second-order moment variable of the target signal gradient, and the formula is as follows: ； ； in : updated momentum variable; : Momentum variable before updating; : momentum decay rate; : The gradient of the loss function with respect to the frequency domain signal; : Updated second-order moment variable; : The second-order moment variable before updating; : Second-order moment decay rate.

6. A high-precision radar target signal noise reduction method according to claim 1, characterized in that: The calculation rules of the symbol function are as follows: 。 7. A high-precision radar target signal noise reduction method according to claim 1, characterized in that: The frequency domain estimation is updated using the update step size, and the formula is as follows: ； in : Updated frequency domain signal; : Estimation signal in frequency domain; : Update step size.

8. A high-precision radar target signal noise reduction method according to claim 7, characterized in that: The signal-to-noise ratio of the updated frequency domain signal is greater than the signal-to-noise ratio of the initial time domain signal.

Images