Millimeter wave radar signal processing method and device, storage medium and computer program product

Abstract

The invention provides a millimeter wave radar signal processing method and device, a storage medium and a computer program product, and relates to the technical field of radar signal processing. The method comprises the following steps: framing and windowing an original time domain signal; performing fast Fourier transform on the windowed time domain signal to generate an initial frequency domain signal; determining a dynamic gain factor according to the signal-to-noise ratio of the current frame; determining a smoothing factor according to the dynamic gain factor; performing smoothing processing on the dynamic gain factor by using the smoothing factor and the smoothed gain factor of the adjacent frame to generate a smoothed gain factor of the current frame; generating a frequency spectrum parameter subjected to smoothing gain according to the smoothed gain factor and the initial frequency spectrum parameter; performing inverse fast Fourier transform on the frequency domain signal subjected to smooth gain processing to obtain an initial de-noised time domain signal; and windowing and overlapping adding are carried out on the frame of the initial denoising time domain signal to generate a final denoising time domain signal. The present invention can suppress music noise.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application generally relate to the technical field of radar signal processing, and more particularly, to a millimeter wave radar signal processing method, device, storage medium and computer program product. BACKGROUND

[0002] The core goal of radar signal preprocessing is to separate the effective target signal from the signal contaminated by background noise and electromagnetic interference, and to preserve the details of the target signal, improve the signal-to-noise ratio (SNR), and provide a pure signal for subsequent target detection and tracking, thereby improving the target detection performance of the radar system.

[0003] The current radar signal preprocessing, the most mainstream frequency domain noise reduction technique is mainly the spectral subtraction method.

[0004] The spectral subtraction method is one of the most widely used techniques in radar frequency domain noise reduction. The core logic is to estimate the noise power spectrum in the frequency domain, subtract the noise component from the noisy signal power spectrum, and then restore it to the time domain signal. The spectral subtraction method has small calculation amount and strong real-time performance, and is particularly suitable for embedded radar systems (such as vehicle-mounted millimeter wave radar and security radar).

[0005] The spectral subtraction method mainly separates the target signal and the background noise in the frequency domain. By estimating the noise spectrum without target, the power spectral density of the noise signal is obtained after Fourier transform, and is subtracted from the power spectral density of the noisy signal to obtain the corrected target signal power spectrum. Finally, the frequency domain signal is restored to the time domain signal to obtain the denoised target signal.

[0006] However, the traditional spectral subtraction method is prone to produce "music noise" when preprocessing the noisy signal. The essence of the "music noise" is the random narrowband noise after spectral subtraction, which can cause noise estimation bias and further cause spectral distortion of the denoised signal. SUMMARY

[0007] To solve the above problems in the prior art, in a first aspect, embodiments of the present application provide a millimeter wave radar signal processing method, the method comprising: frame windowing an original time domain signal of a millimeter wave radar signal to generate a windowed time domain signal; performing a fast Fourier transform on the windowed time domain signal to generate an initial frequency domain signal, the initial frequency domain signal having an initial frequency spectrum parameter; determining a dynamic gain factor of a current frame according to a signal-to-noise ratio of the current frame in the initial frequency domain signal, wherein the dynamic gain factor of the current frame is positively correlated with the signal-to-noise ratio of the current frame; determining a smoothing factor of the current frame according to the dynamic gain factor of the current frame; performing smoothing processing on the dynamic gain factor of the current frame by using the smoothing factor and a smoothed gain factor of a neighboring frame to generate a smoothed gain factor of the current frame; generating a smoothed gain frequency spectrum parameter of the current frame according to the smoothed gain factor of the current frame and the initial frequency spectrum parameter of the current frame; and performing an inverse fast Fourier transform on a smoothed gain processed frequency domain signal represented by the smoothed gain frequency spectrum parameter to obtain an initial denoised time domain signal; and performing windowing and overlap addition on frames of the initial denoised time domain signal to generate a final denoised time domain signal.

[0008] In some embodiments, determining the smoothing factor of the current frame according to the dynamic gain factor of the current frame comprises: determining the smoothing factor of the current frame according to a predefined maximum gain factor and the dynamic gain factor of the current frame.

[0009] In some embodiments, the value of the smoothing factor is negatively correlated with the value of the dynamic gain factor, and performing smoothing processing on the dynamic gain factor of the current frame by using the smoothing factor and the smoothed gain factor of the neighboring frame comprises: respectively weighting the smoothed gain factor of the neighboring frame and the dynamic gain factor of the current frame according to the smoothing factor, wherein the weight of the smoothed gain factor of the neighboring frame is positively correlated with the value of the smoothing factor, and the weight of the dynamic gain factor of the current frame is negatively correlated with the value of the smoothing factor.

[0010] In some embodiments, the smoothed gain factor of the current frame is generated according to the following formula:

[0011]

[0012] wherein, denotes the smoothed gain factor of the current frame, denotes the smoothing factor of the current frame, denotes the smoothed gain factor of a previous frame, denotes the dynamic gain factor of the current frame.

[0013] In some embodiments, determining the dynamic gain factor of the current frame according to the signal-to-noise ratio of the current frame in the initial frequency domain signal comprises: determining the dynamic gain factor of the current frame as a predefined first gain factor value when the signal-to-noise ratio of the current frame is less than or equal to a predefined first signal-to-noise ratio threshold; determining the dynamic gain factor of the current frame as a predefined second gain factor value when the signal-to-noise ratio of the current frame is greater than or equal to a predefined second signal-to-noise ratio threshold; and determining the dynamic gain factor of the current frame as a value that increases with the signal-to-noise ratio between the first gain factor value and the second gain factor value when the signal-to-noise ratio of the current frame is greater than the first signal-to-noise ratio threshold and less than the second signal-to-noise ratio threshold.

[0014] In some embodiments, the square value of the dynamic gain factor of the current frame and the value of the signal-to-noise ratio of the current frame are in a linear relationship when the signal-to-noise ratio of the current frame is greater than the first signal-to-noise ratio threshold and less than the second signal-to-noise ratio threshold.

[0015] In some embodiments, the windowing performed on the original time domain signal uses a 50% overlapping Hamming window.

[0016] In a second aspect, embodiments of the present application provide a millimeter wave radar signal processing device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements the millimeter wave radar signal processing method described in any of the above embodiments.

[0017] In a third aspect, embodiments of the present application provide a storage medium storing computer readable instructions, which, when executed by a processor, perform the millimeter wave radar signal processing method described in any of the above embodiments.

[0018] In a fourth aspect, embodiments of the present application provide a computer program product comprising computer readable instructions, which, when executed by a processor, perform the steps of the millimeter wave radar signal processing method described in any of the above embodiments.

[0019] The millimeter wave radar signal processing method, device, storage medium and computer program product provided by the embodiments of the present application can suppress the defect that the traditional spectral subtraction is prone to cause noise estimation bias and thus produce random "music noise" by using the method of "SNR-driven dynamic gain + iterative smoothing".

[0020] The dynamic gain can accurately match the signal characteristics, the gain values used at different signal-to-noise ratios are different, the dynamic changes of the SNR in the radar signal are adapted, and the noise estimation bias can be avoided from being too large.

[0021] Compared with the "full-band uniform gain" or "fixed threshold gain" commonly used in the prior art, the embodiments of the present application dynamically adjust the gain factor according to the real-time signal-to-noise ratio (SNR) of the millimeter wave radar signal, and introduce a smoothing factor to realize the inter-frame dynamic smoothing of the gain through smoothing iteration. The iterative smoothing can effectively solve the gain mutation that may occur between multiple frames of signals, smooth the random fluctuations, avoid the generation of pulses, and improve the robustness of the overall signal.

[0022] The smoothing strength in the embodiments of the present application is adaptively adjusted according to the SNR, avoiding the problems of "excessive smoothing leading to signal blurring" or "insufficient smoothing leaving musical noise".

[0023] The embodiments of the present application make innovations at the signal reconstruction level, design a multi-frame associated overlap-add strategy, and solve the distortion of frame splicing. The "square root + sine correction" design of the window function further reduces the mutation degree of the frame edge signal compared with the traditional Hamming window, and preserves the credibility of the signal. The square root can smooth the attenuation curve of the window function edge, avoiding the excessive suppression of the original signal. The sine correction can further smooth the mutation that may occur at the frame edge, solve the distortion during frame splicing, and provide higher credibility for the overlap-add restoration of the original signal.

[0024] Since millimeter wave radars (such as vehicle-mounted and security radars) need to be implemented on embedded platforms (such as FPGA and MCU) for microsecond or millisecond level real-time processing, the embodiments of the present application realize the effect of achieving both noise reduction and real-time through low-computational-load algorithm optimization and algorithm process simplification. Based on the FPGA platform test results, the single-frame processing time can be controlled within 1ms, which can meet the real-time requirements of vehicle-mounted radars for processing more than 1000 frames per second. BRIEF DESCRIPTION OF DRAWINGS

[0025] The above and other objects, features and advantages of the embodiments of the present application will become more apparent from the following detailed description, taken in conjunction with the accompanying drawings, in which several embodiments of the present application are shown by way of example, and which are not intended to limit the present application to the specific embodiments, and in which:

[0026] Figure 1 A flowchart of a millimeter wave radar signal processing method according to an embodiment of the present application is shown;

[0027] Figure 2 A window function position diagram according to an embodiment of the present application is shown;

[0028] Figure 3 A time-domain signal simulation diagram according to an embodiment of the present application is shown;

[0029] Figure 4A and Figure 4BSpectrogram diagrams of a millimeter wave radar signal processing method according to a conventional spectral subtraction and according to an embodiment of the present application are shown respectively.

[0030] In the drawings, identical or corresponding reference signs indicate identical or corresponding parts. DETAILED DESCRIPTION

[0031] The principles and spirits of the present application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are given only so that those skilled in the art can better understand and implement the present application, and do not limit the scope of the present application in any way.

[0032] In one aspect, an embodiment of the present application provides a millimeter wave radar signal processing method. Reference is made to Figure 1 which shows a flowchart of a millimeter wave radar signal processing method 100 according to an embodiment of the present application. The method 100 can include steps S101-S108.

[0033] In step S101, the original time domain signal of the millimeter wave radar signal is framed and windowed to generate a windowed time domain signal. This step performs windowing processing on the signal before performing spectral analysis on the signal.

[0034] The original time domain signal s(n) of the millimeter wave radar signal may, for example, be an intermediate frequency echo signal of a millimeter wave radar, etc., which usually contains target signals and background noise.

[0035] As an embodiment of the present application, the windowing performed on the original time domain signal adopts a 50% overlapping Hamming window. Reference is made to Figure 2 which shows a window function position diagram according to an embodiment of the present application. The spectral analysis is performed twice per frame, using two Hamming windows with 50% overlap, in preparation for the subsequent overlap-add algorithm. The window function is shown in the following formula (1):

[0036]

[0037] wherein W FFT (n) is the window function, n = 0, …, 255.

[0038] The use of the 50% overlapping Hamming window can more effectively avoid the spectral leakage caused by framing.

[0039] In step S102, a Fast Fourier Transform (FFT) is performed on the windowed time-domain signal to generate an initial frequency-domain signal X(n), which has initial spectral parameters. This step performs spectral analysis on the sampled signal, that is, analyzes the spectral structure of the signal through Fourier transform. By performing the Fast Fourier Transform (FFT), the time-domain signal is converted into a frequency-domain signal so as to separate the target signal from noise in the frequency domain.

[0040] After performing a Fourier transform on the windowed signal, two sets of initial spectral parameters are obtained, which can also be called initial spectral analysis parameters. The initial spectral parameters can be expressed, for example, as follows: and They represent the real and imaginary parts of the nth frequency point, respectively.

[0041] In step S103, the dynamic gain factor g of the current frame is determined based on the signal-to-noise ratio (SNR) of the current frame in the initial frequency domain signal. s The dynamic gain factor of the current frame is positively correlated with the signal-to-noise ratio (SNR) of the current frame. That is, the dynamic gain factor is adaptively adjusted according to the SNR, with a small gain for low SNR and a high gain or full gain for high SNR.

[0042] As an embodiment of the present invention, step S103 may include: when the signal-to-noise ratio (SNR) of the current frame is less than or equal to a predefined first SNR threshold, determining the dynamic gain factor of the current frame as a predefined first gain factor value, also known as the minimum gain factor value; when the SNR of the current frame is greater than or equal to a predefined second SNR threshold, determining the dynamic gain factor of the current frame as a predefined second gain factor value, also known as the maximum gain factor value; when the SNR of the current frame is greater than the first SNR threshold and less than the second SNR threshold, determining the dynamic gain factor of the current frame as a value that increases with the increase of the SNR between the first gain factor value and the second gain factor value.

[0043] As an example only, a dynamic gain factor g can be defined. s The range of values ​​for g is min ≤g s ≤1, where is a predefined first gain factor value, i.e., the minimum gain factor; 1 is a predefined second gain factor value, i.e., the maximum gain factor. The first signal-to-noise ratio threshold can be set to 1, for example, when the SNR of the current frame ≤ 1, g s The value is g min This avoids excessive signal amplification leading to noise enhancement in low signal-to-noise ratio scenarios. (For example only, g...) min =0.19953, this value is obtained by calculating the gain at -14dB.

[0044] When the signal-to-noise ratio of the current frame is greater than or equal to a predefined second signal-to-noise ratio threshold, it can be generally considered that the target echo signal is already much stronger than the noise signal, and thus no additional gain adjustment is needed, and g s is taken as a fixed gain factor maximum value. In the case where the predefined second gain factor value is 1, g s = 1. By way of example only, the second signal-to-noise ratio threshold can have a value of 45. When SNR ≥ 45, g s = 1.

[0045] As an embodiment of the present application, when the signal-to-noise ratio of the current frame is greater than a first signal-to-noise ratio threshold and less than a second signal-to-noise ratio threshold, the square value of the dynamic gain factor of the current frame has a linear relationship with the value of the signal-to-noise ratio of the current frame.

[0046] By way of example, with the first signal-to-noise ratio threshold being 1 and the second signal-to-noise ratio threshold being 45 and the second gain factor value (gain factor maximum value) being 1, when 1 ≤ SNR ≤ 45, g s is determined by equation (2):

[0047]

[0048] where SNR is the signal-to-noise ratio of the input signal, and is a fixed coefficient for gain factor calculation, is the slope of the linear relationship, and determines the steepness when the gain factor transitions from g min to 1, is the intercept of the linear relationship. and have values determined according to equations (3) and (4) below:

[0049]

[0050] In step S104, a smoothing factor for the current frame is determined according to the dynamic gain factor of the current frame. In an embodiment of the present application, the smoothing factor is derived according to the dynamic gain factor.

[0051] As an embodiment of the present application, step S104 can include determining the smoothing factor for the current frame according to a predefined gain factor maximum value and the dynamic gain factor of the current frame .

[0052] As a first example, the smoothing factor may be the predefined gain factor maximum value minus the value of the dynamic gain factor of the current frame, i.e., where denotes the smoothing factor, denotes the predefined gain factor maximum value, denotes the dynamic gain factor of the current frame. In particular, when

[0053] As a second example, the smoothing factor may be a ratio of the dynamic gain factor of the current frame to a predefined gain factor maximum, i.e., wherein denotes the smoothing factor, denotes the predefined gain factor maximum, and denotes the dynamic gain factor of the current frame.

[0054] As a third example, the smoothing factor may be an inverse of a value of the predefined gain factor maximum minus the dynamic gain factor of the current frame, i.e., wherein denotes the smoothing factor, denotes the predefined gain factor maximum, and denotes the dynamic gain factor of the current frame.

[0055] As a fourth example, the smoothing factor may be 1 minus a ratio of the dynamic gain factor of the current frame to the predefined gain factor maximum, i.e., wherein denotes the smoothing factor, denotes the predefined gain factor maximum, and denotes the dynamic gain factor of the current frame.

[0056] It can be seen that, in the above first and fourth examples, the value of the smoothing factor is negatively correlated with the value of the dynamic gain factor, i.e., the greater the dynamic gain factor, the smaller the smoothing factor. In the above second and third examples, the value of the smoothing factor is positively correlated with the value of the dynamic gain factor, i.e., the greater the dynamic gain factor, the greater the smoothing factor. In practical applications, the determination manner of the smoothing factor can be flexibly adjusted and various forms of normalization processing can be adopted according to needs.

[0057] In step S105, the smoothing processing is performed on the dynamic gain factor of the current frame by using the smoothing factor and the smoothed gain factors of the adjacent frames, to generate the smoothed gain factor of the current frame.

[0058] As one embodiment of the present application, step S105 can include: respectively performing the weighting processing on the smoothed gain factors of the adjacent frames and the dynamic gain factor of the current frame according to the smoothing factor.

[0059] ​​As an embodiment of the present application, the value of the smoothing factor can be negatively correlated with the value of the dynamic gain factor. In this case, the weight of the smoothed gain factor of the neighboring frame is positively correlated with the value of the smoothing factor, and the weight of the dynamic gain factor of the current frame is negatively correlated with the value of the smoothing factor.

[0060] As an embodiment of the present application, the neighboring frame can be the previous frame, and the smoothed gain factor of the current frame is generated according to the following formula (5):

[0061]

[0062] wherein, denotes the smoothed gain factor of the current frame, denotes the smoothing factor of the current frame, denotes the smoothed gain factor of the previous frame, denotes the dynamic gain factor of the current frame, the initial value of is 1.

[0063] As an embodiment of the present application, the neighboring frame can be the previous frame, and the smoothed gain factor of the current frame is generated according to the following formula (5):

[0064] As an embodiment of the present application, if the frequency domain signal is processed in the order from back to front, the neighboring frame can be the Nth frame after the current frame, N≥1, in particular, the next frame.

[0065] In the case of including N (N>1) neighboring frames, as an example, the same weight can be given to each neighboring frame. As another example, different weights can be given to each neighboring frame, for example, the closer to the current frame, the higher the weight.

[0066] The embodiment of the present application introduces a smoothing factor The smoothing processing of the gain factor can slow down the change speed of the gain factor, make the change of the gain factor tend to be stable, prevent the jitter of the gain factor, and thus weaken the music noise after the spectral subtraction.

[0067] The smoothing of the gain factor of the current frame using the smoothed gain factor of the neighboring frame uses the smoothing iteration logic, which actually includes a feedback link, and uses the result of the previous round of processing as the reference for the next round of processing, avoids the music noise caused by the gain mutation, and realizes the inter-frame smoothing update of the gain factor.

[0068] As another embodiment of the present application, the value of the smoothing factor can be positively correlated with the value of the dynamic gain factor. In this case, the weight of the smoothed gain factor of the neighboring frame is negatively correlated with the value of the smoothing factor, and the weight of the dynamic gain factor of the current frame is positively correlated with the value of the smoothing factor.

[0069] It can be seen that, no matter what the case, the gain factor weight of the current frame is positively correlated with the signal-to-noise ratio of the current frame.

[0070] Optionally, the smoothing factor used for performing smoothing processing on the dynamic gain factor of the current frame in step S105 can be the smoothing factor of the current frame, or the smoothing factor of a neighboring frame, for example, the smoothing factor of the previous frame.

[0071] In step S106, smoothed gain spectrum parameters of the current frame are generated according to the smoothed gain factor of the current frame and the initial spectrum parameters of the current frame.

[0072] This step uses the smoothed gain factor to weight the frequency domain signal, obtaining an optimized frequency domain signal, the core of which is to suppress the frequency domain components corresponding to the noise and retain the target signal components.

[0073] As an example, the gain factor is multiplied by the spectrum parameters of the current frame and the previous frame after Fourier transform respectively, to obtain new spectrum parameters, and two groups of .

[0074] In step S107, inverse fast Fourier transform (IFFT) is performed on the smoothed gain spectrum parameters to obtain an initial denoising time domain signal. This step restores the frequency domain signal to the time domain signal, obtaining a preliminary denoising time domain signal.

[0075] In step S108, windowing and overlap addition are performed on the frames of the initial denoising time domain signal to generate a final denoising time domain signal snr(t). This step reconstructs the IFFT time domain signal using the overlap addition method, for example, combining the signal of “past frame + current frame” to eliminate the time domain splicing distortion caused by the frame processing.

[0076] Only as an example, after inverse Fourier transform, two denoising signals After windowing according to formula (1), two windowed denoising signals Finally, the denoising signal is reconstructed using the overlap addition algorithm, as shown in the following formulas (6) and (7):

[0077]

[0078] wherein, is the final denoising time domain signal, is the windowed denoising signal, and the superscripts [-1], [0], and [1] represent the second spectrum analysis of the past frame and the first and second spectrum analyses of the current frame, respectively, and 128 is the number of sampling points.

[0079] Step S108 outputs the final clean radar signal, which is used for subsequent target detection, tracking and other processes.

[0080] The millimeter-wave radar signal processing method proposed in the embodiments of the present invention realizes a millimeter-wave radar noise suppression method based on spectral subtraction and overlapping addition. It is a systematic innovation in millimeter-wave radar noise suppression, which achieves effective noise suppression through the synergy of multiple methods, meeting real-time requirements and engineering feasibility.

[0081] The embodiments of this invention perform noise reduction preprocessing on noisy signals based on spectral subtraction and overlapping addition methods. Traditional spectral subtraction preprocessing of noisy signals easily generates "musical noise," which is essentially random narrowband noise after spectral subtraction. This leads to noise estimation errors and further causes spectral distortion of the denoised signal. To mitigate this impact, the embodiments of this invention employ a gain factor smoothing method. By smoothing the gain factor, estimation errors are effectively reduced, thus decreasing musical noise.

[0082] The embodiments of the present invention also use an overlapping addition method. By analyzing the spectrum of past frames and the current frame, noise is suppressed in the noisy signal. This method can reduce splicing distortion caused by frame processing and can efficiently process long signals, thereby effectively reconstructing the denoised signal.

[0083] The millimeter-wave radar signal processing method proposed in this invention achieves noise suppression of radar signals through a "time domain → frequency domain noise reduction → time domain reconstruction" link. Specifically, the "gain factor weighting" in the frequency domain solves the musical noise problem of traditional spectral subtraction; "overlapping addition" solves the splicing distortion problem of frame processing; and the feedback link realizes smooth iteration of gain, further improving the stability of noise reduction.

[0084] As can be seen from the above signal processing flow, the method proposed in the embodiments of the present invention only requires Fourier transform and spectral subtraction operation, with low computational load, no complex parameter update and iteration process, and extremely strong real-time performance, making it suitable for high-speed radar data processing flow.

[0085] On the other hand, embodiments of the present invention provide a millimeter-wave radar signal processing device, the device including a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, implements the millimeter-wave radar signal processing method described in any of the above embodiments.

[0086] In another aspect, embodiments of the present invention provide a storage medium storing computer-readable instructions that, when executed by a processor, perform the millimeter-wave radar signal processing method described in any of the above embodiments.

[0087] In another aspect, embodiments of the present invention provide a computer program product including computer-readable instructions that, when executed by a processor, perform the steps of the millimeter-wave radar signal processing method described in any of the above embodiments.

[0088] The millimeter-wave radar signal processing method, device, storage medium, and computer program products proposed in the embodiments of the present invention have very beneficial effects.

[0089] To address the issue that traditional spectral subtraction can easily lead to noise estimation bias and generate random "musical noise," the implementation of this invention uses a method of "SNR-driven dynamic gain + iterative smoothing" to suppress this defect at its source.

[0090] Dynamic gain can accurately match signal characteristics. Different gain values ​​are used for different signal-to-noise ratios, adapting to the dynamic changes in SNR in radar signals, while avoiding excessive noise estimation deviation.

[0091] Compared to the "uniform gain across the entire frequency band" or "fixed threshold gain" commonly used in existing technologies, the embodiments of this invention dynamically adjust the gain factor based on the real-time signal-to-noise ratio (SNR) of the millimeter-wave radar signal and introduce a smoothing factor to achieve dynamic inter-frame smoothing of the gain through smoothing iteration. Iterative smoothing can effectively solve the gain abrupt changes that may occur between multiple frames of signal, smooth out random fluctuations, avoid the generation of pulses, and improve the overall signal robustness.

[0092] In the embodiments of the present invention, the smoothing intensity is adaptively adjusted with SNR, avoiding the problems of "oversmoothing leading to signal blurring" or "insufficient smoothing leaving residual music noise".

[0093] This invention innovates at the signal reconstruction level by designing a multi-frame correlation and overlapping summation strategy, thus solving the problem of frame stitching distortion. The "square root + sine correction" design of the window function, compared to the traditional Hamming window, further reduces the abrupt changes in frame edge signals, preserving signal reliability. The square root smooths the attenuation curve at the window function edges, preventing excessive suppression of the original signal. The sine correction further smooths out any abrupt changes that may occur at frame edges, resolving distortion during frame stitching and providing higher reliability for the overlapping summation to restore the original signal.

[0094] Since millimeter-wave radar (such as automotive and security radar) requires real-time processing at the microsecond or millisecond level on embedded platforms (such as FPGAs and MCUs), the implementation method of this invention achieves both noise reduction and real-time performance through low-computational-load algorithm optimization and algorithm process simplification. Based on FPGA platform test results, the processing time for a single frame can be controlled within 1ms, which can meet the real-time requirement of automotive radar to process more than 1000 frames per second.

[0095] Reference is made to Figure 3 , which shows a time-domain signal simulation diagram according to an embodiment of the present application, wherein a time-domain signal simulation diagram of a pure signal, a noisy signal and a denoised signal is shown. The noisy signal is a random generated Gaussian noise signal added on the basis of the pure signal. After the millimeter wave radar signal processing technical solution proposed by the embodiment of the present application, the denoised signal is obtained. It can be seen that the denoised signal basically retains the details of the target signal, and provides a relatively pure signal for subsequent signal processing.

[0096] Reference is made to Figure 4A and Figure 4B , which respectively show a frequency spectrum diagram according to a traditional spectral subtraction and a millimeter wave radar signal processing method according to an embodiment of the present application. Wherein the original signal spectrum, the noisy signal spectrum and the denoised signal spectrum are included respectively, Figure 4A is denoising processing according to a traditional spectral subtraction, Figure 4B is denoising processing according to the signal processing method proposed by the embodiment of the present application. It can be seen from the frequency spectrum waveform that the signal processing method proposed by the embodiment of the present application better retains the characteristics and details of the normal signal in the original signal, and suppresses the abnormal signal.

[0097] The millimeter wave radar signal processing method, device, storage medium and computer program product proposed by the embodiment of the present application can be widely applied to millimeter wave radar signal processing, and are especially suitable for vehicle-mounted FMCW millimeter wave radar or security PMCW millimeter wave radar systems which need high real-time performance.

[0098] The foregoing description of the embodiments of the present application has been given for illustrative purposes only and is not exhaustive or limited to the precise forms disclosed. It will be understood by those skilled in the art that various changes can be made and equivalents substituted for elements without departing from the scope of the present application. In addition, many modifications can be made to adapt a particular situation or material to the teachings of the present application without departing from the basic scope thereof. Therefore, the present application is not intended to be limited to the specific embodiments disclosed as the best mode contemplated for carrying out the present application, and it will include all embodiments falling within the scope of the appended claims.

Claims

1. A millimeter-wave radar signal processing method, characterized in that, The method includes: The original time-domain signal of the millimeter-wave radar signal is framed and windowed to generate a windowed time-domain signal; Perform a fast Fourier transform on the windowed time-domain signal to generate an initial frequency-domain signal, which has initial spectral parameters; The dynamic gain factor of the current frame is determined based on the signal-to-noise ratio of the current frame in the initial frequency domain signal, wherein the dynamic gain factor of the current frame is positively correlated with the signal-to-noise ratio of the current frame. Determine the smoothing factor of the current frame based on the dynamic gain factor of the current frame; By using a smoothing factor and the smoothed gain factors of neighboring frames, the dynamic gain factor of the current frame is smoothed to generate the smoothed gain factor of the current frame. Generate the smoothed gain spectrum parameters of the current frame based on the smoothed gain factor and the initial spectrum parameters of the current frame. Perform an inverse fast Fourier transform on the frequency domain signal represented by the smoothed gain spectral parameters to obtain the initial denoised time domain signal; The frames of the initial denoised time-domain signal are windowed and overlapped to generate the final denoised time-domain signal.

2. The method according to claim 1, characterized in that, Determining the smoothing factor of the current frame based on the dynamic gain factor of the current frame includes: The smoothing factor of the current frame is determined based on the predefined maximum gain factor and the dynamic gain factor of the current frame.

3. The method according to claim 1, characterized in that, The value of the smoothing factor is negatively correlated with the value of the dynamic gain factor, and, Smoothing the dynamic gain factor of the current frame using a smoothing factor and smoothed gain factors from neighboring frames includes: The smoothed gain factors of neighboring frames and the dynamic gain factor of the current frame are weighted according to the smoothing factor. The weight of the smoothed gain factor of neighboring frames is positively correlated with the value of the smoothing factor, while the weight of the dynamic gain factor of the current frame is negatively correlated with the value of the smoothing factor.

4. The method according to claim 3, characterized in that, The smoothed gain factor for the current frame is generated according to the following formula: in, This represents the smoothed gain factor of the current frame. This represents the smoothing factor for the current frame. This represents the smoothed gain factor of the previous frame. This represents the dynamic gain factor for the current frame.

5. The method according to any one of claims 1-4, characterized in that, The dynamic gain factor for the current frame is determined based on the signal-to-noise ratio of the current frame in the initial frequency domain signal, including: When the signal-to-noise ratio of the current frame is less than or equal to a predefined first signal-to-noise ratio threshold, the dynamic gain factor of the current frame is determined to be the predefined first gain factor value. When the signal-to-noise ratio of the current frame is greater than or equal to a predefined second signal-to-noise ratio threshold, the dynamic gain factor of the current frame is determined to be the predefined second gain factor value. When the signal-to-noise ratio (SNR) of the current frame is greater than the first SNR threshold and less than the second SNR threshold, the dynamic gain factor of the current frame is determined to be a value that increases with the increase of the SNR, between the first gain factor value and the second gain factor value.

6. The method according to claim 5, characterized in that, When the signal-to-noise ratio (SNR) of the current frame is greater than the first SNR threshold and less than the second SNR threshold, the square of the dynamic gain factor of the current frame is linearly related to the SNR of the current frame.

7. The method according to any one of claims 1-4, characterized in that, The windowing of the original time-domain signal uses a 50% overlapping Hamming window.

8. A millimeter-wave radar signal processing device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, implements the millimeter-wave radar signal processing method according to any one of claims 1-7.

9. A storage medium, characterized in that, The device stores computer-readable instructions that, when executed by a processor, perform the millimeter-wave radar signal processing method according to any one of claims 1-7.

10. A computer program product comprising computer-readable instructions, characterized in that, When the instruction is executed by the processor, the steps of the millimeter-wave radar signal processing method according to any one of claims 1-7 are performed.

Images