Radar signal processing method and device, terminal equipment and storage medium

Abstract

The invention provides a radar signal processing method and device, terminal equipment and a storage medium, and the method comprises the steps: obtaining an amplitude vector and a phase vector of each received signal, and generating an angular frequency estimation value and an initial phase estimation value based on maximum likelihood estimation; calculating the GLRT statistical magnitude of the received signals according to the prior knowledge condition of the signal amplitude and the noise power, comparing the GLRT statistical magnitude with a preset threshold value, and generating a target detection result of each received signal; wherein the GLRT statistic is the ratio of the likelihood when the alternative hypothesis is established to the likelihood when the zero hypothesis is established; receiving signals with targets are extracted according to the target detection result, an effective detection signal index sequence is established, and the target distance and the target relative speed are generated according to the angular frequency estimation value and the initial phase estimation value of each receiving signal in the effective detection signal index sequence. According to the invention, the robustness and reliability of the radar system can be improved without improving the hardware level.

Description

Technical Field

[0001] This invention relates to the field of radar signal processing technology, and in particular to radar signal processing methods, apparatus, terminal equipment and storage media. Background Technology

[0002] Frequency-Modulated Continuous Wave (FMCW) radar is a radar system that achieves ranging and velocity measurement by transmitting linearly frequency-modulated signals, receiving target-reflected echoes, and obtaining intermediate frequency signals through mixing. With its core advantages of simple structure, low manufacturing cost, and high ranging and velocity measurement accuracy, it has been widely used in many key fields such as modern radar sensing, autonomous driving, intelligent healthcare, and industrial monitoring. Target detection, as the core function of FMCW radar, directly determines whether the radar system can accurately identify the presence of targets and extract key parameters. It is a crucial prerequisite for subsequent decision-making and control, and is of paramount importance to the reliability of the entire radar application system.

[0003] However, in practical applications, oscillators, as core components for radar signal transmission and processing, inherently exhibit phase instability, a phenomenon known as phase noise. The Wiener phase noise model accurately describes the phase noise behavior of many practical oscillators (such as semiconductor lasers and RF oscillators), representing a random walk process. In FMCW systems, phase noise causes random phase drift in the intermediate frequency signal, leading to problems like spectral broadening and decreased signal-to-noise ratio, resulting in reduced accuracy in range and velocity estimation and impacting system performance. Existing radar signal processing algorithms, such as constant false alarm rate detection based on Fast Fourier Transform (FFT) and two-dimensional Fourier Transform parameter estimation, are mostly designed under the ideal assumption of negligible phase noise. When phase noise exists in the system, these algorithms are significantly affected and struggle to achieve their performance. Although some hardware-level phase noise suppression techniques exist, such as adding auxiliary monitoring channels or using high-performance oscillators, these solutions often significantly increase system complexity and manufacturing costs, making them difficult to promote in low-cost, miniaturized applications. Improving the robustness and reliability of radar systems under non-ideal hardware conditions is a pressing technical problem. Summary of the Invention

[0004] The present invention aims to provide a radar signal processing method, apparatus, terminal equipment and storage medium to solve the above-mentioned technical problems and improve the robustness and reliability of radar systems without making improvements at the hardware level.

[0005] To address the aforementioned technical problems, this invention provides a radar signal processing method, comprising: Obtain the amplitude vector and phase vector of each received signal, and generate the angular frequency estimate and initial phase estimate of the received signal based on the maximum likelihood estimation; Using the amplitude vector and phase vector as input, the GLRT statistic for each received signal is calculated based on prior knowledge of signal amplitude and noise power. The GLRT statistic is then compared with a preset threshold, and a target detection result for each received signal is generated based on the comparison result. The GLRT statistic is the ratio of the likelihood when the alternative hypothesis is true to the likelihood when the null hypothesis is true. The null hypothesis characterizes the absence of a target in the first received signal, which contains only additive observation phase noise. The alternative hypothesis characterizes the presence of a target in the second received signal, which contains Wiener phase noise, additive observation phase noise, and distance and velocity information reflected from the target. Based on the target detection results, the received signals of the present target are extracted, and an effective detection signal index sequence is established. The target distance and target relative velocity are generated based on the angular frequency estimate and initial phase estimate of each received signal in the effective detection signal index sequence.

[0006] In the above scheme, by pre-defining null and alternative hypotheses, the signal characteristic differences in determining the presence or absence of a target are defined, providing a rigorous theoretical foundation for subsequent detection. By extracting amplitude and phase vectors, the multidimensional information of the signal is fully explored, and the angular frequency and initial phase estimate are generated by combining maximum likelihood estimation, ensuring the asymptotic optimality of parameter estimation. For different prior knowledge of signal amplitude and noise power, GLRT statistics are adaptively calculated and compared with preset thresholds, achieving accurate target detection in all scenarios. Finally, by filtering data through effective detection index sequences and fusing parameter estimation results from multiple effective signals, the target distance and relative velocity are calculated, further reducing noise interference. The entire scheme requires no additional hardware modification, but only explicitly incorporates Wiener phase noise statistical characteristics through algorithm optimization, breaking through the ideal assumption limitations of traditional algorithms. This avoids the increase in cost and complexity of hardware solutions and improves the robustness, reliability, and parameter estimation accuracy of the radar system under non-ideal conditions, making it widely adaptable to the high-precision target detection needs of autonomous driving, industrial monitoring, and other fields.

[0007] In one implementation, the amplitude vector and phase vector of each received signal are obtained, specifically including: The radar transmission signals transmitted sequentially within a frame and the corresponding received echo signals are mixed one by one to obtain several intermediate frequency signals. Each intermediate frequency (IF) signal is continuously sampled to obtain a complex baseband signal. The magnitude and phase of the complex baseband signal are extracted as the amplitude vector and phase vector of the received signal; the expressions for the amplitude vector and phase vector are as follows: ; ; In the formula, For the first An amplitude vector of the received signal; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; It is the transpose operator; For the first The phase vector of the received signal.

[0008] In the above scheme, the amplitude and phase vectors of the received signal are obtained through frequency mixing and continuous sampling, providing core input data for subsequent covariance matrix construction, parameter estimation, and GLRT statistics calculation. Frequency mixing transforms the high-frequency radar signal into an easily processed intermediate-frequency signal, reducing the difficulty of sampling and computation. The amplitude and phase vectors constructed according to the sampling order fully preserve the signal's time-domain characteristics and target information, ensuring that subsequent algorithms can fully utilize the signal's multi-dimensional information, thus laying a data foundation for improving the accuracy of target detection and parameter estimation.

[0009] In one implementation, before generating the angular frequency estimate and initial phase estimate of the received signal based on maximum likelihood estimation, the method further includes constructing the covariance matrix of the additive observation phase noise and Wiener phase noise, specifically: The additive observation phase noise vector in each received signal is determined based on the amplitude vector, and a first covariance matrix is ​​constructed based on the additive observation phase noise vector; wherein, the expression for the first covariance matrix is: ; In the formula, This is the first covariance matrix; Construct a function for a diagonal matrix; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; Noise power; A second covariance matrix is ​​constructed based on the Wiener phase noise vector corresponding to each received signal; the expression for the second covariance matrix is ​​as follows: ; In the formula, For the first Wiener phase noise vector of the received signal; These are the row and column element indices of the matrix, respectively. Let V be the variance of Wiener phase noise; Construct a third covariance matrix based on the first and second covariance matrices; the expression for the third covariance matrix is: ; In the formula, This is the third covariance matrix.

[0010] In the above scheme, the statistical characteristics of the two types of noise are comprehensively quantified by constructing a first covariance matrix of additive observation phase noise, a second covariance matrix of Wiener phase noise, and a third covariance matrix that fuses the two. The first covariance matrix, combined with the amplitude vector, reflects the fluctuation difference of additive noise; the second covariance matrix reflects the correlation of Wiener noise through random walk characteristics; and the third covariance matrix integrates the overall noise distribution of the signal, providing key weighting basis for subsequent weighted least squares estimation and GLRT statistics calculation. This enables parameter estimation and target detection to adaptively suppress noise interference, improving the stability of the algorithm in complex noise environments.

[0011] In one implementation, the angular frequency estimate and initial phase estimate of the received signal are generated based on maximum likelihood estimation, specifically including: Expressions for the angular frequency estimate and the initial phase estimate are constructed with weighted least squares as the objective; where the expression is: ; In the formula, For the first The estimated angular frequency of the received signal; For the first The initial phase estimate of the received signal; For the first The phase vector of the received signal; For the first The angular frequency of the received signal; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; For the first The initial phase of the received signal; It is a vector of all 1s. ; By maximizing the log-likelihood function to solve the given expression, the angular frequency estimate and the initial phase estimate are obtained; where the expressions for the angular frequency estimate and the initial phase estimate are: ; ; In the formula, It is the inverse of the third covariance matrix.

[0012] In the above scheme, the estimation of angular frequency and initial phase is transformed into a weighted least squares optimization problem. The inverse of the third covariance matrix is ​​used as the weights to highlight the effective signal components and suppress noise interference, ensuring the accuracy of the estimation results. A closed-form estimation expression is obtained by maximizing the log-likelihood function, avoiding complex iterative calculations and improving computational efficiency. This estimation method fully utilizes the phase information of the signal and the statistical characteristics of noise, achieving high-precision angular frequency and initial phase estimates even in scenarios with significant Wiener phase noise, providing reliable parameter support for subsequent target distance and velocity calculations.

[0013] In one implementation, the null hypothesis characterizes the absence of a target in the first received signal, which contains only additive observation phase noise. The alternative hypothesis characterizes the presence of a target in the second received signal, which contains Wiener phase noise, additive observation phase noise, and range and velocity information reflected from the target. Specifically: Define the phase vector expression under the null and alternative hypotheses; where the phase vector expression is: ; In the formula, For the first The phase vector of the received signal; For the first The additive observation phase noise vector corresponding to each received signal; This is the null hypothesis identifier; For the first The angular frequency of the received signal; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; For the first The initial phase of the received signal; It is a vector of all 1s. ; This is the Wiener phase noise vector; For alternative hypothesis identifiers; A composite hypothesis testing model is constructed based on the phase vector expressions of the null hypothesis and alternative hypotheses; the composite hypothesis testing model is used to characterize the phase observation results of whether the target exists or does not exist in the received signal.

[0014] In the above scheme, by clearly defining the phase vector expressions under the null and alternative hypotheses, the abstract judgment of whether a target exists or not is transformed into a specific difference in phase vector composition, giving the composite hypothesis testing model a clear mathematical foundation and operability. This expression precisely decomposes the additive observation phase noise, Wiener phase noise, target range-related angular frequency term, and velocity-related initial phase term, allowing the influence of different components on the received signal to be quantified. This ensures that the composite hypothesis testing model can accurately characterize the phase observation results of target existence or non-existence, providing a precise basis for subsequent GLRT statistic calculations and target detection, effectively improving the logical rigor and discriminative ability of the detection model.

[0015] In one implementation, based on prior knowledge of signal amplitude and noise power, using amplitude and phase vectors as inputs, a GLRT statistic for each received signal is generated. This GLRT statistic is then compared to a preset threshold, and a target detection result for each received signal is generated based on the comparison result. Specifically, this includes: When the signal amplitude and noise power are completely known a priori, the first GLRT statistic is calculated; where the expression for the first GLRT statistic is: ; In the formula, For the first An amplitude vector of the received signal; The signal amplitude; It is a vector of all 1s. ; Noise power; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; It is the inverse of the first covariance matrix; For the first The phase vector of the received signal; This is the third covariance matrix; This is the first covariance matrix; For the first The estimated angular frequency of the received signal; For the first The initial phase estimate of the received signal; It is the inverse of the third covariance matrix; When the signal amplitude is known and the noise power is unknown, calculate the maximum likelihood estimate of the noise power under the null hypothesis and the alternative hypothesis respectively, and select the minimum maximum likelihood estimate of the two as the noise power estimate. The first and third covariance matrices are reconstructed based on the noise power estimate. The second GLRT statistic is then calculated based on the reconstructed first and third covariance matrices and the noise power estimate. The expression for the second GLRT statistic is as follows: ; In the formula, To reconstruct the inverse of the first covariance matrix, reconstruct the first covariance matrix. ; To reconstruct the inverse of the third covariance matrix, reconstruct the third covariance matrix. ; This is the maximum likelihood estimate of the noise power under the null assumption. ; This represents the maximum likelihood estimate of the noise power under the alternative assumptions. ; When the signal amplitude is unknown and the noise power is known, the maximum likelihood estimate of the signal amplitude is calculated based on the observed data of the received signal; the expression for the maximum likelihood estimate of the signal amplitude is: ; In the formula, This is the maximum likelihood estimate of the signal amplitude; An index of the sampling time within a single received signal; For the first The received signal is at the first The complex baseband signal at the sampling time; The third GLRT statistic is calculated based on the maximum likelihood estimate of the signal amplitude; the expression for the third GLRT statistic is as follows: ; When both signal amplitude and noise power are unknown, the fourth GLRT statistic is calculated based on the reconstructed first covariance matrix, the reconstructed third covariance matrix, the noise power estimate, and the maximum likelihood estimate of the signal amplitude; the expression for the fourth GLRT statistic is as follows: ; Based on prior knowledge of signal amplitude and noise power, the corresponding test threshold is invoked. The GLRT statistic for each received signal is compared with the test threshold to generate a target detection result for each received signal. When the GLRT statistic is greater than the test threshold, it is determined that a target exists in the received signal.

[0016] In the above scheme, four adaptive GLRT statistic calculation schemes are designed for different prior knowledge conditions of signal amplitude and noise power, achieving accurate target detection with full scene coverage. The first GLRT statistic in a completely known scene fully utilizes prior parameters to improve detection accuracy; in partially known or completely unknown scenes, unknown parameters are estimated and the covariance matrix is ​​reconstructed to ensure that the statistic can still effectively distinguish between targets and noise; combined with a decision rule based on a preset threshold, the criteria for determining the presence of targets are clarified, controlling the false alarm probability while ensuring the detection probability, significantly improving the algorithm's adaptability and practicality.

[0017] In one implementation, the received signal indicating the presence of a target is extracted based on the target detection result, and an effective detection signal index sequence is established. The target range and target relative velocity are generated based on the estimated angular frequency and initial phase of each received signal in the effective detection signal index sequence. Specifically, this includes: Using frames as units, the estimated angular frequency and initial phase values ​​of each received signal within a frame are statistically analyzed to establish a sequence of angular frequency estimates and a sequence of initial phase estimates. An effective detection index sequence is constructed based on the target detection results of each received signal. Then, effective angular frequency estimates and effective initial phase estimates are extracted from the angular frequency estimate sequence and the initial phase estimate sequence based on the effective detection index sequence. Calculate the average of the effective angular frequency estimates, and then calculate the target distance based on the average; the expressions for the average and the target distance are as follows: ; ; In the formula, This represents the average of the estimated effective angular frequencies. To effectively detect index sequences; This is a sequence of estimated angular frequencies; It is a vector of all 1s. ; The target distance; The speed of light; The frequency modulation slope of the received signal; The average phase difference between adjacent elements in the effective initial phase estimate is calculated using a difference matrix, and the target relative velocity is calculated based on this average phase difference. The expressions for the average phase difference and the target relative velocity are as follows: ; ; In the formula, This represents the average phase difference. For one Matrix; The effective initial phase vector is obtained based on the effective initial phase estimate. For carrier frequency; The duration of a single received signal.

[0018] In the above scheme, signal data containing targets is filtered by effectively detecting index sequences, and interference data without targets is eliminated, thereby reducing the impact of noise on range and velocity calculations. The calculation of the average effective angular frequency suppresses random errors of individual signals, and the average phase difference captures the continuous characteristics of target motion through a difference matrix. Combined with the signal principles of FMCW radar, target range and relative velocity are derived, ensuring the accuracy and stability of the calculation results.

[0019] Secondly, this application also provides a radar signal processing device, including: a signal processing module, a target detection module, and an information generation module; The signal processing module is used to obtain the amplitude vector and phase vector of each received signal, and generate the angular frequency estimate and initial phase estimate of the received signal based on the maximum likelihood estimation. The target detection module takes the amplitude vector and phase vector as input, calculates the GLRT statistic for each received signal based on prior knowledge of signal amplitude and noise power, compares the GLRT statistic with a preset threshold, and generates a target detection result for each received signal based on the comparison result. The GLRT statistic is the ratio of the likelihood when the alternative hypothesis is true to the likelihood when the null hypothesis is true. The null hypothesis characterizes the absence of a target in the first received signal, which contains only additive observation phase noise. The alternative hypothesis characterizes the presence of a target in the second received signal, which contains Wiener phase noise, additive observation phase noise, and distance and velocity information reflected from the target. The information generation module is used to extract the received signals of the present target based on the target detection results, establish an index sequence of effective detection signals, and generate the target distance and target relative velocity based on the estimated angular frequency and initial phase of each received signal in the index sequence of effective detection signals.

[0020] Thirdly, this application also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the radar signal processing method described above.

[0021] Fourthly, this application also provides a computer-readable storage medium, which includes a stored computer program, wherein the computer program controls the device where the computer-readable storage medium is located to perform the radar signal processing method described above when it is running. Attached Figure Description

[0022] Figure 1This is a schematic flowchart of a radar signal processing method provided in one embodiment of the present invention; Figure 2 This is a schematic diagram of a radar signal processing device provided in one embodiment of the present invention. Detailed Implementation

[0023] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0024] The terms "first" and "second," etc., in the specification, claims, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such processes, methods, products, or apparatus.

[0025] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0026] Example 1 See Figure 1 , Figure 1 This is a flowchart illustrating a radar signal processing method according to an embodiment of the present invention. The embodiment includes steps 101 to 103, each step of which is detailed below: Step 101: Obtain the amplitude vector and phase vector of each received signal, and generate the angular frequency estimate and initial phase estimate of the received signal based on the maximum likelihood estimation.

[0027] In this embodiment of the invention, the amplitude vector and phase vector of each received signal are obtained. Based on the principle of maximum likelihood estimation, the angular frequency estimate and initial phase estimate of each signal are obtained by optimizing the objective function.

[0028] In one embodiment, obtaining the amplitude vector and phase vector of each received signal specifically includes: The radar transmission signals transmitted sequentially within a frame and the corresponding received echo signals are mixed one by one to obtain several intermediate frequency signals. Each intermediate frequency (IF) signal is continuously sampled to obtain a complex baseband signal. The magnitude and phase of the complex baseband signal are extracted as the amplitude vector and phase vector of the received signal; the expressions for the amplitude vector and phase vector are as follows: ; ; In the formula, For the first An amplitude vector of the received signal; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; It is the transpose operator; For the first The phase vector of the received signal.

[0029] The radar system sequentially transmits Q linear frequency modulated (LFM) signals within one frame. Each transmitted signal is reflected by the target to form a corresponding echo signal. The receiver performs frequency mixing on each transmitted signal and its corresponding echo signal one by one. In this embodiment of the invention, the transmitted signal is used as a reference and its frequency is superimposed with the echo signal carrying target information to obtain several intermediate frequency (IF) signals. The key function of frequency mixing is to reduce the signal frequency, avoiding high-frequency performance requirements on the hardware during subsequent sampling and processing, while retaining target range and velocity-related information. Each IF signal is continuously sampled by an ADC. The sampling process must be completed within the time domain for a duration of Tc, and a total of N consecutive sampling points are collected to finally obtain the complex baseband signal. Where k is the sampling time index. Complex baseband signals contain both amplitude and phase information, which can be extracted separately through mathematical operations: the magnitude of the complex baseband signal at each sampling point is calculated and arranged in sampling order to obtain the amplitude vector. After calculating the phase of the complex baseband signal at each sampling point, the phases of the N sampling points are arranged in the same sampling order to obtain the phase vector. .

[0030] For example, the radar system transmits Q=16 linear frequency modulated signals per frame, each with a duration Tc=80μs, a bandwidth B=500MHz, a carrier frequency fc=77GHz, and N=128 sampling points for each intermediate frequency signal (IF signal) (i.e., sampling times k=0,1,…,127), and a sampling frequency fs=10MHz. The radar sequentially transmits 16 chirp signals. The third transmitted signal (i=3) is reflected by the vehicle in front, forming an echo signal. The receiver inputs this transmitted signal and the corresponding echo signal into a mixer, and obtains the third IF signal through frequency aliasing. This signal retains the distance and speed information of the vehicle in front. The third IF signal is continuously sampled at 128 points to obtain complex baseband signals r3(0), r3(1),…, r3(127). The magnitude of each sampling point is calculated as follows: |r3(0)|=2.3V, |r3(1)|=2.28V, ..., |r3(127)|=2.32V. Arrange these values ​​in order to obtain the amplitude vector r3=[2.3,2.28,...,2.32]^T. The phase of each sampling point is calculated as follows: ∠r3(0)=0.12rad, ∠r3(1)=0.15rad, ..., ∠r3(127)=0.13rad. Arrange these values ​​in order to obtain the phase vector ∠r3=[0.12,0.15,...,0.13]^T. Both the final amplitude and phase vectors are 128-dimensional vectors.

[0031] In one embodiment, before generating the angular frequency estimate and initial phase estimate of the received signal based on the maximum likelihood estimation, the method further includes constructing a covariance matrix for the additive observation phase noise and the Wiener phase noise, specifically: The additive observation phase noise vector in each received signal is determined based on the amplitude vector, and a first covariance matrix is ​​constructed based on the additive observation phase noise vector; wherein, the expression for the first covariance matrix is: ; In the formula, This is the first covariance matrix; Construct a function for a diagonal matrix; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; Noise power; A second covariance matrix is ​​constructed based on the Wiener phase noise vector corresponding to each received signal; the expression for the second covariance matrix is ​​as follows: ; In the formula, For the first Wiener phase noise vector of the received signal; These are the row and column element indices of the matrix, respectively. Let V be the variance of Wiener phase noise; Construct a third covariance matrix based on the first and second covariance matrices; the expression for the third covariance matrix is: ; In the formula, This is the third covariance matrix.

[0032] In this embodiment of the invention, the first covariance matrix is ​​used to describe the statistical distribution characteristics of additive observation phase noise. Its core construction method utilizes the amplitude vector of the received signal to quantify the degree of noise fluctuation. Additive observation phase noise is a phase perturbation caused by additive white Gaussian noise (AWGN). Under high signal-to-noise ratio conditions, the noise at each sampling point can be modeled as having a mean of 0 and a variance of... The signal is a Gaussian random variable, and its noise variance is inversely proportional to the square of the received signal amplitude. That is, the larger the signal amplitude, the less noise interferes with the phase, and the smaller the variance. Therefore, based on the amplitude vector... The noise variance corresponding to each sampling point is ( These variances are then used as diagonal elements in the sampling order, and a diagonal matrix is ​​constructed. The first covariance matrix can then be constructed. The second covariance matrix is ​​used to characterize the statistical properties of Wiener phase noise. Its core function is to reflect the random walk nature of Wiener phase noise, and the phase jitter is cumulative, with correlations existing between noise samples from different times. Wiener phase noise vector. It follows a Gaussian distribution with a mean of 0, and its covariance matrix is... The elements are the minimum value of the sampling point indices m and n and the Wiener phase noise variance. The product of the two factors determines the phase noise. This matrix is ​​symmetric, with diagonal elements increasing with the index, reflecting the characteristic that the cumulative fluctuation of Wiener phase noise increases as the sampling time progresses. The off-diagonal elements are not zero, reflecting the correlation of noise at different times. The third covariance matrix is ​​a comprehensive statistical description of the overall phase noise of the received signal, used for weighted calculations in subsequent maximum likelihood estimation and generalized likelihood ratio tests. The phase noise of the received signal is composed of additive observation phase noise and Wiener phase noise, and the two types of noise are independent; therefore, their covariance matrix satisfies additivity. This matrix integrates the statistical characteristics of the two types of noise, reflecting both the independence of additive observation phase noise and the correlation of Wiener phase noise. It provides a precise weighting basis for subsequently quantifying the deviation between the observed and theoretical phase values, ensuring the reliability of the estimation and detection results.

[0033] In one embodiment, generating the angular frequency estimate and initial phase estimate of the received signal based on maximum likelihood estimation specifically includes: Expressions for the angular frequency estimate and the initial phase estimate are constructed with weighted least squares as the objective; where the expression is: ; In the formula, For the first The estimated angular frequency of the received signal; For the first The initial phase estimate of the received signal; For the first The phase vector of the received signal; For the first The angular frequency of the received signal; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; For the first The initial phase of the received signal; It is a vector of all 1s. ; By maximizing the log-likelihood function to solve the given expression, the angular frequency estimate and the initial phase estimate are obtained; where the expressions for the angular frequency estimate and the initial phase estimate are: ; ; In the formula, It is the inverse of the third covariance matrix.

[0034] In Wiener phase noise environment, the phase vector of the received signal Includes angular frequency Initial phase Wiener phase noise Additive observation phase noise In this embodiment of the invention, to accurately separate target information from noise, the goal is to minimize the error between the observed and theoretical phase values, and the error weight is determined by the inverse of the third covariance matrix of the received signal phase. The third covariance matrix is ​​the sum of the Wiener phase noise covariance matrix and the additive observation phase noise covariance matrix, and its inverse matrix reflects the reliability of the phase at different sampling points, allowing sampling points with high reliability to account for a higher proportion in the optimization. Assuming that both the Wiener phase noise and the additive observation phase noise follow a Gaussian distribution, maximizing the log-likelihood function of the received signal is equivalent to minimizing the weighted least squares problem. Therefore, maximizing the likelihood function means maximizing the probability of the observed data, corresponding to the case of minimum deviation, thus obtaining the expressions for the angular frequency estimate and the initial phase estimate.

[0035] Step 102: Using the amplitude vector and phase vector as input, calculate the GLRT statistic for each received signal based on prior knowledge of signal amplitude and noise power, and compare the GLRT statistic with a preset threshold. Generate the target detection result for each received signal based on the comparison result. The GLRT statistic is the ratio of the likelihood when the alternative hypothesis is true to the likelihood when the null hypothesis is true. The null hypothesis is used to characterize that the target does not exist in the first received signal and that the first received signal contains only additive observation phase noise. The alternative hypothesis is used to characterize that the target exists in the second received signal and that the second received signal contains Wiener phase noise, additive observation phase noise, and the distance and velocity information reflected by the target.

[0036] In this embodiment of the invention, a null hypothesis and alternative hypotheses are predefined, and the mathematical expressions for phase observations under different scenarios are clarified based on these two hypotheses. The engineering problem of whether a target exists is transformed into a quantifiable statistical decision problem, providing a unified logical framework for subsequent detection and estimation. The statistical characteristics of two types of phase noise are introduced, solving the performance degradation problem caused by the neglect of Wiener phase noise in traditional models and improving the model's adaptability to non-ideal environments. Then, using the amplitude vector and phase vector as input, based on prior knowledge of signal amplitude and noise power, the corresponding GLRT statistic is selected and compared with a preset threshold to determine whether a target exists in each signal, generating a detection result. Adaptively selecting statistics for different prior knowledge scenarios can balance detection efficiency and scenario adaptability.

[0037] In one embodiment, the null hypothesis is used to characterize the absence of a target in the first received signal, which contains only additive observation phase noise. The alternative hypothesis is used to characterize the presence of a target in the second received signal, which contains Wiener phase noise, additive observation phase noise, and range and velocity information reflected from the target. Specifically: Define the phase vector expression under the null and alternative hypotheses; where the phase vector expression is: ; In the formula, For the first The phase vector of the received signal; For the first The additive observation phase noise vector corresponding to each received signal; This is the null hypothesis identifier; For the first The angular frequency of the received signal; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; For the first The initial phase of the received signal; It is a vector of all 1s. ; This is the Wiener phase noise vector; For alternative hypothesis identifiers; A composite hypothesis testing model is constructed based on the phase vector expressions of the null hypothesis and alternative hypotheses; the composite hypothesis testing model is used to characterize the phase observation results of whether the target exists or does not exist in the received signal.

[0038] In this invention, the null hypothesis is defined as the absence of a target in the received signal. In this case, the phase of the received signal is only affected by additive observation phase noise, and there is no angular frequency or initial phase information due to target reflection, nor is it necessary to consider target-related phase change factors. The alternative hypothesis is defined as the presence of a target in the received signal. In this case, the phase of the received signal not only includes additive observation phase noise and Wiener phase noise caused by the oscillator's inherent phase jitter, but also carries the target's range and velocity information. It should be noted that this range and velocity information is represented by the angular frequency and initial phase of the received signal. These two hypotheses fully cover the possible scenarios of the received signal, providing a clear judgment benchmark for subsequent model construction and detection decisions.

[0039] In one embodiment, based on prior knowledge of signal amplitude and noise power, using amplitude vector and phase vector as input, a GLRT statistic for each received signal is generated. The GLRT statistic is then compared with a preset threshold, and a target detection result for each received signal is generated based on the comparison result. Specifically, this includes: When the signal amplitude and noise power are completely known a priori, the first GLRT statistic is calculated; where the expression for the first GLRT statistic is: ; In the formula, For the first An amplitude vector of the received signal; The signal amplitude; It is a vector of all 1s. ; Noise power; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; It is the inverse of the first covariance matrix; For the first The phase vector of the received signal; This is the third covariance matrix; This is the first covariance matrix; For the first The estimated angular frequency of the received signal; For the first The initial phase estimate of the received signal; It is the inverse of the third covariance matrix; When the signal amplitude is known and the noise power is unknown, calculate the maximum likelihood estimate of the noise power under the null hypothesis and the alternative hypothesis respectively, and select the minimum maximum likelihood estimate of the two as the noise power estimate. The first and third covariance matrices are reconstructed based on the noise power estimate. The second GLRT statistic is then calculated based on the reconstructed first and third covariance matrices and the noise power estimate. The expression for the second GLRT statistic is as follows: ; In the formula, To reconstruct the inverse of the first covariance matrix, reconstruct the first covariance matrix. ; To reconstruct the inverse of the third covariance matrix, reconstruct the third covariance matrix. ; This is the maximum likelihood estimate of the noise power under the null assumption. ; This represents the maximum likelihood estimate of the noise power under the alternative assumptions. ; When the signal amplitude is unknown and the noise power is known, the maximum likelihood estimate of the signal amplitude is calculated based on the observed data of the received signal; the expression for the maximum likelihood estimate of the signal amplitude is: ; In the formula, This is the maximum likelihood estimate of the signal amplitude; An index of the sampling time within a single received signal; For the first The received signal is at the first The complex baseband signal at the sampling time; The third GLRT statistic is calculated based on the maximum likelihood estimate of the signal amplitude; the expression for the third GLRT statistic is as follows: ; When both signal amplitude and noise power are unknown, the fourth GLRT statistic is calculated based on the reconstructed first covariance matrix, the reconstructed third covariance matrix, the noise power estimate, and the maximum likelihood estimate of the signal amplitude; the expression for the fourth GLRT statistic is as follows: ; Based on prior knowledge of signal amplitude and noise power, the corresponding test threshold is invoked. The GLRT statistic for each received signal is compared with the test threshold to generate a target detection result for each received signal. When the GLRT statistic is greater than the test threshold, it is determined that a target exists in the received signal.

[0040] In this embodiment of the invention, based on the prior known information of the signal amplitude and noise power, the calculation method of the GLRT statistic is adaptively selected. By quantifying the degree of fit between the received signal and the assumption of the presence / absence of the target, and combining it with a preset threshold, target detection is achieved. The core idea is to use the information of the amplitude vector and phase vector, and incorporate the covariance matrix characteristics of the two types of noise, so that the statistic can accurately distinguish between the target signal and the noise interference. (1) When the signal amplitude and noise power can be predetermined by system calibration (such as standard reflector calibration, noise floor measurement in a targetless scene), the known parameters are directly used for calculation. This statistic fully integrates effective information on amplitude and phase. The amplitude contribution of the target signal is quantified by the degree of matching between the amplitude vector and the known signal amplitude A. The larger the value, the better the amplitude dimension matches the assumption of the existence of the target. By combining the inverse of the covariance matrix of the additive observation phase noise, the effective signal components in the phase vector are highlighted, and additive noise interference is suppressed; The influence of Wiener phase noise on the covariance matrix is ​​demonstrated, and the deviation of the phase statistical distribution is corrected by the logarithm of the matrix determinant ratio. Based on the angular frequency estimate and initial phase estimate obtained from the maximum likelihood estimation, the weighted deviation between the phase observation and theoretical values ​​is calculated. The weight is the inverse of the third covariance matrix. The smaller the deviation, the better the phase dimension fits the assumption of the existence of the target. (2) When the signal amplitude is known and the noise power needs to be estimated from the observation data, the noise power estimation and covariance matrix reconstruction must be completed first, and then the statistics are calculated. The maximum likelihood estimate of the noise power under the null hypothesis and the alternative hypothesis is calculated respectively. and The smaller of the two values ​​is selected as the final noise power estimate to ensure conservatism. The first and third covariance matrices are reconstructed based on the noise power estimate to ensure that the matrices reflect the actual noise characteristics. A new expression is added to the expression of the second GLRT statistic. , indicating that the uncertainty of noise power estimation is incorporated, the quantization accuracy of the statistic is corrected, and the detection results are made more consistent with the actual noise environment. (3) When the noise power is known and the signal amplitude needs to be estimated from the observation data, the maximum likelihood estimate of the signal amplitude is obtained through the mean of the amplitude vector. The expression of the third GLRT statistic has been updated with The target contribution of the amplitude dimension is quantified by the maximum likelihood estimate of the signal amplitude, ensuring that amplitude information can be fully utilized and the discriminative power of the statistic can be maintained even when the signal amplitude is unknown. (4) When there is no prior information on both the signal amplitude and noise power, the two parameters are estimated and recalculated using the aforementioned method. For GLRT statistics under different prior knowledge scenarios, test thresholds are set accordingly. , , and The detection threshold is determined by the system's allowed false alarm probability. Confirmed. In radar signal detection, the false alarm probability... These are system-level parameters pre-set based on the security and reliability requirements of specific applications (e.g., when the Monte Carlo simulation count is 10,000, it is usually set to...). (Order level). Corresponding detection threshold. This is determined through theoretical analysis or Monte Carlo simulation, based on the principle that the GLRT statistic exceeds the detection threshold under noise-only conditions. The probability equals the false alarm probability. There is a clear negative correlation between the two. With fixed system parameters, the false alarm probability... The lower the setting, the lower the required detection threshold. The higher the false alarm probability, the better; if the system noise characteristics change, in order to maintain a constant false alarm probability... Detection threshold Adjustments are also necessary. The GLRT statistic for each received signal is compared to its corresponding threshold. If the statistic is greater than the threshold, it indicates that the received signal better fits the candidate hypothesis of target existence, and the target is determined to exist. If the statistic is less than or equal to the threshold, it indicates that it better fits the null hypothesis of target non-existence, and the target is determined to not exist. Finally, a corresponding target detection result is generated for each received signal.

[0041] Step 103: Extract the received signals containing the target based on the target detection results, establish an effective detection signal index sequence, and generate the target distance and target relative velocity based on the angular frequency estimate and initial phase estimate of each received signal in the effective detection signal index sequence.

[0042] In this embodiment of the invention, valid signals containing targets are selected based on the detection results, an index sequence is established, the average angular frequency is calculated using the estimated angular frequency of the valid signals, and the target distance is then derived; the relative velocity of the target is derived using the differential operation of the initial phase of the valid signals.

[0043] In one embodiment, the received signals indicating the presence of a target are extracted based on the target detection results, and an effective detection signal index sequence is established. The target distance and target relative velocity are generated based on the estimated angular frequency and initial phase of each received signal in the effective detection signal index sequence. Specifically, this includes: Using frames as units, the estimated angular frequency and initial phase values ​​of each received signal within a frame are statistically analyzed to establish a sequence of angular frequency estimates and a sequence of initial phase estimates. An effective detection index sequence is constructed based on the target detection results of each received signal. Then, effective angular frequency estimates and effective initial phase estimates are extracted from the angular frequency estimate sequence and the initial phase estimate sequence based on the effective detection index sequence. Calculate the average of the effective angular frequency estimates, and then calculate the target distance based on the average; the expressions for the average and the target distance are as follows: ; ; In the formula, This represents the average of the estimated effective angular frequencies. To effectively detect index sequences; This is a sequence of estimated angular frequencies; It is a vector of all 1s. ; The target distance; The speed of light; The frequency modulation slope of the received signal; The average phase difference between adjacent elements in the effective initial phase estimate is calculated using a difference matrix, and the target relative velocity is calculated based on this average phase difference. The expressions for the average phase difference and the target relative velocity are as follows: ; ; In the formula, This represents the average phase difference. For one Matrix; The effective initial phase vector is obtained based on the effective initial phase estimate. For carrier frequency; The duration of a single received signal.

[0044] In this embodiment of the invention, the parameter estimation results of all received signals within a single frame are integrated. Using a single frame of radar-transmitted signal as a processing unit, for each received signal, the estimated angular frequency and initial phase values ​​are arranged sequentially according to the order of the received signals to construct an angular frequency estimation sequence. and initial phase estimate sequence This ensures a one-to-one correspondence between each estimated value and its corresponding received signal. Based on the target detection result of each received signal, a detection label is assigned to each signal: 1 for a detected target, and 0 for no detected target. An effective detection index sequence is then constructed based on the labeling results. Based on this index sequence, in the angular frequency estimation sequence Extract all The corresponding elements form a set of effective angular frequency estimates; simultaneously, in the initial phase estimate sequence... Extract all The corresponding elements form a valid initial phase vector. ( , (This is the index of the effective signal), ensuring that subsequent calculations are based only on effective data containing the target. Then, the average of the effective angular frequency estimates is calculated. Among these... To effectively detect the dot product of the index sequence and the angular frequency estimate sequence, effective angular frequency estimates are filtered and accumulated. This average value effectively suppresses noise interference from individual signals and better reflects the angular frequency characteristics of the real target. Then, based on the range-angular frequency relationship of the FMCW radar, the target range is calculated using a formula. This formula essentially uses the angular frequency to infer the frequency difference between the target's reflected echo and the transmitted signal, thus converting it into range. Furthermore, the target's relative velocity is calculated using the phase change of continuous effective signals, combined with the Doppler effect principle. The core is capturing the cumulative phase change over time. A difference matrix D is constructed, which is a (K-1)×K matrix, with each row containing only one -1 and one 1, and the remaining elements being 0. For example, when K=3, D=[[-1,1,0],[0,-1,1]], which is used to calculate the effective initial phase vector. The phase difference between adjacent elements. The average phase difference is calculated using the formula for the average phase difference; where We obtain a (K-1) dimension vector composed of adjacent phase differences. This involves summing up all phase differences. Then, the target relative velocity is calculated using the expression for the target relative velocity. This formula uses the phase change rate to inversely deduce the Doppler frequency shift caused by the target motion, and then converts it into relative velocity.

[0045] In this embodiment of the invention, a radar signal processing device is also provided, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the radar signal processing method described above.

[0046] In this embodiment of the invention, a computer-readable storage medium is also provided, which includes a stored computer program, wherein the computer program controls the device where the computer-readable storage medium is located to execute the above-described radar signal processing method when it is running.

[0047] For example, a computer program can be divided into one or more modules, one or more of which are stored in memory and executed by a processor to perform the present invention. The one or more modules can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in a radar signal processing device.

[0048] Radar signal processing equipment can be computing devices such as desktop computers, laptops, handheld computers, and cloud servers. Radar signal processing equipment may include, but is not limited to, processors, memory, and displays. Those skilled in the art will understand that the above-mentioned components are merely examples of radar signal processing equipment and do not constitute a limitation on the radar signal processing equipment. It may include more or fewer components than those listed above, or a combination of certain components, or different components. For example, radar signal processing equipment may also include input / output devices, network access devices, buses, etc.

[0049] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the radar signal processing equipment, connecting all parts of the equipment through various interfaces and lines.

[0050] The memory can be used to store computer programs and / or modules. The processor implements various functions of the radar signal processing equipment by running or executing the computer programs and / or modules stored in the memory, and by calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area. The program storage area can store the operating system, application programs required for at least one function (such as sound playback function, text conversion function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, text message data, etc.). In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0051] In this invention, if the radar signal processing module is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. Those skilled in the art can understand and implement this invention without any inventive effort.

[0052] This invention provides a radar signal processing method. By pre-defining null and alternative hypotheses, it defines the signal characteristic differences between the presence and absence of a target, providing a rigorous theoretical basis for subsequent detection. By extracting amplitude and phase vectors, it fully mines the multidimensional information of the signal and combines maximum likelihood estimation to generate angular frequency and initial phase estimates, ensuring the asymptotic optimality of parameter estimation. For different prior knowledge situations regarding signal amplitude and noise power, it adaptively calculates GLRT statistics and compares them with preset thresholds, achieving accurate target detection across all scenarios. Finally, by filtering data through an effective detection index sequence and fusing parameter estimation results from multiple effective signals, it calculates the target distance and relative velocity, further reducing noise interference. The entire solution requires no additional hardware modification; it only explicitly incorporates Wiener phase noise statistical characteristics through algorithm optimization, breaking through the ideal assumption limitations of traditional algorithms. This avoids increased cost and complexity of hardware solutions while improving the robustness, reliability, and parameter estimation accuracy of the radar system under non-ideal conditions, making it widely adaptable to the high-precision target detection needs of various fields such as autonomous driving and industrial monitoring.

[0053] Example 2 See Figure 2 , Figure 2 This is a schematic diagram of a radar signal processing device according to an embodiment of the present invention. The radar signal processing device provided in this embodiment includes: a signal processing module 201, a target detection module 202, and an information generation module 203; The signal processing module 201 is used to acquire the amplitude vector and phase vector of each received signal, and generate the angular frequency estimate and initial phase estimate of the received signal based on the maximum likelihood estimation. The target detection module 202 is used to calculate the GLRT statistic for each received signal based on prior knowledge of signal amplitude and noise power, using amplitude vector and phase vector as inputs, and compare the GLRT statistic with a preset threshold. Based on the comparison result, it generates a target detection result for each received signal. The GLRT statistic is the ratio of the likelihood when the alternative hypothesis is true to the likelihood when the null hypothesis is true. The null hypothesis is used to characterize that the target does not exist in the first received signal and that the first received signal contains only additive observation phase noise. The alternative hypothesis is used to characterize that the target exists in the second received signal and that the second received signal contains Wiener phase noise, additive observation phase noise, and distance and velocity information of the target reflection. The information generation module 203 is used to extract the received signals of the present target based on the target detection results, establish an effective detection signal index sequence, and generate the target distance and target relative velocity based on the angular frequency estimate and initial phase estimate of each received signal in the effective detection signal index sequence.

[0054] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0055] This invention provides a radar signal processing device. By pre-defining null and alternative hypotheses, it defines the signal characteristic differences between the presence and absence of a target, providing a rigorous theoretical basis for subsequent detection. By extracting amplitude and phase vectors, it fully mines the multidimensional information of the signal and combines maximum likelihood estimation to generate angular frequency and initial phase estimates, ensuring the asymptotic optimality of parameter estimation. For different prior knowledge situations regarding signal amplitude and noise power, it adaptively calculates GLRT statistics and compares them with preset thresholds, achieving accurate target detection across all scenarios. Finally, by filtering data through an effective detection index sequence and fusing parameter estimation results from multiple effective signals, it calculates the target distance and relative velocity, further reducing noise interference. The entire solution requires no additional hardware modification; it only explicitly incorporates Wiener phase noise statistical characteristics through algorithm optimization, breaking through the ideal assumption limitations of traditional algorithms. This avoids increasing the cost and complexity of hardware solutions while improving the robustness, reliability, and parameter estimation accuracy of the radar system under non-ideal conditions. It can be widely adapted to the high-precision target detection needs of various fields such as autonomous driving and industrial monitoring.

[0056] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make several improvements and substitutions without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention.

Claims

1. A radar signal processing method, characterized in that, include: Obtain the amplitude vector and phase vector of each received signal, and generate the angular frequency estimate and initial phase estimate of the received signal based on the maximum likelihood estimation; Using the amplitude vector and the phase vector as input, the GLRT statistic for each received signal is calculated based on prior knowledge of signal amplitude and noise power. The GLRT statistic is then compared with a preset threshold, and a target detection result for each received signal is generated based on the comparison result. The GLRT statistic is the ratio of the likelihood when the alternative hypothesis is true to the likelihood when the null hypothesis is true. The null hypothesis characterizes the absence of a target in the first received signal, which contains only additive observation phase noise. The alternative hypothesis characterizes the presence of a target in the second received signal, which contains Wiener phase noise, additive observation phase noise, and distance and velocity information of the target reflection. Based on the target detection results, the received signals of the present target are extracted, an effective detection signal index sequence is established, and the target distance and target relative velocity are generated based on the angular frequency estimate and initial phase estimate of each received signal in the effective detection signal index sequence.

2. The radar signal processing method as described in claim 1, characterized in that, The acquisition of the amplitude vector and phase vector of each received signal specifically includes: The radar transmission signals transmitted sequentially within a frame and the corresponding received echo signals are mixed one by one to obtain several intermediate frequency signals. Each intermediate frequency signal is continuously sampled to obtain a complex baseband signal. The magnitude and phase of the complex baseband signal are extracted as the amplitude vector and phase vector of the received signal; wherein, the expressions for the amplitude vector and the phase vector are: ； ； In the formula, For the first An amplitude vector of the received signal; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; It is the transpose operator; For the first The phase vector of the received signal.

3. The radar signal processing method as described in claim 1, characterized in that, Before generating the angular frequency estimate and initial phase estimate of the received signal based on maximum likelihood estimation, the method further includes constructing the covariance matrix of the additive observation phase noise and the Wiener phase noise, specifically: The additive observation phase noise vector in each received signal is determined based on the amplitude vector, and a first covariance matrix is ​​constructed based on the additive observation phase noise vector; wherein, the expression for the first covariance matrix is: ； In the formula, This is the first covariance matrix; Construct a function for a diagonal matrix; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; Noise power; A second covariance matrix is ​​constructed based on the Wiener phase noise vector corresponding to each received signal; wherein the expression for the second covariance matrix is: ； In the formula, For the first Wiener phase noise vector of the received signal; These are the row and column element indices of the matrix, respectively. Let V be the variance of Wiener phase noise; A third covariance matrix is ​​constructed based on the first and second covariance matrices; wherein the expression for the third covariance matrix is: ； In the formula, This is the third covariance matrix.

4. The radar signal processing method as described in claim 3, characterized in that, The process of generating the angular frequency estimate and initial phase estimate of the received signal based on maximum likelihood estimation specifically includes: Expressions for the angular frequency estimate and the initial phase estimate are constructed with weighted least squares as the objective; wherein, the expressions are: ； In the formula, For the first The estimated angular frequency of the received signal; For the first The initial phase estimate of the received signal; For the first The phase vector of the received signal; For the first The angular frequency of the received signal; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; For the first The initial phase of the received signal; It is a vector of all 1s. ; The estimated angular frequency and the estimated initial phase are obtained by solving the expression based on the maximization of the log-likelihood function; wherein, the expressions for the estimated angular frequency and the estimated initial phase are: ； ； In the formula, It is the inverse of the third covariance matrix.

5. The radar signal processing method as described in claim 3, characterized in that, The null hypothesis characterizes the absence of a target in the first received signal, which contains only additive observation phase noise. The alternative hypothesis characterizes the presence of a target in the second received signal, which contains Wiener phase noise, additive observation phase noise, and range and velocity information reflected from the target. Specifically: Define the phase vector expression under the null hypothesis and the alternative hypothesis; wherein, the phase vector expression is: ； In the formula, For the first The phase vector of the received signal; For the first The additive observation phase noise vector corresponding to each received signal; This is the null hypothesis identifier; For the first The angular frequency of the received signal; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; For the first The initial phase of the received signal; It is a vector of all 1s. ; This is the Wiener phase noise vector; For alternative hypothesis identifiers; A composite hypothesis testing model is constructed based on the phase vector expressions of the null hypothesis and the alternative hypotheses; wherein, the composite hypothesis testing model is used to characterize the phase observation results of whether the target exists or does not exist in the received signal.

6. The radar signal processing method as described in claim 5, characterized in that, Based on prior knowledge of signal amplitude and noise power, and using the amplitude vector and phase vector as inputs, a GLRT statistic for each received signal is generated. This GLRT statistic is then compared with a preset threshold, and a target detection result for each received signal is generated based on the comparison result. Specifically, this includes: When the signal amplitude and noise power are completely known a priori, the first GLRT statistic is calculated; wherein, the expression for the first GLRT statistic is: ; In the formula, For the first An amplitude vector of the received signal; The signal amplitude; It is a vector of all 1s. ; Noise power; This represents the number of consecutive sampling points for a single intermediate frequency signal. ; It is the inverse of the first covariance matrix; For the first The phase vector of the received signal; This is the third covariance matrix; This is the first covariance matrix; For the first The estimated angular frequency of the received signal; For the first The initial phase estimate of the received signal; It is the inverse of the third covariance matrix; When the signal amplitude is known and the noise power is unknown, calculate the maximum likelihood estimate of the noise power under the null hypothesis and the alternative hypothesis respectively, and select the minimum maximum likelihood estimate of the two as the noise power estimate. Based on the noise power estimate, the first covariance matrix and the third covariance matrix are reconstructed. Based on the obtained reconstructed first covariance matrix, reconstructed third covariance matrix, and the noise power estimate, a second GLRT statistic is calculated. The expression for the second GLRT statistic is: ； In the formula, To reconstruct the inverse of the first covariance matrix, reconstruct the first covariance matrix. ; To reconstruct the inverse of the third covariance matrix, reconstruct the third covariance matrix. ; This is the maximum likelihood estimate of the noise power under the null assumption. ; This represents the maximum likelihood estimate of the noise power under the alternative assumptions. ; When the signal amplitude is unknown and the noise power is known, the maximum likelihood estimate of the signal amplitude is calculated based on the observation data of the received signal; wherein, the expression for the maximum likelihood estimate of the signal amplitude is: ； In the formula, This is the maximum likelihood estimate of the signal amplitude; An index of the sampling time within a single received signal; For the first The received signal is at the first The complex baseband signal at the sampling time; The third GLRT statistic is calculated based on the maximum likelihood estimate of the signal amplitude; wherein, the expression for the third GLRT statistic is: ； When both the signal amplitude and the noise power are unknown, a fourth GLRT statistic is calculated based on the reconstructed first covariance matrix, the reconstructed third covariance matrix, the noise power estimate, and the maximum likelihood estimate of the signal amplitude; wherein, the expression for the fourth GLRT statistic is: ； Based on prior knowledge of the signal amplitude and noise power, a corresponding test threshold is invoked, and the GLRT statistic corresponding to each received signal is compared with the test threshold to generate a target detection result for each received signal; wherein, when the GLRT statistic is greater than the test threshold, it is determined that a target exists in the received signal.

7. The radar signal processing method as described in claim 1, characterized in that, The step of extracting the received signal containing the target based on the target detection result, establishing an effective detection signal index sequence, and generating the target distance and target relative velocity based on the angular frequency estimate and initial phase estimate of each received signal in the effective detection signal index sequence specifically includes: Using frames as units, the estimated angular frequency and initial phase values ​​of each received signal within a frame are statistically analyzed to establish a sequence of angular frequency estimates and a sequence of initial phase estimates. An effective detection index sequence is constructed based on the target detection result of each received signal, and effective angular frequency estimates and effective initial phase estimates are extracted from the angular frequency estimate sequence and the initial phase estimate sequence based on the effective detection index sequence. Calculate the average value of the effective angular frequency estimates, and calculate the target distance based on the average value; wherein, the expressions for the average value and the target distance are: ； ； In the formula, This represents the average of the estimated effective angular frequencies. To effectively detect index sequences; This is a sequence of estimated angular frequencies; It is a vector of all 1s. ; The target distance; The speed of light; The frequency modulation slope of the received signal; The average phase difference between adjacent elements in the effective initial phase estimate is calculated using a difference matrix, and the target relative velocity is calculated based on the average phase difference; wherein, the expressions for the average phase difference and the target relative velocity are: ； ； In the formula, This represents the average phase difference. For one Matrix; The effective initial phase vector is obtained based on the effective initial phase estimate. For carrier frequency; The duration of a single received signal.

8. A radar signal processing device, characterized in that, include: Signal processing module, target detection module, and information generation module; The signal processing module is used to obtain the amplitude vector and phase vector of each received signal, and generate the angular frequency estimate and initial phase estimate of the received signal based on the maximum likelihood estimation. The target detection module is used to calculate the GLRT statistic for each received signal based on prior knowledge of signal amplitude and noise power, using the amplitude vector and phase vector as inputs. The GLRT statistic is then compared with a preset threshold, and a target detection result is generated for each received signal based on the comparison result. The GLRT statistic is the ratio of the likelihood when the alternative hypothesis is true to the likelihood when the null hypothesis is true. The null hypothesis characterizes the absence of a target in the first received signal, which contains only additive observation phase noise. The alternative hypothesis characterizes the presence of a target in the second received signal, which contains Wiener phase noise, additive observation phase noise, and distance and velocity information reflected from the target. The information generation module is used to extract the received signals of the presence of the target based on the target detection results, establish an effective detection signal index sequence, and generate the target distance and target relative velocity based on the angular frequency estimate and initial phase estimate of each received signal in the effective detection signal index sequence.

9. A terminal device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the radar signal processing method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the radar signal processing method as described in any one of claims 1 to 7.

Images