Indoor personnel trajectory tracking and anti-interference method and system based on 24G millimeter wave radar

Abstract

The invention discloses an indoor personnel trajectory tracking and anti-interference method based on a millimeter wave radar. The method comprises the following steps: firstly, dividing echo data into dynamic and static point cloud branches for respective processing; performing static clutter suppression on the dynamic point cloud branch, then performing speed dimension fast Fourier transform, and performing direct processing on the static point cloud branch to obtain a distance-Doppler spectrum; determining a target signal and a background clutter; carrying out point cloud screening by taking the point with the strongest energy as a spectrum peak center, and constructing compact representation of a target; and finally, predicting the target state of the current frame, and carrying out target track association. And carrying out track intersection judgment and recording an intersection state, and outputting a human body target position and a track. The method provided by the invention can dynamically adapt to human body sensing requirements in different scenes by combining various correlation analysis technologies, has relatively low calculation complexity and relatively high robustness, and is particularly suitable for human body sensing detection tasks in low-power-consumption application scenes. The system has the advantages of low cost, small size and low power consumption.

Description

Technical Field

[0001] This invention relates to the field of intelligent sensing technology, and in particular to a method and system for indoor personnel trajectory tracking and anti-interference based on millimeter-wave radar. Background Technology

[0002] Existing indoor people tracking technologies (such as infrared and cameras) suffer from privacy leaks and light dependence. While millimeter-wave radar offers advantages in terms of strong penetration and privacy protection, it faces two major interference challenges in the home environment: dynamic interference sources, such as the periodic rotation of fan blades, which produce micro-Doppler characteristics similar to those of the human body; and static micro-motion interference, such as the slight swaying of green plants under airflow, which can be easily misidentified as a stationary person. Traditional multi-channel radar is expensive, while low-cost single-transmitter, dual-receiver radars have weak anti-interference capabilities due to their limited number of channels.

[0003] Therefore, there is a need for an indoor personnel trajectory tracking and anti-interference method and system based on millimeter-wave radar. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide an indoor personnel trajectory tracking and anti-interference method and system based on millimeter-wave radar. The method realizes dynamic identification of human presence, location and movement status; and can accurately distinguish human body from common household interference.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides an indoor personnel trajectory tracking and anti-interference method based on millimeter-wave radar, which includes acquiring echo data of multiple chirped signals using a millimeter-wave radar with a single-transmitter dual-receiver architecture, specifically including the following steps: S1: The echo data is divided into dynamic point cloud branches and static point cloud branches for separate processing; S2: For the dynamic point cloud branch, after static clutter suppression, perform a velocity-dimensional fast Fourier transform to obtain the range-Doppler spectrum; S3: For static point cloud branches, perform a fast Fourier transform directly to obtain the distance-Doppler spectrum; S4: Apply a sliding window mean filter to the distance and velocity dimensions for smoothing. S5: Perform constant false alarm rate detection on the smoothed and filtered data to determine the target signal and background clutter; S6: Use the point with the strongest energy as the center of the spectral peak to screen the point cloud and construct a compact representation of the target; S7: Predict the target state in the current frame based on the target's state vector from the previous frame, and select the nearest neighbor measurement point by Euclidean distance to associate the target track; S8: Determine the intersection of tracks and record the intersection status. Select the update method based on the intersection status and output the human target position and trajectory.

[0006] Furthermore, the static clutter suppression process in step S2 is expressed as follows: in, This represents the distance spectrum obtained by performing a Fast Fourier Transform (FFT) on all sampling points n=0,1,…,N-1 of the m-th chirp; Indicates the number of chirps; Furthermore, the point cloud screening in step S6 includes retaining three neighboring points in front of and behind the spectral peak center, compressing the point cloud width of a single target to within 1 meter.

[0007] Furthermore, in step S7, the prediction of the target state in the current frame is achieved using a motion state transition matrix, and the state vector includes position, velocity, and acceleration components.

[0008] Furthermore, in step S8, the intersection determination is based on the Euclidean distance between trajectories and a preset threshold, classifying the intersection state into intersection between moving trajectories, intersection between moving trajectories and stationary trajectories, or intersection between stationary trajectories, and selecting an update strategy accordingly.

[0009] The present invention provides an indoor personnel trajectory tracking and anti-interference system based on millimeter-wave radar, comprising a millimeter-wave radar module with a single-transmitter dual-receiver architecture for acquiring echo data of multiple chirped signals, and a processing module configured to execute: (1) Divide the echo data into dynamic point cloud branches and static point cloud branches; (2) Perform static clutter suppression and velocity-dimensional fast Fourier transform on the dynamic point cloud branch to generate the range-Doppler spectrum; (3) Perform fast Fourier transform on the static point cloud branches to generate a distance-Doppler spectrum; (4) Apply sliding window mean filtering to the distance and velocity dimensions; (5) Implement constant false alarm rate detection to distinguish target signals from background clutter; (6) Select point clouds based on the points with the strongest energy to compress the target representation width; (7) Predict the target state and associate it with the nearest neighbor measurement points to maintain the track; (8) Determine the trajectory intersection status and update the output target trajectory.

[0010] Furthermore, the processing module also includes a self-calibration unit, which is used to achieve point cloud width compression control by retaining points near the center of the spectral peak during point cloud screening.

[0011] Furthermore, the processing module is integrated into a chip that includes a Cortex-M0+ processor core and an RF transceiver.

[0012] The present invention provides an indoor personnel trajectory tracking and anti-interference system based on millimeter-wave radar, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-mentioned method.

[0013] The beneficial effects of this invention are as follows: The indoor personnel trajectory tracking and anti-interference method and system based on millimeter-wave radar provided by this invention has the advantages of low cost, small size, and low power consumption. This method, by combining multiple correlation analysis techniques, can dynamically adapt to the human perception needs in different scenarios, exhibiting low computational complexity and high robustness, making it particularly suitable for human perception detection tasks in low-power application scenarios.

[0014] This method and system utilizes a 24GHz millimeter-wave radar with a single-transmitter, dual-receiver (one transmitter, two receivers) architecture to achieve high-precision tracking of human movement trajectories in a home environment. It is particularly suitable for complex indoor scenarios with dynamic interference (such as rotating fans) and static micro-motion interference (such as green plants). This method dynamically identifies the presence, location, and movement of human beings; it can accurately distinguish between human bodies and common household disturbances.

[0015] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0016] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following drawings are provided for illustration.

[0017] Figure 1 This is a flowchart of an indoor personnel trajectory tracking and anti-interference method based on millimeter-wave radar.

[0018] Figure 2 This is a system architecture diagram for this embodiment.

[0019] Figure 3 This is a flowchart of the point cloud filtering process in this embodiment.

[0020] Figure 4 This is a flowchart of the track matching and maintenance process in this embodiment.

[0021] Figure 5 These are the RD spectrum and static target spectrum in this embodiment.

[0022] Figure 6 This is a schematic diagram of the CFAR method and a flowchart of the point cloud filtering process.

[0023] Figure 7 This is a diagram illustrating the effect under conditions of multiple people moving in this embodiment.

[0024] Figure 8 This is a diagram illustrating the effect under the conditions of green plants and fan interference in this embodiment. Detailed Implementation

[0025] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0026] Example 1 like Figure 1 As shown, the indoor personnel trajectory tracking and anti-interference method based on millimeter-wave radar provided in this embodiment includes the following steps: S1: Use millimeter-wave radar to transmit multiple chirp signals and receive their reflected echo signals. Utilize the echo data to process the dynamic point cloud branch and the static point cloud branch separately. S2: For the processing branch that acquires dynamic point clouds, static clutter suppression is first performed, and the process can be represented as follows: After static clutter filtering is completed, the result is subjected to velocity-dimensional FFT to obtain the range-Doppler spectrum (RD spectrum) of the signal: in, This is the output, representing the distance-Doppler spectrum at a specific distance index. and Doppler frequency index The complex range at the location; This represents the total number of pulses in the slow time dimension; Indicates the Doppler frequency index; Indicates the impulse index in the slow time dimension; S3: For the processing branch that acquires static point clouds, FFT is performed directly to obtain its RD spectrum and distance spectrum.

[0027] S4: In this embodiment, a sliding window mean filter is applied to the distance and velocity dimensions for smoothing filtering, which can effectively suppress local false high-energy points caused by sudden noise or spectral leakage, and significantly improve the overall cleanliness and usability of the point cloud.

[0028] S5: Then, CFAR (Constant False Alarm Rate) detection is performed. The target signal is searched by sliding within a range-velocity two-dimensional matrix. By setting fixed-size sliding windows in the horizontal and vertical directions, the mean and standard deviation of the background energy within the window are calculated, and a dynamic threshold is determined based on the set false alarm rate threshold. The system classifies the current point as either a "target signal" or "background clutter" according to the calculated threshold value, thus completing target detection.

[0029] S6: In this embodiment, to address the target fusion and misjudgment issues caused by multiple targets approaching or overlapping, a point cloud filtering strategy is implemented using the point with the strongest energy as the spectral peak center. This strategy retains three neighboring points in front and behind the target, constructing a compact representation of the target and compressing the point cloud width of a single target to within 1 meter. This effectively improves the system's multi-target discrimination capability and spatial resolution.

[0030] S7: In this embodiment, target trajectory matching and maintenance are performed. First, the target state in the current frame is predicted based on the target's state vector in the previous frame to match the target trajectory. Then, the measurement point that is the nearest neighbor of the predicted position of the tracked target by Euclidean distance is selected as the measurement associated with the trajectory.

[0031] S8: After successful association, the intersection of the tracks is determined and the intersection status is recorded. The update method is selected according to different statuses, and finally the position and trajectory of the human target are obtained.

[0032] In this embodiment, the system can be continuously monitored and dynamically adjusted based on the target location and different user needs. This method can be applied to low-power 24GHz millimeter-wave radar for real-time detection of human presence within a target area. It features low computational complexity and high robustness, making it particularly suitable for smart home, security monitoring, and energy-saving control scenarios. Example 2 The indoor personnel trajectory tracking anti-interference system based on millimeter-wave radar provided in this embodiment includes a millimeter-wave radar module with a single-transmitter dual-receiver architecture for acquiring echo data of multiple chirped signals. Its distinguishing feature is that it also includes a processing module configured to execute: (1) Divide the echo data into dynamic point cloud branches and static point cloud branches; (2) Perform static clutter suppression and velocity-dimensional fast Fourier transform on the dynamic point cloud branch to generate the range-Doppler spectrum; (3) Perform fast Fourier transform on the static point cloud branches to generate a distance-Doppler spectrum; (4) Apply sliding window mean filtering to the distance and velocity dimensions; (5) Implement constant false alarm rate detection to distinguish target signals from background clutter; (6) Select point clouds based on the points with the strongest energy to compress the target representation width; (7) Predict the target state and associate it with the nearest neighbor measurement points to maintain the track; (8) Determine the trajectory intersection status and update the output target trajectory.

[0033] The processing module in this embodiment also includes a self-calibration unit, which is used to achieve point cloud width compression control by retaining points near the center of the spectral peak during point cloud screening.

[0034] In this embodiment, the processing module is integrated into a chip that includes a Cortex-M0+ processor core and an RF transceiver.

[0035] The indoor personnel trajectory tracking method and anti-interference system based on millimeter-wave radar provided in this embodiment employs a one-transmitter, two-receiver millimeter-wave radar, which has the advantages of low cost, small size, and low power consumption. By combining various correlation analysis techniques, this system can dynamically adapt to the human perception needs in different scenarios, exhibiting low computational complexity and high robustness, making it particularly suitable for human perception detection tasks in low-power application scenarios.

[0036] like Figure 2 As shown, Figure 2 This is a system architecture diagram for this embodiment. Figure 2 The complete processing module architecture of the anti-interference system is shown, and the meaning and function of each module are as follows: The input module is used to acquire echo data; ADC sampling converts analog echo signals into digital signals for subsequent digital processing; Interpolation + 1DFFT is used to increase the sampling rate and improve the range resolution; a fast time Fourier transform is performed on each chirp to obtain the range information. MTD performs Doppler FFT on each range cell to extract velocity information and suppress stationary clutter; Intra-frame cancellation, inter-frame differential cancellation, and spectral accumulation suppress fixed clutter between antennas and remove static background between frames; multiple chirps are accumulated to enhance the energy of weak targets and improve SNR. 2DFFT + RD spectrum smoothing filter + environmental calibration, then perform FFT again in the slow time direction to obtain a higher resolution distance-velocity map; smoothing filter can reduce noise and suppress isolated noise points; environmental calibration can eliminate interference from known stationary objects through a static background template; 2D CFAR + point cloud filtering: Detect target points (point clouds) in the RD spectrum, adaptively set thresholds to remove isolated points and low-energy points, and improve the reliability of effective target points; Angle determination (spectral method, phase detection), estimating target direction through beamforming or subspace methods; The point cloud of the moving target is fused with distance and Doppler information to form the point cloud data of the moving target; Static target point cloud, used to determine the presence of targets; Cancel the spectrum, store the residual signal after inter-frame cancellation, and determine the micro-moving target; Kalman filtering or a-β-y filtering is used for target motion state estimation and smoothing, improving the stability of target tracking; Track prediction predicts the possible location of the target in the next frame for use in matching and joint updates; Observation space trajectory clustering groups adjacent / similar point clouds / trajectories into one category, reducing false alarms and improving robustness; The predicted trajectory is matched with the observed trajectory, and the new observed data is associated with the historical predicted trajectory to update the trajectory status; Dynamic and static trajectory determination + intersection determination: Determine the target's motion state and determine whether there are multiple target trajectories intersecting or crossing to prevent false association; Crossing detection is used to determine whether a target crosses a designated area; Track updates under different strategies: Select different update strategies to improve accuracy; Track management includes starting new tracks, maintaining existing tracks, and deleting outdated tracks, thus achieving track lifecycle management. Adaptive parameter feedback dynamically adjusts thresholds, filter parameters, and detection sensitivity based on environmental changes; Attitude determination: judging the target state by changes in point cloud; Interference removal: Remove non-human interference targets such as fans, plants, and pets; Human presence determination is based on features such as static target point cloud and cancellation spectrum to determine the presence of a human being even if they are not moving. The system outputs comprehensive trajectory information, ultimately providing target information such as distance and speed for use by higher layers. like Figure 3 As shown, Figure 3 This is a flowchart of the point cloud filtering process in this embodiment; ADC sampling converts the analog intermediate frequency signal received by the radar antenna into an analog-to-digital signal, providing a basis for subsequent signal processing. The distance-dimensional FFT transform is used to perform a fast Fourier transform on the data in each chirp to obtain the reflection intensity distribution of the target in the distance dimension; Static clutter suppression eliminates echoes from fixed reflectors (such as walls and tables) by using inter-frame differential and averaging cancellation methods, thus highlighting the signals of moving targets. Perform a distance-dimensional FFT transformation to improve accuracy. The velocity dimension FFT transform is performed on each distance bin in the slow time direction to extract the Doppler frequency and obtain the target velocity information; Smoothing filtering is used to spatially filter the RD map, remove discrete noise points and isolated artifacts, and improve the coherence and stability of the target point cloud. CFAR adaptively detects points on the RD map where the echo intensity exceeds a threshold, identifying "point clouds" where targets may exist. The point cloud filtering strategy further filters points based on their energy, continuity, velocity, and other characteristics, removing isolated points, low SNR points, and non-human feature points. Dynamic point cloud acquisition involves extracting targets with velocity information from the filtered point cloud to generate a "dynamic point cloud". Static point cloud acquisition extracts targets that remain largely stationary between radar frames but still emit stable echoes.

[0037] Example 3 To better understand the human perception detection method for low-power 24GHz millimeter-wave radar provided in this embodiment, further explanation is given below with reference to specific embodiments. In this embodiment, the 24GHz low-power AT24Y4AP12-4415 chip from the domestic company GeKong (Shanghai) Intelligent Technology Co., Ltd. is used. This chip contains a Cortex-M0+ processor core and an RF transceiver. In this example, the radar is installed at a height of 1.8 meters, with an elevation angle of 10°, a horizontal field of view (FOV) of ±60°, and a coverage range of 5 meters.

[0038] The processing steps of this method are as follows: Step 1: Millimeter-wave radar signal acquisition and preprocessing In this step, data acquisition is performed using a 24GHz single-channel millimeter-wave radar. The radar transmits multiple chirped signals and receives the echo signals reflected back from the human body. The frequency of each chirped signal varies linearly, supporting accurate distance measurement. The received echo signals are mixed to generate an intermediate frequency (IF) signal. Subsequently, an analog-to-digital converter (ADC) samples the IF signal to obtain discrete time-domain signal data.

[0039] The transmitted and received signals are mixed to obtain the intermediate frequency (IF) signal: in, It is the baseband frequency difference. Indicates intermediate frequency signal; Indicates the transmission of a signal; Indicates the conjugate of the received signal; This represents a fixed-frequency offset caused by the target distance; Represents the continuous-time variable within a single pulse; Indicates the duration of a single linear frequency modulated pulse; Indicates the time it takes for a signal to travel to and from the target; Indicates the Doppler frequency; Related to the distance to the target; for the th The first chirp cycle within the [number]th There are sampling points, and the discrete time is: in, This refers to the sampling interval. The corresponding discrete sampled signal can be written as: For the first All sampling points of chirp Perform a Fast Fourier Transform (FFT) to obtain the distance spectrum: in, This represents the intermediate frequency signal value at the nth fast-time sampling point in the mth pulse; This represents the number of sampling points in the fast time dimension within each pulse; This indicates the distance to the bin index, and the FFT peak value corresponds to the target distance.

[0040] Step 2: Point Cloud Generation The point cloud generation process is divided into dynamic point cloud acquisition and static point cloud acquisition. For the processing branch that acquires dynamic point clouds, static clutter suppression is performed first, and its processing can be expressed as follows: After static clutter filtering is completed, the result is subjected to velocity-dimensional FFT to obtain the range-Doppler spectrum (RD spectrum) of the signal: For the processing branch that acquires static point clouds, FFT is directly performed to obtain its RD spectrum and distance spectrum.

[0041] In this embodiment, a sliding window mean filter is applied to the distance and velocity dimensions for smoothing, which can effectively suppress local false high-energy points caused by sudden noise or spectral leakage, and significantly improve the overall cleanliness and usability of the point cloud.

[0042] Then, CFAR (Constant False Alarm Rate) detection is performed. The system searches for the target signal by sliding within a range-velocity two-dimensional matrix. By setting fixed-size sliding windows in the horizontal and vertical directions, the mean and standard deviation of the background energy within the window are calculated, and a dynamic threshold is determined based on the set false alarm rate threshold. The system classifies the current point as either a "target signal" or "background clutter" according to the calculated threshold value, thus completing target detection.

[0043] This embodiment addresses the target fusion and misjudgment issues that arise in scenarios where multiple targets are close together or overlap. A point cloud filtering strategy is employed, using the point with the strongest energy as the spectral peak center. This strategy retains three neighboring points in front and behind the target, constructing a compact representation of the target and compressing the point cloud width of a single target to within 1 meter. This effectively improves the system's multi-target discrimination capability and spatial resolution.

[0044] Step 3: Target track matching and maintenance like Figure 4 As shown, Figure 4 The flowchart for track matching and maintenance in this embodiment is as follows: First, based on the target's state vector from the previous frame... The target state in the current frame is predicted using the motion state transition matrix H: in, Indicate The predicted value of the target state vector at time step; Represents the motion state transition matrix; express The state vector of the target at any given time; Indicates the target is Position along the axis; Indicates the target is Position along the axis; Indicates the target is Velocity in the axial direction; Indicates the target is Velocity in the axial direction; Indicates the target is Acceleration in the axial direction; Indicates the target is Acceleration in the axial direction; Then target association is performed. Due to the presence of clutter and noise in the environment, there are multiple false measurements in the positioning results. At the same time, when there are multiple targets in the environment, different targets will also produce measurement results at different locations.

[0045] Therefore, before performing track filtering, it is necessary to determine whether there is a valid track at the present time, and first associate the existing measurement results obtained at the present time with the existing track at the present time.

[0046] To achieve data association, the Nearest Neighbor (NN) algorithm based on Euclidean distance was adopted.

[0047] The basic idea of ​​the NN algorithm is to calculate the Euclidean distance and select the nearest neighbor measurement point to the predicted position of the tracked target as the measurement associated with the track.

[0048] Euclidean distance Indicates two The distance between two vectors is calculated using the following formula: in, Indicates the first vector at the 1st position. Components in the dimension; Indicates the second vector at the th... Components in the dimension; Predict the state vector of the current frame With Dianyunji Trajectory association is performed, and the association threshold uses Euclidean distance, i.e. .

[0049] Based on the trajectory state vector of the previous frame Predict the state vector of the current frame trajectory ; Predict the state vector of the current frame Trajectory association is performed with the point cloud set P = {P1, P2, ..., Pm}, with the association threshold using Euclidean distance. ; in, This represents the first point in the current frame point set 𝑃, including its position information; This represents the second point in the current frame's point cloud, 𝑃. This represents the first element in the current frame point set X. One point; If the association fails, a new trajectory will be created. The creation conditions are: the signal-to-noise ratio (SNR) and the number of distance point clouds must meet the creation conditions, and the turning trajectory will be output smoothly. If the association is successful, the intersection of the tracks will be determined. If the association fails and the signal-to-noise ratio and the number of distance point clouds reach the set threshold, it indicates that a new target has appeared within the detection range, and a new track will be created.

[0050] After successful association, it is necessary to verify the valid trajectories. Perform intersection determination and record the intersection status. The determination criteria are as follows: in, Indicates the first track and the first The Euclidean distance between the tracks; The intersection threshold, Indicates the first The spatial position vector of the current frame of the track; Indicates the first The spatial position vector of the current frame of the track; , Indicates the track numbers of the two tracks to be compared; Based on the threshold and the current target state, the intersection of targets can be divided into three cases: That is, the trajectories of motion intersect with each other; the trajectories of motion intersect with the stationary trajectories; and the trajectories of stationary trajectories intersect with each other. For existing trajectories, when a match is successful, the system will adopt a differentiated update strategy based on different trajectory intersection situations.

[0051] Specifically, trajectory intersections mainly include three types: moving trajectories and moving trajectories, moving trajectories and stationary trajectories, and stationary trajectories and stationary trajectories.

[0052] The system selects different update methods for different intersection types: 1. In intersections involving stationary targets, only the dynamic trajectory state is updated, that is, the dynamic point cloud is updated with the dynamic trajectory, while the stationary trajectory remains unchanged; the intersection involving stationary targets is the intersection of moving trajectories and static trajectories, or the intersection of static trajectories with static trajectories. 2. When the movement trajectories intersect, the points that are successfully associated are updated on the trajectory according to the principle of nearest association; When two motion trajectories intersect, the relevant trajectories are updated synchronously based on the nearest neighbor matching results to prevent trajectory drift and identity confusion, and to ensure the continuity and accuracy of the identities of multiple targets.

[0053] Finally, output the smoothed trajectory; otherwise, return to check if a valid trajectory exists in the next frame (Fram = Fram + 1). During the trajectory maintenance phase, the system predictively grows trajectories that have not been matched with measurements, using motion models to estimate the current state and achieve smooth trajectory continuation, reducing trajectory interruptions caused by short-term occlusion or signal loss. If there are no matches for several consecutive frames, the system determines that the target has left the monitoring range or the signal has been lost, and then terminates the trajectory to release resources.

[0054] like Figure 5-6 As shown, where, Figure 5 These are the RD spectrum and the static target spectrum in this embodiment. Figure 6 Here is a schematic diagram of the CFAR method and a flowchart of the point cloud filtering process; Figure 5The study demonstrates the range-Doppler spectrum and spectral response of the system to static targets in a real indoor environment. It can be observed that due to numerous static reflection sources (such as walls, furniture, and fans) and dynamic interference (such as swaying curtains and air conditioning airflow), the target echo signal is prone to multipath effects and spectral clutter, resulting in multiple spurious peaks in the RD spectrum. These clutter and non-ideal factors interfere with target energy focusing, leading to a large number of false targets in the subsequent point cloud extraction stage, affecting the stability and accuracy of the system in multi-target detection and trajectory tracking.

[0055] Conventional CFAR algorithms are prone to spectral spread when operating under low bandwidth (250 MHz) conditions, resulting in excessively broad energy peaks in the distance dimension of the target point cloud. To address this issue, this embodiment employs a point cloud compression strategy based on energy center constraints, such as... Figure 6 As shown, the point with the strongest energy is used as the center of the spectral peak, and three neighboring points are retained in front and behind to construct a compact representation of the target. This compresses the point cloud width of a single target to less than 1 meter, thereby effectively improving the system's multi-target discrimination capability and spatial resolution.

[0056] like Figure 7-8 As shown, Figure 7 This is a diagram illustrating the effect under conditions of multiple people exercising in this embodiment. Figure 8 This is a diagram illustrating the effect under the interference conditions of green plants and fans in this embodiment. It can be observed that the host computer display includes three parts: radar, infrared, and camera. The upper left part is the radar display, where active personnel are represented by solid red dots. The maximum detection range is 8 meters; when the target exceeds this range, the system will not recognize it. Meanwhile, the debugging bar in the lower right corner provides real-time feedback on target information. This can be seen from... Figure 7 This shows that the process of multiple people converging can accurately identify each target, and at the same time, it can be seen from... Figure 8 The results show that the system can still accurately identify and detect the trajectory of the target even under the interference of the green plant fan.

[0057] The above-described embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. A method for indoor personnel trajectory tracking and anti-interference based on millimeter-wave radar, comprising acquiring echo data of multiple chirped signals using a millimeter-wave radar with a single-transmitter dual-receiver architecture, characterized in that, It also includes the following steps: S1: The echo data is divided into dynamic point cloud branches and static point cloud branches for separate processing; S2: For the dynamic point cloud branch, after static clutter suppression, perform a velocity-dimensional fast Fourier transform to obtain the range-Doppler spectrum; S3: For static point cloud branches, perform a fast Fourier transform directly to obtain the distance-Doppler spectrum; S4: Apply a sliding window mean filter to the distance and velocity dimensions for smoothing. S5: Perform constant false alarm rate detection on the smoothed and filtered data to determine the target signal and background clutter; S6: Use the point with the strongest energy as the center of the spectral peak to screen the point cloud and construct a compact representation of the target; S7: Predict the target state in the current frame based on the target's state vector from the previous frame, and select the nearest neighbor measurement point by Euclidean distance to associate the target track; S8: Determine the intersection of tracks and record the intersection status. Select the update method based on the intersection status and output the human target position and trajectory.

2. The indoor personnel trajectory tracking and anti-interference method based on millimeter-wave radar as described in claim 1, characterized in that: The static clutter suppression process in step S2 is described as follows: in, This represents the distance spectrum obtained by performing a Fast Fourier Transform (FFT) on all sampling points n=0,1,…,N-1 of the m-th chirp; This indicates the number of chirps.

3. The indoor personnel trajectory tracking and anti-interference method based on millimeter-wave radar as described in claim 1, characterized in that: The point cloud filtering in step S6 includes retaining three neighboring points in front of and behind the spectral peak center, compressing the point cloud width of a single target to within 1 meter.

4. The indoor personnel trajectory tracking and anti-interference method based on millimeter-wave radar as described in claim 1, characterized in that: In step S7, the prediction of the target state in the current frame is achieved using a motion state transition matrix, and the state vector includes position, velocity, and acceleration components.

5. The indoor personnel trajectory tracking and anti-interference method based on millimeter-wave radar as described in claim 1, characterized in that: The intersection determination in step S8 is based on the Euclidean distance between trajectories and a preset threshold. The intersection state is classified into intersection between moving trajectories, intersection between moving trajectories and stationary trajectories, or intersection between stationary trajectories, and an update strategy is selected differently.

6. An indoor personnel trajectory tracking and anti-interference system based on millimeter-wave radar, comprising a millimeter-wave radar module with a single-transmitter, dual-receiver architecture, used to acquire echo data of multiple chirped signals, characterized in that, It also includes a processing module, configured to execute: (1) Divide the echo data into dynamic point cloud branches and static point cloud branches; (2) Perform static clutter suppression and velocity-dimensional fast Fourier transform on the dynamic point cloud branch to generate the range-Doppler spectrum; (3) Perform fast Fourier transform on the static point cloud branches to generate a distance-Doppler spectrum; (4) Apply sliding window mean filtering to the distance and velocity dimensions; (5) Implement constant false alarm rate detection to distinguish target signals from background clutter; (6) Select point clouds based on the points with the strongest energy to compress the target representation width; (7) Predict the target state and associate it with the nearest neighbor measurement points to maintain the track; (8) Determine the trajectory intersection status and update the output target trajectory.

7. The indoor personnel trajectory tracking and anti-interference system based on millimeter-wave radar as described in claim 6, characterized in that: The processing module also includes a self-calibration unit, which is used to achieve point cloud width compression control by retaining points near the center of the spectral peak during point cloud screening.

8. The indoor personnel trajectory tracking and anti-interference system based on millimeter-wave radar as described in claim 6, characterized in that: The processing module is integrated into a chip that includes a Cortex-M0+ processor core and an RF transceiver.

9. An indoor personnel trajectory tracking and anti-interference system based on millimeter-wave radar, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method described in any one of claims 1 to 5.

Images