94Ghz radar baseband signal resolving method and system

Abstract

The invention belongs to the technical field of radar signal processing, and provides a 94Ghz radar baseband signal resolving method and system, comprising clock signal synchronization, signal receiving and preprocessing, analog-to-digital conversion, multistage parallel signal processing, dynamic parameter adjustment, multi-target resolving and dynamic resource allocation. According to the invention, by introducing an FPGA multi-stage parallel processing architecture, the multi-target detection and tracking capability of the system in a complex scene is improved; signal processing parameters are dynamically optimized according to a real-time environment through a deep reinforcement learning algorithm, and the accurate detection capability of the system in different scenes is ensured; the real-time tracking and resolving capabilities of the system on multiple targets are enhanced through a graph neural network multi-target cooperative processing algorithm; dynamic resource allocation is realized through an energy sensing algorithm, and the power consumption of the system is greatly reduced while high-speed processing is ensured.

Description

Technical Field

[0001] The present invention relates to the field of radar signal processing technology, and in particular to a 94Ghz radar baseband signal solution method and system. Background Art

[0002] With the widespread application of millimeter-wave radar technology in intelligent transportation, autonomous driving and other fields, the demand for radar signal processing in complex environments is growing. Millimeter-wave radar systems face multiple technical challenges when processing signals in multi-target tracking and dynamic environments.

[0003] In complex scenes, radar signals are easily reflected and scattered by environmental obstacles, resulting in multipath effects. These interferences cause signal attenuation, affect target detection accuracy, and reduce system stability. At the same time, noise interference also increases system errors. Existing radar systems often have difficulty processing multiple targets simultaneously due to limited computing power when detecting multiple targets and tracking high-speed moving targets. The sequential processing architecture limits the real-time performance of the system in complex scenes, especially in applications that require rapid response, such as autonomous driving. Existing technologies show lag and are difficult to respond efficiently. In addition, existing millimeter-wave radar systems lack adaptive capabilities and usually rely on fixed signal processing processes that cannot be dynamically adjusted according to changes in the environment. This fixed processing method limits the flexibility of the system in different scenarios and cannot cope with the diverse needs in complex scenarios. In summary, the existing technology has technical deficiencies in signal attenuation, real-time detection of multiple targets, high-speed target tracking, and adaptive capabilities. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a 94Ghz radar baseband signal solution method and system, which introduces an FPGA multi-level parallel processing architecture to improve the system's multi-target detection and tracking capabilities in complex scenarios; uses a deep reinforcement learning algorithm to dynamically optimize signal processing parameters according to the real-time environment to ensure the system's accurate detection capabilities in different scenarios; uses a graph neural network multi-target collaborative processing algorithm to enhance the system's real-time tracking and solution capabilities for multiple targets; and uses an energy-aware algorithm to achieve dynamic resource allocation, which greatly reduces the system's power consumption while ensuring high-speed processing.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A 94Ghz radar baseband signal solution method, comprising:

[0007] Use the system clock and synchronization module to provide synchronous clock signals for the RF front-end module, analog-to-digital converter, FPGA processing module, deep learning optimization module, target detection and solution module, target detection and solution module, and data transmission and control module;

[0008] Utilize the RF front-end module to receive a 94 GHz millimeter wave signal, and perform mixing, amplification, noise reduction, and analog-to-digital conversion on the 94 GHz millimeter wave signal to obtain an intermediate frequency signal;

[0009] Converting the intermediate frequency signal into a digital signal using the analog-to-digital converter;

[0010] The FPGA processing module is used to perform noise filtering, normalization and multi-level parallel signal processing on the digital signal to obtain a multi-target difference comparison heat map;

[0011] Dynamically adjust the FFT window, filter bandwidth, and gain control parameters of the FPGA processing module using the deep reinforcement learning algorithm of the deep learning optimization module;

[0012] Utilizing the multi-target collaborative processing algorithm based on graph neural network in the target detection and solution module to detect and solve the distance, speed and angle of multiple targets according to the multi-target difference comparison heat map, a multi-target solution map is obtained; the multi-target solution map includes: distance difference, speed difference and angle difference;

[0013] The importance function and resource allocation optimization function of the energy sensing algorithm in the data transmission and control module are used to dynamically allocate the computing resources of the FPGA processing module, the deep learning optimization module and the target detection and solution module, and the multi-objective solution graph is uploaded to the target host computer.

[0014] Preferably, the construction process of the FPGA processing module includes:

[0015] The state-action value function of the deep reinforcement learning algorithm is used to extract the Doppler frequency shift characteristics, echo intensity characteristics and noise characteristics of the pre-collected intermediate frequency signal to obtain a feature data set; the calculation formula of the state-action value function is: Among them, Q(s,a) is the value of taking action a in state s; r is the immediate reward; γ is the discount factor; s' represents the state after action a; a' represents the action in state s'; Represents state s'

[0016] Next, we select the expected value of the action a' with the largest value;

[0017] Performing dimensionality reduction processing on the feature data set to obtain a training data set;

[0018] Based on the fast Fourier transform and multi-objective collaborative processing algorithm, the FPGA processing module is iterated by using a preset loss function according to the training data set; the calculation formula of the loss function is: Where L is the total loss; R i and are the true distance and predicted distance of the i-th target respectively; V j and are the actual speed and predicted speed of the jth target respectively; λ is the weight factor; N is the number of targets used for distance calculation; M is the number of targets used for speed calculation.

[0019] Preferably, the noise filtering process uses an autocorrelation function and a low-pass filter; the calculation formula of the autocorrelation function is: The calculation formula of the low-pass filter is:

[0020]

[0021] Among them, R xx (τ) is the autocorrelation function; H(f) is the low-pass filter; t is the integral variable; x(t) is the target filter signal; τ is the time constant; f is the signal frequency; f c is the filter cutoff frequency.

[0022] Preferably, the calculation function of the normalization process is: Among them, x norm (t) is the normalized signal; y(t) is the target normalized signal; μ is the signal mean; σ is the signal standard deviation.

[0023] Preferably, the calculation formula of the multi-objective solution graph includes:

[0024] as well as

[0025] Where R is the distance difference; V target is the speed difference; θ is the angle difference; c is the speed of light; T s is the signal scanning period; Δf is the frequency difference; B w is the radar signal bandwidth; Λ is the radar wavelength; Δf doppler is the Doppler frequency shift; Δφ is the phase difference; d is the antenna unit spacing.

[0026] Preferably, the importance function is: The resource allocation optimization function is:

[0027]

[0028] Among them, w i is the importance weight of the i-th target; R alloc is the optimal resource allocation scheme; R i and V i are the distance feature and speed feature of the i-th target respectively; Rj and V j are the distance feature and speed feature of the jth target respectively; α and β are weight factors; U(R i ) is the utility function of the ith objective; Represents finding the optimal solution of the distance difference.

[0029] Preferably, a 94Ghz radar target detection and resolution module comprises:

[0030] A receiving unit, configured to receive a 94 GHz millimeter wave signal, and perform mixing, amplification, noise reduction, and analog-to-digital conversion on the 94 GHz millimeter wave signal to obtain an intermediate frequency signal;

[0031] A conversion unit, used for converting the intermediate frequency signal into the digital signal;

[0032] A parallel processing unit, used for performing noise filtering, normalization and multi-level parallel signal processing on the digital signal to obtain a multi-target difference comparison thermal map;

[0033] A calculation unit, used for detecting and calculating the distance, speed and angle of multiple targets according to the multi-target difference comparison heat map, to obtain a multi-target calculation map;

[0034] An optimization unit, used for dynamically adjusting the FFT window, filter bandwidth and gain control parameters of the parallel processing unit;

[0035] A control unit, used to dynamically adjust the computing resources of the parallel processing unit, the solving unit and the optimization unit by using the importance function and the resource allocation optimization function, and upload the multi-objective solving graph to a target host computer;

[0036] A synchronization unit is used to provide a synchronization clock signal for the receiving unit, the conversion unit, the parallel processing unit, the solving unit, the optimization unit and the control unit.

[0037] The present invention discloses the following technical effects:

[0038] The present invention provides a 94Ghz radar baseband signal solution method and system. By introducing an FPGA multi-level parallel processing architecture, the defect of poor multi-target detection effect in the prior art is solved, and multi-target detection and tracking functions in complex scenarios are realized; through a deep reinforcement learning algorithm, the problem of poor generalization and reliability caused by fixed parameters of the FPGA processing module is solved, and dynamic optimization of signal processing parameters according to the real-time environment is realized; through a graph neural network multi-target collaborative processing algorithm, the problem of poor solution ability of conventional technologies is solved, and real-time tracking and solution functions for multiple targets are realized; through an energy sensing algorithm, the problem of low efficiency and high power consumption caused by fixed computing power allocation in conventional technologies is solved, and dynamic resource allocation is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0040] Figure 1 A schematic diagram of a 94 GHz radar baseband signal solution process according to an embodiment of the present invention;

[0041] Figure 2 A schematic diagram of the overall architecture of a 94GHz millimeter-wave radar baseband signal solution system based on FPGA technology provided in an embodiment of the present invention;

[0042] Figure 3 A multi-objective difference comparison heat map provided by an embodiment of the present invention;

[0043] Figure 4 A schematic diagram of a signal parallel processing flow provided by an embodiment of the present invention;

[0044] Figure 5 A signal processing optimization flow chart provided for an embodiment of the present invention;

[0045] Figure 6 A multi-objective collaborative processing flow chart provided for an embodiment of the present invention;

[0046] Figure 7 A target correlation analysis radar chart provided by an embodiment of the present invention;

[0047] Figure 8 A schematic diagram of dynamic resource allocation provided by an embodiment of the present invention;

[0048] Fig. 9 A target importance ranking diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] The purpose of the present invention is to provide a 94Ghz radar baseband signal solution method and system, which improves the system's multi-target detection and tracking capabilities in complex scenarios by introducing an FPGA multi-level parallel processing architecture; dynamically optimizes signal processing parameters according to the real-time environment through a deep reinforcement learning algorithm to ensure the system's accurate detection capabilities in different scenarios; enhances the system's real-time tracking and solution capabilities for multiple targets through a graph neural network multi-target collaborative processing algorithm; and realizes dynamic resource allocation through an energy sensing algorithm to significantly reduce the system's power consumption while ensuring high-speed processing.

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] Figure 1 A schematic diagram of a 94 GHz radar baseband signal solution process according to an embodiment of the present invention is shown in FIG. Figure 1 As shown, the present invention provides a 94Ghz radar baseband signal solution method, comprising:

[0053] Step 100: Use the system clock and synchronization module to provide a synchronous clock signal for the RF front-end module, the analog-to-digital converter, the FPGA processing module, the deep learning optimization module, the target detection and solution module, the target detection and solution module, and the data transmission and control module;

[0054] Step 200: Receive a 94 GHz millimeter wave signal using the RF front-end module, and perform mixing, amplification, noise reduction, and analog-to-digital conversion on the 94 GHz millimeter wave signal to obtain an intermediate frequency signal;

[0055] Step 300: Convert the intermediate frequency signal into a digital signal using the analog-to-digital converter;

[0056] Step 400: using the FPGA processing module to perform noise filtering, normalization, and multi-level parallel signal processing on the digital signal to obtain a multi-target difference comparison heat map;

[0057] Step 500: Dynamically adjust the FFT window, filter bandwidth, and gain control parameters of the FPGA processing module using the deep reinforcement learning algorithm of the deep learning optimization module;

[0058] Step 600: Detect and calculate the distance, speed and angle of multiple targets according to the multi-target difference comparison heat map using the multi-target collaborative processing algorithm based on graph neural network in the target detection and calculation module to obtain a multi-target calculation map; the multi-target calculation map includes: distance difference, speed difference and angle difference;

[0059] Step 700: Utilize the importance function and resource allocation optimization function of the energy sensing algorithm in the data transmission and control module to dynamically allocate computing resources of the FPGA processing module, the deep learning optimization module, and the target detection and solution module, and upload the multi-target solution graph to the target host computer (uploaded in the form of UDP packets).

[0060] Specifically, the construction process of the FPGA processing module includes:

[0061] The state-action value function of the deep reinforcement learning algorithm is used to extract the Doppler frequency shift characteristics, echo intensity characteristics and noise characteristics of the pre-collected intermediate frequency signal to obtain a feature data set; the calculation formula of the state-action value function is: Among them, Q(s,a) is the value of taking action a in state s; r is the immediate reward; γ is the discount factor; s' represents the state after action a; a' represents the action in state s'; Represents state s'

[0062] Next, we select the expected value of the action a' with the largest value;

[0063] Performing dimensionality reduction processing on the feature data set to obtain a training data set;

[0064] Based on the fast Fourier transform and multi-objective collaborative processing algorithm, the FPGA processing module is iterated by using a preset loss function according to the training data set; the calculation formula of the loss function is: Where L is the total loss; R i and are the true distance and predicted distance of the i-th target respectively; V j and are the actual speed and predicted speed of the jth target respectively; λ is the weight factor; N is the number of targets used for distance calculation; M is the number of targets used for speed calculation.

[0065] Specifically, the state-action value function Q(s,a): Q(s,a) represents the expected value of taking action a in the current state s. It measures the value of taking an action by the accumulated rewards to optimize the signal processing strategy. Immediate reward r: r is the direct return obtained after taking an action in the current state, reflecting the immediate effect of the action, and is used to guide the learning direction of the deep reinforcement learning model. Discount factor γ: γ is a value between 0 and 1, which is used to discount future rewards. It determines the importance of the model to long-term benefits. A larger γ value means that the model pays more attention to the future cumulative benefits, which is suitable for complex signal processing scenarios that require long-term adjustments. Future state s′: s′ represents the state after the current action a is applied, and the model optimizes the current strategy by predicting the subsequent state. Future action a′: a′ is the action that may be taken in the future state s′. The model optimizes the decision of the current action by predicting the value of future actions. Maximize maxQ(s′,a′): This term represents the expected value of selecting the action a′ with the largest value in the future state s′, which is used to update the action value of the current state. It guides the model to choose the best action at each step to improve the cumulative reward.

[0066] Furthermore, the total loss L is the target value of model training, and the prediction accuracy of the model is improved by minimizing this value. It combines the errors of the two indicators of distance and speed to achieve the high-precision requirements for multi-objective solution. The actual distance R i and predicted distance is the true distance to the target, is the distance of the i-th target predicted by the model. The square of the difference between the two Used to calculate the error of distance prediction. True speed V j and predicted speed V j ^\hat: V j is the true velocity of the jth target, V j ^\hat is the jth target speed predicted by the model. The square of the difference between the two Used to calculate the error of speed prediction. The weight factor λ is a weight factor used to adjust the proportion of distance and speed error terms in the total loss. By adjusting λ, the model's attention to distance prediction error and speed prediction error can be changed to adapt to different detection accuracy requirements. Number of targets N and M: N represents the number of detected targets (related to distance calculation), while M represents the number of detected targets (related to speed calculation). They are used to determine the summation range of each type of error in the loss function.

[0067] Preferably, the noise filtering process uses an autocorrelation function and a low-pass filter; the calculation formula of the autocorrelation function is: The calculation formula of the low-pass filter is:

[0068]

[0069] Among them, R xx (τ) is the autocorrelation function; H(f) is the low-pass filter; t is the integral variable; x(t) is the target filter signal; τ is the time constant; f is the signal frequency; f c is the filter cutoff frequency. By adjusting f c , high-frequency noise can be filtered out to obtain a purer echo signal.

[0070] Specifically, the calculation function of the normalization process is: Among them, x norm (t) is the normalized signal; y(t) is the target normalized signal; μ is the signal mean; σ is the signal standard deviation.

[0071] Preferably, the calculation formula of the multi-objective solution graph includes:

[0072] as well as

[0073] Where R is the distance difference; V target is the speed difference; θ is the angle difference; c is the speed of light; T s is the signal scanning period; Δf is the frequency difference, that is, the frequency difference between the radar transmitting signal and the echo signal; B w is the radar signal bandwidth; Λ is the radar wavelength, defined as Where f0 is the center frequency of the radar transmission (for the 94 GHz band, it is approximately 3.19×10 -3 m); Δf doppler is the Doppler shift, which is obtained by measuring the frequency offset of the echo signal; Δφ is the phase difference; and d is the antenna unit spacing. The distance difference formula calculates the frequency difference between the transmitted signal and the echo signal to accurately calculate the distance of the target. This algorithm is used to eliminate the speed ambiguity problem caused by the Doppler effect and ensure that the radar system accurately calculates the target speed in high-speed moving scenarios. The angle difference formula measures the phase difference of the echo signals received by adjacent antenna units in the phased array antenna array to calculate the relative angle between the target and the radar, ensuring high-precision calculation of the target's azimuth.

[0074] Specifically, the importance function is: The resource allocation optimization function is:

[0075]

[0076] Among them, w i is the importance weight of the i-th target; R alloc is the optimal resource allocation scheme; R iand V i are the distance feature and speed feature of the i-th target respectively; R j and V j are the distance feature and speed feature of the jth target respectively; α and β are weight factors used to adjust the influence of different features on importance; U(R i ) is the utility function of the ith objective; Represents finding the optimal solution of the distance difference.

[0077] Furthermore, a 94Ghz radar target detection and resolution module includes:

[0078] A receiving unit, configured to receive a 94 GHz millimeter wave signal, and perform mixing, amplification, noise reduction, and analog-to-digital conversion on the 94 GHz millimeter wave signal to obtain an intermediate frequency signal;

[0079] A conversion unit, used for converting the intermediate frequency signal into the digital signal;

[0080] A parallel processing unit, used for performing noise filtering, normalization and multi-level parallel signal processing on the digital signal to obtain a multi-target difference comparison thermal map;

[0081] A calculation unit, used for detecting and calculating the distance, speed and angle of multiple targets according to the multi-target difference comparison heat map, to obtain a multi-target calculation map;

[0082] An optimization unit, used for dynamically adjusting the FFT window, filter bandwidth and gain control parameters of the parallel processing unit;

[0083] A control unit, used to dynamically adjust the computing resources of the parallel processing unit, the solving unit and the optimization unit by using the importance function and the resource allocation optimization function, and upload the multi-objective solving graph to a target host computer;

[0084] A synchronization unit is used to provide a synchronization clock signal for the receiving unit, the conversion unit, the parallel processing unit, the solving unit, the optimization unit and the control unit.

[0085] Preferably, the method for generating a multi-objective solution graph comprises:

[0086] Multiple target feature parameters are processed in pairs to obtain interactive information such as distance, speed difference, angle change, etc. between two targets;

[0087] Utilizing the multi-objective solution model, calculating the mutual influence between various objectives;

[0088] Generate a multi-target solution diagram based on the calculation results, and optimize the target tracking and signal processing strategies in real time.

[0089] Specifically, the fast Fourier transform (FFT) formula used in the signal solution process is:

[0090]

[0091] Where: FFT(f) is the frequency spectrum of the echo signal in the frequency domain. This formula is used to convert the time domain signal into a frequency domain signal, which is used to further calculate the distance, speed and angle of the target.

[0092] Optionally, the autocorrelation function is often used for time domain preprocessing of signals. It can analyze the correlation between signals and noise and perform preliminary denoising at the early stage of the signal processing process. Low-pass filters (LPFs) are mainly used in the frequency domain processing stage, especially when specific frequency noise needs to be filtered out. It is a way to accurately process signals by adjusting the frequency. In complex signal processing, the sources of noise can be very diverse. In order to obtain a clearer signal, a combination of various technical means is usually used to gradually reduce the interference of noise on the signal. The autocorrelation function captures the noise characteristics in the time domain and initially removes it, while the low-pass filter further refines the removal of high-frequency noise in the frequency domain. This combination ensures that the system can optimize signal quality at different levels.

[0093] refer to Figure 2 , the specific functions of each module are as follows:

[0094] RF front-end module: This module is used to receive 94GHz millimeter wave signals. Through mixing, amplification and noise reduction, it is converted into an intermediate frequency signal. This step provides a stable signal input source for the subsequent digital processing of the system.

[0095] Analog-to-digital converter (ADC): Converts analog intermediate frequency signals into digital signals. The ADC ensures that the analog signals from the RF front-end module are digitized, providing a basis for subsequent precise calculations by the FPGA processing module. The ADC has high sampling rate and low power consumption, supporting real-time data acquisition and processing.

[0096] FPGA processing module: This is the core processing unit of the system, responsible for performing multi-level parallel signal processing.

[0097] Target detection and calculation module: This module performs target detection tasks by receiving data from the FPGA processing module. It calculates the distance, speed and angle of multiple targets in real time and analyzes the target's motion trajectory. Figure 3The target detection and solution module can generate a multi-target difference comparison heat map based on the differences in distance, speed and angle of different targets. This map shows the differences between multiple targets in different parameters, provides an intuitive comparison reference for the system's multi-target solution, and helps to identify the difference characteristics between targets. The module is embedded with a real-time multi-target solution process, which can analyze and process the mutual influence of multiple targets.

[0098] Deep Learning Optimization Module: Through the deep reinforcement learning (DRL) algorithm, this module dynamically adjusts key parameters such as FFT window, filter bandwidth and gain control.

[0099] Data transmission and control module: This module is responsible for managing the data transmission of the entire system and the communication between modules, ensuring the normal operation and efficient collaboration of all modules. This module uses a high-speed data bus and low-latency communication protocol to ensure the real-time and high efficiency of the radar signal processing process. The data transmission and control module is also responsible for the interface connection with the external system to realize the integration and sharing of multiple sensor data.

[0100] System clock and synchronization module: This module provides a synchronous clock signal for the entire system, ensuring that all modules work in coordination under a unified time base. This module achieves efficient coordination of the entire radar system through a high-precision clock source and clock synchronization algorithm. This module also supports time synchronization across multiple radar devices and is suitable for multi-radar coordination scenarios.

[0101] refer to Figure 3, the target comparison heat map shows the comparative differences between multiple targets in three features: distance difference, speed difference, and angle difference. The values ​​and color depths in the heat map reflect the size of the feature differences between the targets. The darker the color, the greater the difference, and the lighter the color, the smaller the difference. The following is a detailed description of the content in the figure: Targets A, B, C, D: represent different targets detected in the system. In radar detection, the differences between these targets are quantified by the characteristic values ​​of distance, speed, and angle. Distance difference, speed difference, angle difference: represent the distance difference, speed difference, and angle difference between different targets, respectively. These features are used to evaluate and compare the relative position and motion state between each target. Target pairs such as AB, AC: Each cell of the heat map represents the difference between two targets in a certain feature. Specifically: Distance difference row: represents the distance difference between each target pair. For example, the distance difference between target A and target B is 4, and the distance difference between target A and target C is 12. Speed ​​difference row: represents the speed difference between each target pair. For example, the speed difference between target A and target B is 4, and the speed difference between target A and target C is 2. Angle difference row: Indicates the angle difference between each target pair. For example, the angle difference between target A and target C is 25, and the angle difference between target A and target D is 15. Color description: The darker the color, the greater the difference and the larger the value; the lighter the color, the smaller the difference. The darker the color, the more obvious the difference between the corresponding target pairs in this feature. For example, the angle difference between target A and target C is 25, which is the darkest color, indicating that the angle difference is large. Overall analysis: The angle difference between targets A and C is the largest (25), followed by the distance difference (12), indicating that the two targets are significantly different in relative orientation and position. The speed difference between each target pair is generally small, and the values ​​are all low, indicating that their relative motion states are relatively close. This graphical representation method facilitates quick observation of the feature differences between different targets, and can help the radar system give priority to target pairs with significant differences when solving targets, thereby optimizing resource allocation and processing strategies.

[0102] refer to Figure 4 ,The signal parallel processing flow in the FPGA processing module specifically includes the following steps:

[0103] S1: Reception and preprocessing of intermediate frequency signals. The RF front-end module receives 94GHz millimeter wave signals and generates intermediate frequency signals. The analog signal is converted into a digital signal through an analog-to-digital converter (ADC) and enters the FPGA processing module for parallel processing.

[0104] S2: Fast Fourier Transform (FFT). The FPGA processing module first performs a fast Fourier transform (FFT) on the input digital signal to convert the time domain signal into a frequency domain signal. The size of the FFT window is dynamically adjusted by the deep learning optimization module to adapt to the signal characteristics and noise levels in different environments, ensuring the accuracy and efficiency of signal processing.

[0105] S3: Filtering and gain control. The signal after FFT conversion is filtered through a bandpass filter to remove noise and unnecessary frequency band signals. The system further enhances the signal strength through gain control to ensure the clarity and resolution accuracy of the target signal.

[0106] S4: Target detection and parameter calculation. Further, the filtered signal is transmitted to the target detection and calculation module to extract the target parameters such as distance, speed and angle. Through the multi-target collaborative processing algorithm based on graph neural network (GNN), the system can analyze the correlation information of multiple targets, generate a multi-target calculation graph, and realize real-time target tracking and prediction.

[0107] Further, beamforming: Beamforming is a signal processing method based on phased array technology, which is mainly used in angle channels. By adjusting the signal phase and amplitude of each antenna unit, beamforming technology can enhance the signal strength in a specific direction while suppressing interference signals in other directions. This technology is used in this system to improve the angular resolution of the target, so that the system can more accurately locate the target's position. The result of beamforming can effectively improve the detection performance of the system in a multi-target environment. DOA (Direction of Arrival) angle of arrival algorithm: The DOA angle of arrival algorithm is used to determine the direction of signal arrival. This system receives data through multiple channels, and uses the DOA algorithm to accurately calculate the angle information of the target signal based on the signal phase difference of each channel and the beamforming result. The core of this algorithm is to calculate the phase difference received by multiple antennas, thereby inferring the incident angle of the signal. The DOA algorithm is used in conjunction with beamforming to further improve the accuracy of angle measurement. Doppler analysis: Doppler analysis is used for signal processing in the velocity channel, mainly by detecting the Doppler frequency shift to determine the relative velocity of the target. Doppler shift is the frequency offset caused by the relative motion of the target. The system extracts the frequency shift information through time domain and frequency domain processing techniques and calculates the speed of the target. The results of Doppler analysis are further processed by the speed threshold filter to filter out the target signals that meet the speed detection requirements. Array calibration: Array calibration is used to ensure the accuracy of the phased array antenna. This process involves correcting the phase and amplitude differences of each unit in the antenna array to ensure consistency in signal processing. Array calibration is a basic step to ensure the accuracy of beamforming and DOA algorithms, and can maintain the reliability of signal processing when hardware deviations or environmental interference exist. Kalman filter: Kalman filter is used in the target summary module to filter and track the multi-target fused signal. It is a recursive algorithm that can combine historical states and current observations to predict the target's motion trajectory and filter out noise interference. The application of Kalman filter ensures the stability of distance, speed and angle output, enabling the system to achieve accurate tracking in complex multi-target scenarios.

[0108] refer to Figure 5 , the process of adaptive optimization of signal processing parameters through deep reinforcement learning (DRL) algorithm is as follows:

[0109] S1: The system collects intermediate frequency signals in real time through the FPGA processing module and extracts key signal features from them, such as Doppler frequency shift, echo intensity, etc. The system can effectively capture the real-time dynamic changes of signals, improve the accuracy and response speed of signal processing, and ensure that signal features can be captured in a timely manner even in complex and changing environments.

[0110] S2: Establish a historical database. The system uses a large amount of historical data to train the deep reinforcement learning model. Specifically, the model continuously learns the historical feature changes of the signal and the external environment information, and dynamically adjusts the FFT window, filter bandwidth, and gain control parameters to optimize the signal processing process. The training process uses simulated signal data for different scenarios to ensure that the model has a wide range of adaptability.

[0111] S3: The system adjusts the FFT window size, filter bandwidth, and gain parameters in real time based on the output of the deep reinforcement learning model. By updating these parameters in real time, the system can adaptively process signals in different scenarios to ensure that the optimized signal processing effect reaches the optimal state. During the target detection process, the system can make detailed parameter adjustments for different frequency bands and signal strengths to improve detection accuracy in complex environments.

[0112] S4: The system uses a feedback mechanism to detect the effect of signal processing in real time and returns the processing results to the deep reinforcement learning model for optimization. The model continuously adjusts the processing strategy based on new feedback to further enhance the model's adaptive optimization capabilities. In this way, the system ensures high accuracy and stability of signal processing, reduces errors, and improves the overall system response speed.

[0113] refer to Figure 6 ,The operation process of the multi-objective collaborative processing algorithm based on graph neural network (GNN) is as follows:

[0114] S1: The system obtains preliminary solution data of multiple targets from the FPGA processing module, including basic information such as the target's distance, speed, and angle. These data are analyzed to generate a preliminary multi-target solution graph. The system can extract more accurate correlation information based on the spatiotemporal dependencies between different targets. The model learns the dynamic correlation between multiple targets, captures the spatiotemporal dependencies between targets, and predicts the future motion trajectory of the target.

[0115] S2: The graph neural network analyzes the correlation information of multiple targets, predicts the future motion trajectories of multiple targets, and generates an optimized multi-target solution graph through model optimization. This optimized graph combines the correlation of multiple targets with historical trajectories to improve the accuracy of target detection and solution.

[0116] S3: Based on the optimized multi-target solution graph, the system realizes real-time tracking and motion prediction of multiple targets. The system ensures that multiple targets can be tracked with high accuracy and low latency in complex environments, and is particularly suitable for multi-target tracking and behavior prediction in complex traffic scenarios. Figure 7 ,The system also generates a target correlation analysis diagram through ,datadriven analysis, which shows the correlation relationship among multiple ,targets, providing more accurate data support for the prediction of the ,target’s motion trajectory.

[0117] refer to Figure 8 , the dynamic optimization allocation of FPGA resources is realized through the energy-aware algorithm, which specifically includes the following steps:

[0118] S1: Load monitoring and signal acquisition. The system monitors the current load status in real time and detects the computing pressure and signal processing requirements of the FPGA module. By monitoring the FPGA load, the system accurately evaluates the current resource usage.

[0119] S2: Execution of energy-aware algorithms. The system executes energy-aware algorithms based on real-time load data to evaluate the energy consumption and computing requirements of each processing module. Through energy consumption evaluation, the system predicts the energy consumption of each processing module under the current load and allocates resources according to actual needs.

[0120] S3: Optimize resource allocation. Based on the evaluation results of the energy-aware algorithm, the system optimizes the allocation of resources to the FPGA processing module. By reducing the resource usage of low-priority tasks, it ensures that the resource requirements of high-priority tasks are fully met, especially when the system is under high load, it can still efficiently perform important tasks. Fig. 9 , based on the data processing results of the system, an importance ranking diagram of the target is generated. This diagram analyzes the importance of multiple targets, displays the priority of the targets, and provides a reference for resource allocation and decision-making.

[0121] S4: Energy consumption optimization feedback. The system continuously adjusts the resource allocation strategy through the feedback mechanism to ensure that the energy consumption optimization effect continues to improve. By optimizing the energy consumption of the system in real time, the system can significantly reduce energy consumption while meeting computing needs, improve overall energy efficiency, and is suitable for energy-constrained application scenarios to solve the problem of uneven energy consumption distribution at the same time.

[0122] Furthermore, the energy detection module is used to monitor the power, signal strength and spectrum occupancy of the input signal. This module can analyze the characteristics of the signal so that the subsequent resource allocation module can appropriately allocate signals of different frequency bands and power levels. The resource allocation module performs dynamic resource allocation based on the signal characteristics output by the energy detection module. This module optimizes the allocation of spectrum based on the power and strength characteristics of the signal to ensure the rational use of system resources. Spectrum allocation is one of the key functions of the resource allocation module. It dynamically allocates different frequency resources based on the spectrum occupancy information obtained by energy detection to ensure that the bandwidth requirements of the signal during transmission are met and reduce signal interference and spectrum overlap. Power control is used to adjust the power size of the signal transmission to ensure that the signal achieves a balance between coverage and power consumption. Power control can adaptively adjust the transmission power according to the target distance and signal environment, reduce power consumption and maintain the transmission quality of the signal. Time slot control is another key function in the resource allocation module. Time slot control avoids multiple signals from being transmitted at the same time by allocating different time intervals (time slots) to each signal, thereby reducing signal interference. This time slot scheduling mechanism is particularly important for multi-target detection and multi-signal processing environments. It can ensure that the system's signal processing timing is carried out in an orderly manner, improving the system's real-time performance and stability. The final output signal is a signal that has been optimized in terms of spectrum, power, and time slot. Through these controls and allocations, the system can ensure that the signal is transmitted in the best state, effectively cope with complex signal environments, and achieve stable signal output.

[0123] Preferably, the allocation of dynamic resource allocation module. Target detection and solution module: In the radar system, the target detection and solution module is a part with high computing requirements. The dynamic resource allocation module needs to ensure that the module obtains sufficient resources when the computational complexity is high, especially in the multi-target detection scenario, where the computational pressure is high. FPGA processing module: As the core processing unit, FPGA can adjust the resource occupancy of different signal processing stages under high load by dynamically allocating resources, such as filtering, gain control, FFT processing, etc. Deep learning optimization module: This module uses the DRL algorithm to adaptively optimize signal processing parameters. Dynamic resource allocation ensures that the deep learning optimization module can prioritize sufficient FPGA computing power to support real-time optimization when a large amount of computing resources is required. Feedback mechanism: Dynamic resource allocation not only adjusts resources according to the order of target importance, but also provides feedback according to the real-time load status of the system. By real-time monitoring of the resource usage of each module (such as FPGA load), the feedback path will feed this data back to the resource allocation module to optimize the subsequent resource scheduling. The system can dynamically adjust the signal processing process through the feedback mechanism to cope with different scene changes.

[0124] The beneficial effects of the present invention are as follows:

[0125] The present invention improves the system's multi-target detection and tracking capabilities in complex scenarios by introducing an FPGA multi-level parallel processing architecture; dynamically optimizes signal processing parameters according to the real-time environment through a deep reinforcement learning algorithm, ensuring the system's accurate detection capabilities in different scenarios; enhances the system's real-time tracking and solving capabilities for multiple targets through a graph neural network multi-target collaborative processing algorithm; and realizes dynamic resource allocation through an energy-aware algorithm, which significantly reduces the system's power consumption while ensuring high-speed processing.

[0126] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0127] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A 94Ghz radar baseband signal solution method, characterized in that: include: Use the system clock and synchronization module to provide synchronous clock signals for the RF front-end module, analog-to-digital converter, FPGA processing module, deep learning optimization module, target detection and solution module, target detection and solution module, and data transmission and control module; Utilize the RF front-end module to receive a 94 GHz millimeter wave signal, and perform mixing, amplification, noise reduction, and analog-to-digital conversion on the 94 GHz millimeter wave signal to obtain an intermediate frequency signal; Converting the intermediate frequency signal into a digital signal using the analog-to-digital converter; The FPGA processing module is used to perform noise filtering, normalization and multi-level parallel signal processing on the digital signal to obtain a multi-target difference comparison heat map; Dynamically adjust the FFT window, filter bandwidth, and gain control parameters of the FPGA processing module using the deep reinforcement learning algorithm of the deep learning optimization module; Utilizing the multi-target collaborative processing algorithm based on graph neural network in the target detection and solution module to detect and solve the distance, speed and angle of multiple targets according to the multi-target difference comparison heat map, a multi-target solution map is obtained; the multi-target solution map includes: distance difference, speed difference and angle difference; The importance function and resource allocation optimization function of the energy sensing algorithm in the data transmission and control module are used to dynamically allocate the computing resources of the FPGA processing module, the deep learning optimization module and the target detection and solution module, and the multi-objective solution graph is uploaded to the target host computer.

2. A 94Ghz radar baseband signal solution method according to claim 1, characterized in that: The construction process of the FPGA processing module includes: The state-action value function of the deep reinforcement learning algorithm is used to extract the Doppler frequency shift characteristics, echo intensity characteristics and noise characteristics of the pre-collected intermediate frequency signal to obtain a feature data set; the calculation formula of the state-action value function is: Among them, Q(s,a) is the value of taking action a in state s; r is the immediate reward; γ is the discount factor; s' represents the state after action a; a' represents the action in state s'; represents the expected value of selecting the action a' with the greatest value in state s'; Performing dimensionality reduction processing on the feature data set to obtain a training data set; Based on the fast Fourier transform and multi-objective collaborative processing algorithm, the FPGA processing module is iterated by using a preset loss function according to the training data set; the calculation formula of the loss function is: Where L is the total loss; R i and are the true distance and predicted distance of the i-th target respectively; V j and are the actual speed and predicted speed of the jth target respectively; λ is the weight factor; N is the number of targets used for distance calculation; M is the number of targets used for speed calculation.

3. A 94Ghz radar baseband signal solution method according to claim 1, characterized in that: The noise filtering process uses an autocorrelation function and a low-pass filter; the calculation formula of the autocorrelation function is: The calculation formula of the low-pass filter is: Among them, R xx (τ) is the autocorrelation function; H(f) is the low-pass filter; t is the integral variable; x(t) is the target filter signal; τ is the time constant; f is the signal frequency; f c is the filter cutoff frequency.

4. A 94 GHz radar baseband signal solution method according to claim 1, characterized in that: The calculation function of the normalization process is: Among them, x norm (t) is the normalized signal; y(t) is the target normalized signal; μ is the signal mean; σ is the signal standard deviation.

5. A 94 GHz radar baseband signal solution method according to claim 2, characterized in that: The calculation formula of the multi-objective solution graph includes: as well as Where R is the distance difference; V target is the speed difference; θ is the angle difference; c is the speed of light; T s is the signal scanning period; Δf is the frequency difference; B w is the radar signal bandwidth; Λ is the radar wavelength; Δf doppler is the Doppler frequency shift; Δφ is the phase difference; d is the antenna unit spacing.

6. A 94 GHz radar baseband signal solution method according to claim 5, characterized in that: The importance function is: The resource allocation optimization function is: Among them, w i is the importance weight of the i-th target; R alloc is the optimal resource allocation scheme; R i and V i are the distance feature and speed feature of the i-th target respectively; R j and V j are the distance feature and speed feature of the jth target respectively; α and β are weight factors; U(R i ) is the utility function of the ith objective; Represents finding the optimal solution of the distance difference.

7. A 94Ghz radar target detection and resolution module, characterized in that: A 94 GHz radar baseband signal solution method according to claim 1, wherein the system comprises: A receiving unit, configured to receive a 94 GHz millimeter wave signal, and perform mixing, amplification, noise reduction, and analog-to-digital conversion on the 94 GHz millimeter wave signal to obtain an intermediate frequency signal; A conversion unit, used for converting the intermediate frequency signal into the digital signal; A parallel processing unit, used for performing noise filtering, normalization and multi-level parallel signal processing on the digital signal to obtain a multi-target difference comparison thermal map; A calculation unit, used for detecting and calculating the distance, speed and angle of multiple targets according to the multi-target difference comparison heat map, to obtain a multi-target calculation map; An optimization unit, used for dynamically adjusting the FFT window, filter bandwidth and gain control parameters of the parallel processing unit; A control unit, used to dynamically adjust the computing resources of the parallel processing unit, the solving unit and the optimization unit by using the importance function and the resource allocation optimization function, and upload the multi-objective solving graph to a target host computer; A synchronization unit is used to provide a synchronization clock signal for the receiving unit, the conversion unit, the parallel processing unit, the solving unit, the optimization unit and the control unit.

Images