A real-time human body detection method and system based on microwave radar

Abstract

This invention provides a real-time human detection method and system based on microwave radar, applicable to the field of signal data processing. By constructing a sparse reconstruction model and combining it with multidimensional feature decomposition, this invention effectively alleviates the recognition difficulties caused by target signal aliasing. It continuously outputs and receives radar signals, extracts range, Doppler, and angle features, and constructs a target distribution point cloud in a three-dimensional feature space. Furthermore, by utilizing angle continuity judgment and point cloud density analysis, it accurately estimates the number and distribution of overlapping targets. For scenarios where no continuous region is formed, a Kalman filter is introduced for dynamic tracking, achieving continuous separation of multiple targets and pseudo-color image reconstruction, thereby significantly improving the resolution accuracy and target positioning capability of radar in scenarios where multiple people are gathered at close range.

Description

Technical Field

[0001] This invention relates to the field of signal data processing, and in particular to a real-time human body detection method and system based on microwave radar. Background Technology

[0002] Microwave radar transmits electromagnetic waves at specific frequencies (such as 24GHz, 60GHz, or 77GHz), which are reflected upon encountering the human body, and the radar receives the echo signals. Because of the subtle movements of the human body (such as breathing fluctuations, heartbeat vibrations, and micro-movements), these movements cause minute changes in the frequency (micro-Doppler effect), amplitude, and phase of the echo signals. By analyzing these changes, it is possible to detect human presence, recognize movement, and even monitor vital signs.

[0003] However, when two or more targets are very close in space (e.g., only a few tens of centimeters apart), the radar waves they reflect almost overlap in time delay (range information) and may also be very close in spectrum and angle, resulting in severe superposition of received signals. The superimposed signals are difficult to distinguish through traditional matched filters or beamformers, causing errors in quantity estimation or misidentification. Summary of the Invention

[0004] This invention aims to address the problem that radar detects multiple human body reflections with similar echo time, frequency, and amplitude characteristics, making it impossible to accurately estimate the number of targets and their respective locations. It provides a real-time human body detection method and system based on microwave radar.

[0005] The present invention employs the following technical means to solve the technical problem:

[0006] This invention provides a real-time human detection method based on microwave radar, comprising:

[0007] Based on the preset detection scenario of the microwave radar, the echo signal of the detection scenario is repeatedly output and received within a preset time period.

[0008] Determine whether the echo signal detects a peak value of the merged signal;

[0009] If so, based on the pre-collected original radar signal, the original radar signal is used as superimposed data of sparse source to construct a corresponding sparse reconstruction model. The aliased signal is decomposed into corresponding sparse units using a preset sparse coding. Multidimensional features are extracted from the sparse units. The target distribution point cloud of the multidimensional features is established through a preset three-dimensional feature space. The multidimensional features specifically include distance features, Doppler features, and angle features.

[0010] Determine whether the angular direction of the target distribution point cloud forms a continuous region;

[0011] If no target distribution point cloud is formed, the local point cloud density of the target distribution point cloud is identified. Based on the local point cloud density, the corresponding number of targets is dynamically estimated. A preset Kalman filter is applied to continuously track the target parameters of each target, generate the separated multi-target echo, and reconstruct the multi-target echo into a pseudo-color radar image. The target parameters specifically include target velocity and target velocity change.

[0012] Furthermore, the step of decomposing the aliased signal into corresponding sparse units using a preset sparse coding method and extracting multidimensional features from the sparse units further includes:

[0013] Calculate the squared amplitude integral of the sparse unit, normalize the squared amplitude integral to a preset standard scale, apply a preset continuous wavelet transform to the sparse unit, and generate a corresponding time-frequency spectrum, wherein the time-frequency spectrum specifically includes the dominant frequency, bandwidth, and frequency change rate.

[0014] Determine whether the time-frequency spectrum detects a preset low-frequency periodic motion feature, wherein the low-frequency periodic motion feature specifically includes breathing and heartbeat;

[0015] If so, based on the echo delay of the sparse unit, the target distance corresponding to the echo delay is collected, and according to the target distance, the internal point distribution pattern of the sparse unit is statistically analyzed. Within a preset time window, the time trajectory evolution of the sparse unit is tracked. The internal point distribution pattern specifically includes the principal axis length, flatness, and density distribution characteristics.

[0016] Furthermore, after the step of establishing the target distribution point cloud of the multidimensional features through a preset three-dimensional feature space, the method further includes:

[0017] Based on the point cloud density of the target distribution point cloud, identify the number of neighboring points of each point cloud within a preset radius;

[0018] Determine whether the number of neighboring points is less than a preset value;

[0019] If so, then detect extreme outliers of the multidimensional features, evaluate the dispersion of each point cloud based on the extreme outliers, statistically analyze the topological characteristics of the target distribution point cloud, and add additional features to each point cloud. Specifically, the extreme outliers include reflection energy supermassive and frequency jumps, the topological characteristics include the main direction of the point cloud and the overall shape of the point cloud, and the additional features include local curvature and local velocity gradient.

[0020] Furthermore, the step of identifying the local point cloud density of the target distribution point cloud and dynamically estimating the corresponding number of targets based on the local point cloud density further includes:

[0021] Identify continuous frame point clouds from the target distribution point cloud, perform temporal fusion of the continuous frame point clouds, and generate corresponding high-density point cloud regions.

[0022] Determine whether the high-density region of the point cloud recurs within a preset frequency;

[0023] If so, the location information of the high-density region of the point cloud is obtained, and a preset micro-motion spectrum feature is extracted from the high-density region of the point cloud. Based on the location information and the micro-motion spectrum feature, the target type of the continuous frame point cloud is dynamically divided. The micro-motion spectrum feature specifically includes breathing frequency and arm swing frequency, and the target type specifically includes dynamic human body, interference object and non-human body object.

[0024] Furthermore, the step of determining whether the echo signal detects a peak value of the combined signal also includes:

[0025] Obtain the full width at half maximum (FWHM) of the main peak of the echo signal;

[0026] Determine whether the full width at half maximum (FWHM) of the main peak exceeds a preset multiple of the historical single-target waveform width;

[0027] If so, then detect several small peaks of the full width at half maximum (FWHM) of the main peak, identify the integrated energy on the left and right sides of the small peaks, and dynamically collect the superimposed waveforms caused by the reflection amplitudes of each target based on the symmetry of the integrated energy. Specifically, the small peaks include shoulder peak sub-valleys, plateau-shaped waveform sub-valleys, and inter-peak sub-valleys.

[0028] Furthermore, the step of determining whether the angular direction of the target distribution point cloud forms a continuous region also includes:

[0029] The target distribution point cloud is sorted according to a preset angle value, and the sorting constructs a corresponding angle sequence. The angle increment of the angle sequence is then calculated.

[0030] Determine whether the angle increment has increased;

[0031] If so, identify the break point after the angle increment increases, perform clustering and cutting at the break point as the separation boundary of the point cloud target, divide the point cloud target into several continuous segments based on the position of the break point, and use the several continuous segments as the angle region of a single candidate target.

[0032] Furthermore, the step of repeatedly outputting and receiving the echo signal of the detection scene based on the preset detection scene of the microwave radar within a preset time period also includes:

[0033] Based on the preset scene type of the detection scenario, the beam coverage strategy of the microwave radar is adaptively switched. The scene type specifically includes corridor, room and doorway, and the beam coverage strategy specifically includes omnidirectional, directional and variable beam.

[0034] Determine whether the beam coverage strategy can meet the preset target dynamic detection;

[0035] If not, then according to the preset detection parameters of the microwave radar, the preset static shielding area is dynamically activated to block the obstacle echoes pre-collected by the microwave radar through the static shielding area. The detection parameters specifically include the horizontal scanning angle range, the vertical coverage range, and the effective distance.

[0036] This invention also provides a real-time human detection system based on microwave radar, comprising:

[0037] The receiving module is used to repeatedly output and receive the echo signal of the detection scene within a preset time period based on the preset detection scene of the microwave radar.

[0038] The judgment module is used to determine whether the echo signal detects a peak value of the merged signal;

[0039] The execution module is used to, if so, construct a corresponding sparse reconstruction model based on the pre-acquired original radar signal as superimposed data of a sparse source, decompose the aliased signal into corresponding sparse units using a preset sparse coding, extract multi-dimensional features from the sparse units, and establish a target distribution point cloud of the multi-dimensional features through a preset three-dimensional feature space, wherein the multi-dimensional features specifically include distance features, Doppler features, and angle features;

[0040] The second judgment module is used to determine whether the angular direction of the target distribution point cloud forms a continuous region;

[0041] The second execution module is used to identify the local point cloud density of the target distribution point cloud if no target distribution point cloud is formed, dynamically estimate the corresponding number of targets based on the local point cloud density, apply a preset Kalman filter to continuously track the target parameters of each target, generate the separated multi-target echo, and reconstruct the multi-target echo into a pseudo-color radar image. The target parameters specifically include target velocity and target velocity change.

[0042] Furthermore, the execution module also includes:

[0043] The computing unit is used to calculate the amplitude squared integral of the sparse unit, normalize the amplitude squared integral to a preset standard scale, apply a preset continuous wavelet transform to the sparse unit, and generate a corresponding time-frequency spectrum, wherein the time-frequency spectrum specifically includes the dominant frequency, bandwidth and frequency change rate.

[0044] The judgment unit is used to determine whether the time-frequency spectrum detects a preset low-frequency periodic motion feature, wherein the low-frequency periodic motion feature specifically includes breathing and heartbeat;

[0045] An execution unit is configured to, if so, acquire the target distance corresponding to the echo delay based on the echo delay of the sparse unit, statistically analyze the internal point distribution pattern of the sparse unit according to the target distance, and track the time trajectory evolution of the sparse unit within a preset time window, wherein the internal point distribution pattern specifically includes the principal axis length, flatness, and density distribution characteristics.

[0046] Furthermore, it also includes:

[0047] The identification module is used to identify the number of neighboring points of each point cloud within a preset radius based on the point cloud density of the target distribution point cloud.

[0048] The third judgment module is used to determine whether the number of neighbor points is less than a preset value;

[0049] The third execution module is used to detect extreme outliers of the multidimensional features if the condition is met, evaluate the dispersion of each point cloud based on the extreme outliers, statistically analyze the topological characteristics of the target distribution point cloud, and add additional features to each point cloud. Specifically, the extreme outliers include reflection energy supermassive and frequency jumps, the topological characteristics include the main direction of the point cloud and the overall shape of the point cloud, and the additional features include local curvature and local velocity gradient.

[0050] This invention provides a real-time human detection method and system based on microwave radar, which has the following beneficial effects:

[0051] This invention effectively alleviates the identification difficulties caused by target signal aliasing by constructing a sparse reconstruction model and combining it with multidimensional feature decomposition. It continuously outputs and receives radar signals, extracts range, Doppler, and angle features, and constructs a target distribution point cloud in a three-dimensional feature space. Furthermore, it uses angle continuity judgment and point cloud density analysis to accurately estimate the number and distribution of overlapping targets. For scenarios where no continuous region is formed, a Kalman filter is introduced for dynamic tracking to achieve continuous separation of multiple targets and pseudo-color image reconstruction, thereby significantly improving the resolution accuracy and target positioning capability of radar in scenarios where multiple people are gathered at close range. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating an embodiment of the real-time human detection method based on microwave radar of the present invention.

[0053] Figure 2 This is a structural block diagram of an embodiment of the real-time human body detection system based on microwave radar of the present invention. Detailed Implementation

[0054] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. The realization of the purpose, functional features, and advantages of the invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings.

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Reference Appendix Figure 1 The present invention provides a real-time human detection method based on microwave radar, comprising:

[0057] S1: Based on the preset detection scenario of the microwave radar, the echo signal of the detection scenario is repeatedly output and received within a preset time period.

[0058] S2: Determine whether the echo signal detects a peak value of the merged signal;

[0059] S3: If so, based on the pre-collected original radar signal, the original radar signal is used as superimposed data of sparse source to construct a corresponding sparse reconstruction model. The aliased signal is decomposed into corresponding sparse units using a preset sparse coding. Multidimensional features are extracted from the sparse units. The target distribution point cloud of the multidimensional features is established through a preset three-dimensional feature space. The multidimensional features specifically include distance features, Doppler features, and angle features.

[0060] S4: Determine whether the angular direction of the target distribution point cloud forms a continuous region;

[0061] S5: If not formed, identify the local point cloud density of the target distribution point cloud, dynamically estimate the corresponding number of targets based on the local point cloud density, apply a preset Kalman filter to continuously track the target parameters of each target, generate the separated multi-target echo, and reconstruct the multi-target echo into a pseudo-color radar image, wherein the target parameters specifically include target velocity and target velocity change.

[0062] In this embodiment, the system repeatedly outputs and receives echo signals from a pre-set detection scenario based on a microwave radar within a pre-defined time period. The system then determines whether these echo signals detect a merged signal peak and executes corresponding steps accordingly. For example, if the system determines that no merged signal peak is detected in the repeatedly output and received echo signals, it assumes that there is a certain distance separation between the various human targets, and that the radar echoes do not exhibit significant overlap or aliasing in time, frequency, and angle dimensions. The system will continue to maintain the current output and reception cycle, increasing its detection capability for weak or distant targets, and may consider adjusting the radar's receiving gain or changing its transmission power to enhance the detection capability for distant or small targets. The system can directly process each independent echo, extracting multi-dimensional features (such as range, velocity, and Doppler effect) and performing target identification and localization in a three-dimensional feature space, since there is no signal peak merging. Furthermore, if the number of targets is small, the system can identify each target individually using a simple separation method based on range and Doppler features, ensuring the accurate position and motion state of each target. Based on the target distribution, the system dynamically adjusts its operating parameters (such as scanning angle and transmission frequency) to better adapt to the scene requirements. If the system detects a relatively static detection area, it can also reduce the output frequency to decrease power consumption. For example, when the system... The system detects a peak in the merged signal when repeatedly outputting and receiving echo signals from the detection scene. At this point, the system assumes that the distances between various human targets are close, easily causing radar echo overlap or aliasing. Based on pre-collected raw radar signals, the system treats these signals as superimposed data from sparse sources, constructs a corresponding sparse reconstruction model, and uses pre-defined sparse coding to decompose the aliased signal into corresponding sparse units. Multidimensional features are extracted from these sparse units, specifically including range features, Doppler features, and angle features. A target distribution point cloud with multidimensional features is established using a pre-defined three-dimensional feature space. The system treats the raw radar signals as superimposed data from multiple sparse sources and employs a sparse reconstruction model... By deconstructing the signal, the merged signal can be effectively separated into sparse units with identifiable features, thus overcoming the limitation of traditional methods in detecting multiple people at close range. At the same time, three-dimensional features of distance, Doppler and angle are extracted from the sparse units to form richer target information, which is conducive to distinguishing targets from multiple dimensions. Especially when targets have similar shapes and similar directions of movement, the accuracy of target classification and recognition is improved. Furthermore, the target distribution point cloud established by the three-dimensional feature space can intuitively show the relative position and density relationship of targets in space, providing a basic data structure for subsequent operations such as target quantity estimation, trajectory separation and dynamic tracking, and improving the system's multi-target management capability in crowded or occluded scenarios.The system then determines whether the angular orientations of these target distribution point clouds form a continuous region, and executes the corresponding steps accordingly. For example, when the system determines that the angular orientations of the target distribution point clouds form a continuous region, it considers that relatively complete target reflection information has been detected within this angular region, which can accurately represent the target's directional range in space. The system adjusts the echo signal processing method, merging the echo signals within the continuous region to avoid unnecessary signal separation. At this point, the target recognition strategy should focus on the analysis of group behavior rather than individual target separation to improve processing efficiency. Simultaneously, based on density analysis, motion patterns, and multidimensional features, the system dynamically estimates the number of targets and uses appropriate classification algorithms (such as cluster-based classification). The system classifies and identifies these targets using a class-based approach. By analyzing the dynamic changes in the targets' positions, it can more accurately determine the relationships between targets, avoiding misidentification in multi-target scenarios. Furthermore, for occasionally occurring continuous angular regions, the system should carefully eliminate noise interference. By setting reasonable thresholds and signal enhancement mechanisms, it can ensure that the continuity of target distribution is not due to misjudgment caused by interference signals or environmental noise. For example, when the system determines that the angular direction of the target distribution point cloud does not form a continuous region, it will consider that no complete target reflection information has been detected within this angular region, and the directional range of the target in space cannot be represented. The system will then identify the local point cloud density of the target distribution point cloud and dynamically estimate the corresponding... The system identifies the number of human targets by continuously tracking their parameters, including velocity and velocity variation, using a pre-set Kalman filter. This generates separated multi-target echoes, which are then reconstructed into a pseudo-color radar image. Further analysis of the local point cloud density reveals multiple high-density reflective sub-regions even in areas lacking angular extension, allowing for dynamic estimation of the actual number of targets. This density-based supplementation mechanism significantly improves the system's robustness in weak signal regions or under non-ideal reflection conditions. Furthermore, when multiple targets are close together or move at similar speeds, their radar echoes often overlap in time, frequency, and angular dimensions. This leads to the difficulty in distinguishing instantaneous targets. To address this, a Kalman filter is introduced. Based on the target's historical motion state and combined with new observation data, the system continuously estimates and corrects the velocity and velocity change parameters of each target. This not only effectively avoids target trajectory interruptions caused by short-term aliasing but also maintains the consistency of target trajectories even with partial obstruction or signal loss. Furthermore, after completing sparse reconstruction, feature extraction, point cloud generation, and target tracking for multiple targets, the system synthesizes the separated target echo signals into a pseudo-color radar image. Each target is assigned a different color code based on its characteristic parameters. This visual approach improves the readability of the system results and the efficiency of human-computer interaction, allowing end users to intuitively identify the number, location, and direction of movement of targets.

[0063] It should be noted that, based on the pre-acquired raw radar signals, the raw radar signals are used as superimposed data of sparse sources to construct a corresponding sparse reconstruction model. A preset sparse coding is used to decompose the aliased signals into corresponding sparse units, and multi-dimensional features are extracted from the sparse units. A specific example is as follows:

[0064] Suppose a microwave radar is installed on the ceiling or side wall of an office corridor to monitor the number, direction, and speed of people passing by. At a certain moment, two employees, A and B, walk side by side, less than 40 centimeters apart, at the same speed, and are of similar size. A distinct echo peak is generated at the radar receiver, but only one set of echo data is available, as shown below:

[0065] There is only one concentrated reflection point (e.g., 2.5 meters) from the reflection peak;

[0066] The Doppler frequency shift is 0.4 m / s (forward travel);

[0067] The angle estimates are close to overlapping (e.g., +5°);

[0068] If a radar system uses traditional CFAR or FFT+angle estimation, it typically cannot distinguish between two people and may misclassify them as a "single target," which will lead to subsequent statistical errors. Therefore, the processing steps and explanations are as follows:

[0069] 1. Upon detecting a peak in the merged signal, the radar system analyzes the echo energy distribution at that moment and finds that the reflected signal is relatively wide (waveform broadening), indicating signal overlap, thus triggering the aliasing detection mechanism.

[0070] 2. Construct a sparse reconstruction model system that calls the "original radar signal library," which contains the following collected data:

[0071] Single-person reflection templates A1, A2, and A3: represent people of different heights, body types, directions, and speeds;

[0072] Each template signal structure contains multiple channels (such as range channel, velocity channel, and angle channel) of raw radar IQ data;

[0073] Construct a model that expresses the current echo signal Y as a linear combination (sparse form) of a set of original templates A:

[0074] Y≈α1*A1+α2*A2+...+αn*An+e

[0075] Where α is the sparsity coefficient and e is the residual;

[0076] 3. Sparse coding decomposition is adopted, and the system uses the orthogonal matching pursuit (OMP) algorithm for decoding: in the sparse dictionary, the most relevant templates are matched one by one; finally, two sets of templates A5 and A9 with significant responses are matched.

[0077] Decoding results indicate that the current signal can be interpreted as the result of the overlap of two sparse target sources;

[0078] 4. Multidimensional feature extraction: For the original templates of A5 and A9 respectively, the system extracts the following features:

[0079] feature Target A (corresponding to A5) Target B (corresponding to A9) distance 2.43m 2.56m speed 0.39m / s 0.41m / s angle +4.3° +6.2°

[0080] The system represents these two sparse units as two independent point cloud clusters in the three-dimensional feature space, which prepares for subsequent point cloud connectivity judgment, target quantity estimation, trajectory tracking, etc.

[0081] In summary, by combining the examples above, the system ultimately determines that there are actually two targets at that moment, rather than one, successfully avoiding the problems of "target merging" or "missed detection" in traditional processing algorithms; even if two people are close to each other and their movements are consistent, they are still separated into independent signals through sparse modeling; it is particularly suitable for human target recognition and statistical analysis in high-density environments such as elevator entrances and classroom lobbies.

[0082] It should be added that the local point cloud density of the target distribution point cloud is identified, and the corresponding number of targets is dynamically estimated based on the local point cloud density. A preset Kalman filter is applied to continuously track the target parameters of each target, generate the separated multi-target echoes, and reconstruct a pseudo-color radar image from the multi-target echoes. A specific example is as follows:

[0083] Imagine a microwave radar deployed above an elevator door in an office building to detect people waiting to board the elevator. At a certain time, the elevator is about to arrive at the station, and several people gather at the door. The radar system needs to determine the number of people waiting and continuously track the movement of the target.

[0084] At this moment, four people enter the radar's field of view almost simultaneously, with two standing side by side and the other two approaching one after the other; the detailed system processing flow is as follows:

[0085] 1. Point cloud data generation: The radar detected multiple reflected signals, but due to the close distance between the personnel (only about 0.3 to 0.5 meters), some echoes merged; a relatively dense but blurry three-dimensional point cloud region was initially formed.

[0086] 2. Local point cloud density recognition: The system divides the 3D point cloud into voxel grids, for example, a unit block is a range of 0.2m × 0.2m × 10° angle; it detects multiple high-density areas: there are two point cloud kernels near the elevator door in front (two people side by side); two relatively sparse but stable point cloud clusters are formed a little further behind (one in front and one behind); the system calculates the density value of each point cloud cluster (if the number of points per voxel is >20, it is considered high density), and dynamically estimates that there are 4 independent targets based on the density and its spatial distribution;

[0087] 3. Kalman filter tracking: Create a Kalman filter for each estimated target and initialize the state vector:

[0088] Position coordinates (X, Y);

[0089] Velocity vector (Vx, Vy);

[0090] The system acquires the updated point cloud center position and Doppler velocity in each frame, and the Kalman filter continuously predicts and corrects based on the time series to solve the problem of partial target occlusion or temporary signal loss.

[0091] For example:

[0092] In frame 1, all four targets are clearly visible.

[0093] In the second frame, the third target is occluded behind it, but the system continues to track its movement based on prediction.

[0094] In the third frame, the third target reappears, and its path closely matches the predicted path.

[0095] 4. Multi-target echo separation: The system back-maps the tracking results to the original radar signal and separates the reflection time sequence of each target; thus obtaining four sets of independent target time-frequency echo data.

[0096] 5. Pseudo-color image reconstruction: The distance, Doppler velocity, and reflection intensity of each target are mapped to color elements on the pseudo-color image.

[0097] Distance → Hue: Near is red, far is blue;

[0098] Speed ​​→ Saturation: Faster speed means higher saturation, slower speed means lower saturation;

[0099] Reflection intensity → Brightness: Stronger is brighter, weaker is darker;

[0100] The final image shows four distinctly different "human shadow areas," which facilitates subsequent system or manual identification.

[0101] In summary, in the examples above, the system can still accurately identify the number of targets in densely populated close-range scenarios, overcoming the problem of traditional radar's "waveform overlap making it difficult to distinguish individuals." Kalman filtering enhances robustness against short-term obstruction or interference signals, making the tracking process more coherent. The pseudo-color image intuitively expresses changes in target status, which is beneficial for equipment maintenance, intelligent identification, or data archiving and analysis.

[0102] In this embodiment, the aliasing signal is decomposed into corresponding sparse units using a preset sparse coding method. Step S3, which extracts multidimensional features from the sparse units, further includes:

[0103] S31: Calculate the amplitude squared integral of the sparse unit, normalize the amplitude squared integral to a preset standard scale, apply a preset continuous wavelet transform to the sparse unit, and generate a corresponding time-frequency spectrum, wherein the time-frequency spectrum specifically includes the dominant frequency, bandwidth and frequency change rate.

[0104] S32: Determine whether the time-frequency spectrum detects a preset low-frequency periodic motion feature, wherein the low-frequency periodic motion feature specifically includes breathing and heartbeat;

[0105] S33: If so, based on the echo delay of the sparse unit, the target distance corresponding to the echo delay is collected, and according to the target distance, the internal point distribution pattern of the sparse unit is statistically analyzed. Within a preset time window, the time trajectory evolution of the sparse unit is tracked. The internal point distribution pattern specifically includes the principal axis length, flatness, and density distribution characteristics.

[0106] In this embodiment, the system calculates the amplitude squared integral of the sparse units, normalizes these amplitude squared integrals to a pre-set standard scale, applies a pre-set continuous wavelet transform to these sparse units, and generates a corresponding time-frequency spectrum. The time-frequency spectrum specifically includes the dominant frequency, bandwidth, and rate of change of frequency. Then, the system determines whether the time-frequency spectrum detects pre-set low-frequency periodic motion features, specifically including breathing and heartbeat, to execute corresponding steps. For example, if the system determines that the time-frequency spectrum does not detect the pre-set low-frequency periodic motion features, the system will consider that multiple human micro-motion signals are superimposed in the sparse units, leading to... If energy diffusion or waveform distortion occurs in the frequency band, making it impossible to clearly identify low-frequency features, the system will temporarily classify the reflective target corresponding to the current sparse cell as a "non-biological target" or "inactive target." It will then combine position and velocity characteristics for target classification and filtering to avoid misidentifying them as interactive or alarm-requiring human objects. If the target is located in the region of interest (such as a bed, sofa, or monitoring area), the analysis period will be automatically extended (e.g., from 5 seconds to 10 seconds) for a second time-frequency analysis to improve weak signal detection capabilities. Furthermore, if no low-frequency periodic features are detected in multiple sparse cells, the system may attempt to adjust the radar gain, antenna beam, or activate auxiliary channels (such as vertical beamforming) to improve detection sensitivity. If such "targets without low-frequency characteristics" appear repeatedly, they should be marked and uploaded to the cloud or backend database as part of behavioral risk identification (e.g., sudden fainting or abnormal stillness). For example, when the system determines that a pre-set low-frequency periodic motion feature has been detected in the time-frequency spectrum, the system will consider the radar signal quality to be good, the micro-motion components (such as breathing, heartbeat, etc.) not masked by noise or interference, the system can clearly distinguish the target's vital signs, and the signal processing effect is good. The system will collect the target distance corresponding to the echo delay based on the sparse unit echo delay, and statistically analyze the internal point distribution pattern of the sparse unit according to different target distances. The morphology specifically includes the principal axis length, flatness, and density distribution characteristics. Within a pre-set time window, the system tracks the temporal trajectory evolution of these sparse units. Based on the echo delay of the sparse units, the system can collect the delay information of the echo signal, thereby accurately calculating the target distance. Through echo delay, the system can determine the target's position. Especially when the target is at different distances, the system can dynamically adjust its detection range to further improve tracking accuracy. At the same time, by statistically analyzing the internal point distribution morphology of the sparse units (including principal axis length, flatness, density distribution characteristics, etc.), the system can further reveal the target's morphological characteristics. For example, the target's density distribution characteristics help to infer the target's size, shape, and motion state.Changes in the spindle length and flatness can be used to determine the target's trajectory and attitude. Within a set time window, the system can track the temporal evolution of sparse units. By continuously tracking the target's trajectory, the system can acquire the target's motion trend (such as changes in speed and direction) in real time, enabling dynamic monitoring and prediction of the target. This is particularly important for applications such as smart homes, health monitoring, and security systems.

[0107] It should be noted that, based on the echo delay of the sparse unit, the target distance corresponding to the echo delay is collected. According to the target distance, the internal point distribution pattern of the sparse unit is statistically analyzed. Within a preset time window, the temporal trajectory evolution of the sparse unit is tracked. A specific example is as follows:

[0108] Suppose a microwave radar system is installed at the entrance of a shopping mall to monitor for unauthorized personnel entering and ensure the mall's security. By detecting, identifying, and tracking targets, the system can effectively distinguish different targets and determine their direction of movement.

[0109] 1. Radar detection delay of multiple sparse units: In this scenario, the radar system is installed at the entrance of a shopping mall and can detect multiple targets. Suppose that the radar system detects two reflected waves with echo delays of 15 nanoseconds and 25 nanoseconds, respectively, corresponding to target distances of 2.5 meters and 3.2 meters.

[0110] Example:

[0111] The first echo has a delay of 15 nanoseconds, and the calculated distance is 2.5 meters. The target is likely the first person approaching the entrance.

[0112] The second echo had a delay of 25 nanoseconds, and the calculated distance was 3.2 meters, suggesting that the target was another intruder approaching the entrance.

[0113] 2. Analyze the point cloud morphology to determine the target state. The system then analyzes the point cloud inside the sparse unit of the two targets to determine their target state.

[0114] Example:

[0115] The first target: its point cloud has a short main axis length (about 0.4 meters), high flatness, and high density, indicating that the target is in a stable standing state and its posture is relatively stable. It may be standing at the doorway to observe or to perform some actions.

[0116] The second target: its point cloud has a relatively long main axis (about 0.8 meters), low flatness, low density, and significant movement, indicating that the target is rapidly crossing the entrance of the shopping mall and may have the intention to enter.

[0117] 3. Time trajectory tracking and behavior determination: The system uses tracking algorithms (such as Kalman filtering) to track the time trajectories of the two targets in order to determine their motion trajectories and behaviors;

[0118] Example:

[0119] First target: Within a 10-second time window, the first target remained relatively stationary at the entrance, without any noticeable movement; the system judged this as normal behavior.

[0120] The second target: The point cloud location of the second target moves rapidly from the entrance into the mall within 10 seconds. The system determines it to be a fast-moving target based on its dynamic changes such as speed and acceleration. Combined with historical data, the system may determine that this target is an intruder and trigger a security alarm.

[0121] In summary, through echo delay and internal point cloud analysis, the system can accurately determine the position, attitude, and behavior of targets, greatly improving the accuracy of target recognition. At the same time, by tracking the target's time trajectory and monitoring its changes in real time, the system can effectively identify the target's behavior patterns, avoid false alarms, and handle interference between multiple targets, accurately distinguishing the behavior trajectory of each target and adapting to complex monitoring scenarios.

[0122] In this embodiment, after step S3 of establishing the target distribution point cloud of the multidimensional features through a preset three-dimensional feature space, the method further includes:

[0123] S301: Based on the point cloud density of the target distribution point cloud, identify the number of neighboring points of each point cloud within a preset radius;

[0124] S302: Determine whether the number of neighbor points is less than a preset value;

[0125] S303: If so, detect extreme outliers of the multidimensional features, evaluate the dispersion of each point cloud based on the extreme outliers, statistically analyze the topological characteristics of the target distribution point cloud, and add additional features to each point cloud. Specifically, the extreme outliers include reflection energy supermassive and frequency jumps, the topological characteristics include the main direction of the point cloud and the overall shape of the point cloud, and the additional features include local curvature and local velocity gradient.

[0126] In this embodiment, the system identifies the number of neighboring points within a pre-defined radius for each point cloud based on the point cloud density of the target distribution point cloud. Then, the system determines whether the number of these neighboring points is less than a pre-defined value to execute corresponding steps. For example, if the system determines that the number of neighboring points within the pre-defined radius for each point cloud is not less than the pre-defined value, the system considers the current point cloud to have high local density and stability, and is not a discrete noise point. Therefore, there is no need to remove noise points. The system uses a clustering algorithm to cluster these dense point cloud regions, distinguishing different targets. The clustering results help the system identify different targets and simultaneously... Clustered point clouds, by extracting features such as geometric shape, boundaries, and centroid location, can be further analyzed to determine the target's orientation, volume, and morphology. Furthermore, if time-series data is available, the system can calculate the target's motion parameters, such as velocity, acceleration, and direction, based on its trajectory. Kalman filtering and other algorithms can then be used to track the target and predict its future position. Based on the target's geometric features and dynamic changes, classification algorithms (such as deep learning models and support vector machines) can be used for target recognition, identifying whether the target is a pedestrian, a stationary object, or another moving object. For example, when the system determines that each point cloud point is within a pre-defined radius... If the number of neighboring points is less than a preset value, the system considers the current point cloud to be unstable and may belong to discrete noise points, requiring noise point removal. The system detects extreme outliers in multi-dimensional features, specifically including excessive reflection energy and frequency jumps. Based on these extreme outliers, the system assesses the dispersion of each point cloud and statistically analyzes the topological characteristics of the target distribution point cloud, including the main orientation and overall shape of the point cloud. Additional features are added to each point cloud, including local curvature and local velocity gradient. By determining the number of neighboring points within a preset radius and combining this with extreme outliers in reflection energy and frequency, the system can more effectively... The system accurately distinguishes between valid target points and discrete noise points, thereby avoiding the accidental deletion of edge targets or weakly reflective targets, improving the purity and reliability of point cloud data. At the same time, it further analyzes the topological characteristics of the point cloud (such as main direction and overall shape) to extract additional features such as local curvature and velocity gradient, which helps to accurately model the geometry and motion morphology of the target, improve the multi-dimensional feature recognition capability of the target and the stability of continuous tracking. Furthermore, with the help of multi-dimensional outlier detection and topological analysis, the system can not only remove invalid point clouds, but also identify anomalous point clouds with characteristics such as severe reflection and sudden velocity changes, which helps to warn of potential security risks or atypical behaviors and enhance the system's intelligent monitoring and response capabilities.

[0127] In this embodiment, step S5, which identifies the local point cloud density of the target distribution point cloud and dynamically estimates the corresponding number of targets based on the local point cloud density, further includes:

[0128] S51: Identify continuous frame point clouds from the target distribution point cloud, perform temporal fusion of the continuous frame point clouds, and generate a corresponding high-density point cloud region.

[0129] S52: Determine whether the high-density region of the point cloud appears repeatedly within a preset frequency;

[0130] S53: If so, obtain the location information of the high-density region of the point cloud, extract the preset micro-motion spectrum features from the high-density region of the point cloud, and dynamically classify the target type of the continuous frame point cloud based on the location information and the micro-motion spectrum features. The micro-motion spectrum features specifically include breathing frequency and arm swing frequency, and the target type specifically includes dynamic human body, interference object and non-human body object.

[0131] In this embodiment, the system identifies continuous frame point clouds from the target distribution point cloud, performs temporal fusion on these continuous frame point clouds to generate corresponding high-density point cloud regions, and then determines whether the high-density point cloud region recurs within a pre-set frequency to execute corresponding steps. For example, when the system determines that the high-density point cloud region does not recur within the pre-set frequency, the system considers that the point cloud changes in this region may not have obvious stability or periodicity, and there may be no continuous target present. The system will identify whether there are large changes or sudden fluctuations in the point cloud in this region. If the high-density region does not recur within the predetermined frequency, it indicates that the target may have moved out of the current detection range, or the target may be masked by noise. At the same time, for high-density regions that do not recur, the system needs to update the target state, perform target tracking or state update for leaving the current region, which involves adjusting the target's motion model. The system predicts paths to avoid mistakenly tracking a non-existent or inactive target. If the high-density area does not recur, the system needs to further filter out potential environmental noise and eliminate false alarms. By dynamically adjusting the sensor sensitivity and optimizing the radar signal processing algorithm, the system can reduce interference from irrelevant factors on target detection and improve the system's sensitivity and accuracy for effective targets. For example, when the system determines that a high-density area in the point cloud recurs within a pre-set frequency, the system will consider that the point cloud changes in that area have obvious stability or periodicity and that a continuous target exists. The system will obtain the location information of the high-density area in the point cloud and extract pre-set micro-motion spectrum features from the high-density area in the point cloud. The micro-motion spectrum features specifically include breathing frequency and arm swing frequency. Based on the location information and micro-motion spectrum features, the system dynamically classifies the target types of continuous frame point clouds. The target types specifically include dynamic human bodies, interference objects, and non-human objects.By extracting micro-motion spectral features, the system can effectively distinguish dynamic human beings from interference objects or non-human objects. For example, human breathing frequency has specific low-frequency stable characteristics, while arm swing frequency reflects a rhythmic movement pattern. These features are clearly distinguishable in the spectrum. Compared to methods that rely solely on static morphology or positional point clouds, this method of fusing micro-motion features significantly improves the accuracy of identifying dynamic human beings, avoiding misidentification of objects such as leaves blowing in the wind or curtains swaying as human targets. Furthermore, when high-density point cloud regions repeatedly appear in the frequency spectrum, the system can determine that the target is a long-term active or stationary object, rather than a transient disturbance. This discrimination based on frequency stability is helpful in… The system prioritizes processing resources for real, continuous human targets, improving overall monitoring efficiency and real-time response capabilities. This makes it particularly suitable for applications such as personnel lingering monitoring and security deployment. Furthermore, by integrating location information and micro-motion spectrum data, the system can dynamically classify targets in continuous frame point clouds, such as identifying moving human figures, periodic interference sources (e.g., fans), and stationary non-human objects (e.g., pillars, sofas). This classification method based on "target behavior + micro-motion features" gives the system a stronger target understanding capability, propelling it from "point cloud detection" to "semantic understanding," providing a solid foundation for subsequent behavior recognition, intelligent decision-making, and early warning control.

[0132] In this embodiment, step S2, which determines whether the echo signal detects a peak value of the combined signal, further includes:

[0133] S21: Obtain the full width at half maximum (FWHM) of the main peak of the echo signal;

[0134] S22: Determine whether the full width at half maximum (FWHM) of the main peak exceeds a preset multiple of the width of a historical single-target waveform;

[0135] S23: If so, detect several small peaks of the full width at half maximum (FWHM) of the main peak, identify the integrated energy on the left and right sides of the small peaks, and dynamically collect the superimposed waveforms caused by the reflection amplitudes of each target based on the symmetry of the integrated energy. Specifically, the small peaks include shoulder peak sub-valleys, plateau-shaped waveform sub-valleys, and inter-peak sub-valleys.

[0136] In this embodiment, the system acquires the full width at half maximum (FWHM) of the echo signal and then determines whether the FWHM exceeds a preset multiple of the historical single-target waveform width to execute corresponding steps. For example, when the system determines that the FWHM of the echo signal does not exceed a preset multiple of the historical single-target waveform width, the system considers that the current echo signal may originate from a relatively concentrated single target, whose signal waveform is relatively stable and has not undergone excessive expansion or overlap. The system confirms that the current echo signal originates from a single target and performs single-target tracking and analysis, extracting relevant information about the target, such as position, velocity, and trajectory, to further predict the target's motion trend and possible behavioral changes. Simultaneously, since the echo signal indicates that the target is single and clear... The system selects and optimizes the target detection algorithm to improve its tracking accuracy. For example, it uses a Kalman filter or other tracking algorithms to accurately track the target, avoiding misjudgments caused by multi-target interference. The system can reduce the need for complex calculations, focusing on the tracking and analysis of a single target, thus improving efficiency. Furthermore, when the system confirms that the echo signal originates from a single target, it can appropriately increase its tolerance to signal noise and interference, ignoring the impact of background noise interference on the signal, thereby further improving the accuracy and stability of detection. In cases with less interference, the system can more clearly analyze the behavioral characteristics of the target; for example, when the system determines that the full width at half maximum (FWHM) of the echo signal exceeds a preset multiple of the historical single-target waveform width, the system will consider the current echo... The signal originates from multiple dispersed targets. The system detects several smaller peaks at half-width (FWHM) of the main peak. These smaller peaks include secondary valleys of shoulder peaks, plateau-shaped waveforms, and inter-peak valleys. The system identifies the integrated energy on both sides of these smaller peaks. Based on the symmetry of different integrated energies, it dynamically acquires the superimposed waveforms caused by the reflection amplitudes of each target. By finely identifying the smaller peaks in the main peak (such as secondary valleys of shoulder peaks and plateau-shaped waveforms) and analyzing the integrated energy on both sides, the system can effectively distinguish multiple reflection sources in the superimposed signal. Compared to directly identifying a wide main peak as a "fuzzy target," this method can more precisely distinguish multiple approaching targets, significantly improving the accuracy of multi-target analysis and avoiding the omission or misidentification of individual targets due to waveform aliasing. Furthermore, based on the various... By analyzing the symmetry of the integrated energy on both sides of the small peak, the system can dynamically infer the reflection amplitude of each target in the composite waveform. This approach helps to reconstruct the independent echo characteristics of each target, enabling the system to extract the true reflection intensity, possible material differences, or body shape distribution characteristics of each target. This improves the ability to judge the target category, shape, and behavior in subsequent identification. Furthermore, in complex scenarios where multiple targets are close together, overlap, or frequently occlude (such as crowds approaching, walking through corridors, or multiple people converging), the system can not only maintain the continuity of echo recognition by refining the processing after the full width at half maximum (FWHM) of the main peak exceeds the limit, but also dynamically adjust the number and distribution information of targets in the perception model, thereby improving the stability and robustness of the system for multi-target tracking and classification in dynamic environments.

[0137] It should be noted that by detecting several smaller peaks in the full width at half maximum (FWHM) of the main peak, identifying the integrated energy on both sides of these smaller peaks, and based on the symmetry of the integrated energy, dynamically acquiring the superimposed waveform caused by the reflection amplitude of each target, a specific example is as follows:

[0138] Suppose a system monitors multiple moving targets in an environment and uses radar signals for target detection; suppose that at a certain moment, the waveform of the echo signal received by the system exhibits a main peak, and several smaller peaks are found around the main peak (such as secondary valleys of shoulder peaks, secondary valleys of plateau waveforms, and secondary valleys between peaks); the following step-by-step analysis explains how the system separates multiple targets based on this information and dynamically collects the reflection amplitude of each target;

[0139] Steps and analysis, calculating the full width at half maximum (FWHM) of the main peak:

[0140] The system first calculates the full width at half maximum (FWHM) of the main peak of the echo signal. If the width of this main peak is significantly greater than a preset multiple of the width of the historical single-target waveform, the system assumes that the signal may be a reflection from multiple targets. For example, if the width of the historical single-target waveform is 1 unit, and the current full width at half maximum (FWHM) of the main peak is 2 units, which exceeds the preset multiple of the historical width (e.g., set to 1.5 times), this indicates that the signal may be a superposition of echoes from multiple targets.

[0141] Identify small peak values:

[0142] Next, the system detects multiple smaller peaks around the main peak; assuming these smaller peaks are secondary valleys of shoulder peaks, secondary valleys of plateau-shaped waveforms, and secondary valleys between peaks, the system will identify the possible sources of the target's echo signal from them;

[0143] Shoulder peak and secondary valley: refers to a small peak that appears to the left of the main peak. The system found that the reflected energy distribution of this small peak is relatively concentrated.

[0144] A plateau-shaped waveform trough refers to a stable region between the main peak and the wave peak, where a gentle small wave peak appears, indicating that there may be a continuous fluctuation of a reflection source here;

[0145] Inter-peak valleys: These refer to the valley points between two peaks, which may correspond to the overlap of echo signals between two targets. The system needs to further analyze their characteristics.

[0146] Calculate the integral energy:

[0147] For each small peak, the system calculates the integrated energy on its left and right sides, that is, calculates the reflection intensity distribution of the waveform;

[0148] For the shoulder peak and the secondary valley, assuming the system calculation finds that the integral energy on the left side of the wavelet peak is 0.8, while the integral energy on the right side is 0.2; this indicates that the echo of this small peak may come from a distant target.

[0149] For the plateau-shaped waveform trough, the system found that the integrated energy on the left and right sides was basically equal (0.5), which means that it may come from a relatively uniform reflection source, such as a human body or object at close range.

[0150] For the secondary valleys between the peaks, the system found a large difference in the integrated energy between the left and right sides (0.7 on the left and 0.3 on the right), which indicates that the echo signals reflected by the two targets may overlap and have a large distance difference.

[0151] Analyze the symmetry of the integral energy:

[0152] The system further analyzes the energy symmetry of each small peak:

[0153] For the shoulder peak and the secondary valley, due to energy asymmetry, the system believes that the small peak may come from a distant target;

[0154] For the second trough of the plateau-shaped waveform, because the energy on the left and right sides is symmetrical, the system assumes that the echo signal may come from a nearby target with uniform reflection.

[0155] For the secondary valleys between wave crests, due to the large energy difference between the left and right sides, the system considers them to be the superposition of echo signals from two targets.

[0156] Dynamically acquire reflection amplitude:

[0157] Based on the analysis of integral energy, the system dynamically adjusts the acquisition of reflection amplitude for each target:

[0158] For targets with acromion and valley, when the system dynamically adjusts the amplitude of the collected reflections, the target is set as a distant target and the estimation of its reflection amplitude is reduced;

[0159] For targets with plateau-shaped waveforms and secondary valleys, the system considers their reflection amplitude to be large, and therefore increases the estimated reflection amplitude of the target.

[0160] For targets with secondary valleys between wave crests, the system dynamically calculates the reflection amplitude of the two targets and separates their echoes;

[0161] Target echo separation and reconstruction:

[0162] The system can effectively separate the echoes of multiple targets and generate independent waveforms for each target; for example, it can separate the echo waveform of a distant target from the echo of a sub-valley of a shoulder peak; it can separate the echo waveform of a near target from the echo of a sub-valley of a plateau waveform; and it can separate the echo waveforms of two targets from the echo of a sub-valley between peaks, representing them as two independent targets. Based on the dynamically acquired reflection amplitude information, the system can further track the motion trajectory of each target and identify and classify the target type (such as dynamic human body, static obstacle, etc.).

[0163] In summary, as illustrated above, the system can separate the reflection waveforms of multiple targets from complex echo signals and dynamically acquire the reflection amplitude of each target through integral energy symmetry analysis, thereby reconstructing the echo signals of multiple targets. This process greatly improves the accuracy and real-time performance of radar signal processing. Especially in multi-target environments, it can effectively avoid target echo aliasing and provide reliable basic data for subsequent target tracking and classification.

[0164] In this embodiment, step S4, which determines whether the angular direction of the target distribution point cloud forms a continuous region, further includes:

[0165] S41: Sort the target distribution point cloud according to a preset angle value, construct a corresponding angle sequence by sorting, and calculate the angle increment of the angle sequence;

[0166] S42: Determine whether the angle increment has increased;

[0167] S43: If so, identify the break point after the angle increment increases, perform clustering and cutting at the break point as the separation boundary of the point cloud target, divide the point cloud target into several continuous segments based on the position of the break point, and use the several continuous segments as the angle region of a single candidate target.

[0168] In this embodiment, the system sorts the target distribution point cloud according to preset angle values, constructs corresponding angle sequences, calculates the angle increments of these angle sequences, and then determines whether these angle increments have increased to execute corresponding steps. For example, when the system determines that the angle increments of the angle sequences have not increased, the system considers the target distribution to be relatively uniform or to have a large stable region. The system sorts the target distribution point cloud according to the angle values ​​to obtain a corresponding angle sequence. Assuming that at a certain moment, the target point cloud distribution is relatively uniform and the interval between each point in the angle sequence is small, the system calculates the angle increments of these angle sequences, i.e., the differences between adjacent angle values. If the target distribution is relatively uniform, the angle increments will be relatively consistent with little difference. The calculated angle increments reflect the target's spatial distribution. If these angle increments remain consistent or change little, it indicates a relatively uniform target distribution, and the system can accurately determine the target's distribution characteristics. For example, within a relatively uniform area, the target may exhibit continuous reflected signals with small changes in angle increments. Furthermore, if the angle increments remain within a certain range without increasing, it indicates a relatively stable target distribution. The system will not identify significant fluctuations or discrete point clouds in the target distribution; in other words, the relative positions of the targets will change little, without dense areas or scattered variations. In such cases, the system assumes the targets are in a relatively stable or uniformly distributed state, potentially eliminating the need for further complex processing such as target separation or noise removal. For instance, when the system detects a significant increase in the angle increment of the angle sequence, it assumes the target distribution is concentrated, exhibiting dense areas or scattered changes. The system identifies the breakpoints where the angle increment increases, performs clustering at these breakpoints as separation boundaries for the point cloud targets, and divides the point cloud targets into several continuous segments based on the locations of these breakpoints. These continuous segments are then used as the angle regions of individual candidate targets. By identifying breakpoints in the angle sequence where the angle increment significantly increases, the system can accurately locate the targets within the point cloud. The transition boundary effectively separates multiple targets that are close to or partially overlap each other, avoiding target confusion and improving the accuracy of subsequent classification and recognition. At the same time, it uses the abrupt change characteristics of angle increment as the basis for clustering and segmentation, so that the system can still flexibly divide the point cloud data into regions when there are multiple targets, especially when there are dense distributions and uneven spacing at angles. This enhances the adaptability of point cloud processing to complex distributions. Furthermore, by dividing the entire point cloud into several continuous segments as candidate target regions, the system can process each segment more independently and in a more targeted manner, such as dynamically setting filtering parameters and assigning tracking numbers, thereby reducing the computational burden on irrelevant regions and improving the overall operating efficiency and real-time performance of the system.

[0169] It should be noted that the breakpoints after the angle increment increases are identified, and clustering and segmentation are performed at these breakpoints as separation boundaries for point cloud targets. Based on the positions of the breakpoints, the point cloud targets are divided into several continuous segments, and these continuous segments are used as the angular regions of a single candidate target. A specific example is as follows:

[0170] Suppose there is a radar system monitoring an indoor space. The radar emits signals and receives reflected point cloud data; each point cloud represents the reflected signal of a target, containing the target's angle information relative to the radar.

[0171] Step 1: Sorting and Constructing the Angle Sequence. First, the system sorts the point cloud data according to the angle values ​​of the target in space based on the reflected signals returned by the radar. Assume the system obtains the following angle data (unit: degrees):

[0172] [12.0, 12.2, 12.4, 12.5, 12.7, 20.1, 20.3, 20.5, 20.6, 30.0, 30.2, 30.3], these angle values ​​represent the angle information of multiple target reflected signals;

[0173] Step 2, Calculate the angle increment. Next, the system calculates the angle increment (angle difference) between every two adjacent points in the angle sequence:

[0174] Incremental sequence: [0.2, 0.2, 0.1, 0.2, 7.4, 0.2, 0.2, 0.1, 9.4, 0.2, 0.1]

[0175] Step 3: Identify breakpoints. The system sets a threshold (e.g., an increment exceeding 1.0° is considered a breakpoint) and identifies breakpoints based on the angle increment sequence.

[0176] The angle increment between the 5th increment (12.7→20.1) is 7.4°, which is much greater than 1.0°, so this is identified as the first break point;

[0177] The angle increment between the 9th increment (20.6→30.0) is 9.4°, which is much greater than 1.0°, so this is identified as the second break point;

[0178] Step 4, clustering and segmentation: Based on the identified breakpoints (at the 5th and 9th increments), the system segments the point cloud data into 3 continuous segments:

[0179] Segment 1 (Target 1): From the angles [12.0, 12.2, 12.4, 12.5, 12.7], this segment has relatively close angles and belongs to the first target area;

[0180] Segment 2 (Target 2): From the angles [20.1, 20.3, 20.5, 20.6], the angle values ​​in this segment are relatively close, belonging to the second target area;

[0181] Segment 3 (Target 3): From the angles [30.0, 30.2, 30.3], the angle intervals in this segment are slightly larger, but it still falls within the angle range of a single target;

[0182] Step 5, Target Segmentation and Subsequent Processing: After target segmentation, the system treats each consecutive angle segment as a separate candidate target; for example:

[0183] Target 1: Located in the angle range [12.0, 12.7], this may represent the first target in the room (e.g., a person standing on one side of the room);

[0184] Target 2: Located in the angle range [20.1, 20.6], representing another target (e.g., another person in the room);

[0185] Target 3: Located in the angle range [30.0, 30.3], representing a distant target (such as an object or person in another corner);

[0186] Step 6, Result Analysis and Application: By dividing these angular regions, the system can separate each target from other targets and apply these angular regions as input in subsequent target tracking or analysis. This enables the system to accurately: improve the accuracy of target separation, especially in environments with multiple densely distributed targets; in real-time monitoring and tracking, perform continuous tracking and analysis based on these angular regions to avoid confusion between different targets; and in dynamic target identification, further combine with other sensor information (such as distance, speed, etc.) to optimize the accuracy of target identification.

[0187] In summary, the above examples demonstrate that by incrementally calculating the angle sequence and identifying breakpoints where the angle increment increases, the system can effectively identify the boundaries between multiple targets and segment the point cloud data into multiple target regions. This method not only helps improve the target separation capability in multi-target scenarios but also provides more accurate data input for subsequent target recognition, tracking, and classification.

[0188] In this embodiment, step S1, which involves repeatedly outputting and receiving the echo signal of the detection scenario within a preset time period based on a preset detection scenario of the microwave radar, further includes:

[0189] S11: Based on the preset scene type of the detection scene, adaptively switch the beam coverage strategy of the microwave radar, wherein the scene type specifically includes corridor, room and doorway, and the beam coverage strategy specifically includes omnidirectional, directional and variable beam.

[0190] S12: Determine whether the beam coverage strategy can meet the preset target dynamic detection;

[0191] S13: If not, then according to the preset detection parameters of the microwave radar, the preset static shielding area is dynamically activated to block the obstacle echo pre-collected by the microwave radar through the static shielding area. The detection parameters specifically include the horizontal scanning angle range, the vertical coverage range, and the effective distance.

[0192] In this embodiment, the system adaptively switches the beam coverage strategy of the microwave radar based on a pre-defined scene type, specifically including corridors, rooms, and doorways. The beam coverage strategy includes omnidirectional, directional, and variable beams. The system then determines whether these beam coverage strategies can meet the pre-defined target dynamic detection requirements and executes the corresponding steps. For example, when the system determines that the microwave radar's beam coverage strategy can meet the pre-defined target dynamic detection requirements, the system considers that the current radar beam configuration has effectively covered the area where the target may appear or move, and has sufficient detection sensitivity and spatial resolution to stably detect the target in the field. In the context of dynamic behavior perception, the system maintains its current omnidirectional, directional, or variable beam coverage to ensure continuous and stable detection performance. Simultaneously, if both detection requirements and energy conservation are needed, it can switch to low-frequency or intermittent scanning modes to reduce energy consumption. Furthermore, provided the beam coverage requirements are met, the system activates subsequent fine-grained tracking functions, such as human posture estimation and micro-motion recognition (e.g., breathing, gait), to further uncover dynamic behavioral characteristics. For example, if the system determines that the microwave radar's beam coverage strategy cannot meet the pre-set target dynamic detection requirements, it will conclude that the current radar beam configuration cannot effectively cover the area where the target may appear. The system dynamically activates pre-set static shielding zones based on pre-defined detection parameters of the microwave radar, including the horizontal scanning angle range, vertical coverage range, and effective distance. These static shielding zones block obstacle echoes pre-collected by the microwave radar. By activating the static shielding zones, the system blocks echo signals from non-target areas, avoiding interference from obstacles. By setting the horizontal scanning angle range, vertical coverage range, and effective distance, the system can accurately locate the radar detection area, reducing unnecessary interference areas and making the radar's detection range more focused. This allows for better coverage of the area required for dynamic target detection. Simultaneously, the introduction of static shielding zones effectively reduces the impact of unnecessary obstacle echo signals on the radar system. By blocking obstacle echoes collected by the radar, these echo signals are prevented from being confused with target echoes, ensuring the system can focus on target detection. This is especially important when target detection conditions change or the environment is complex. Dynamically activating static shielding zones means the radar can adaptively adjust to real-time environmental changes, thereby improving the system's stability and accuracy in complex environments. By shielding interference signals, the system can more accurately capture target signals, especially in situations with multiple targets or complex backgrounds, contributing to improved detection reliability.

[0193] Reference Appendix Figure 2 A real-time human detection system based on microwave radar, as described in one embodiment of the present invention, includes:

[0194] The receiving module 10 is used to repeatedly output and receive the echo signal of the detection scene within a preset time period based on the preset detection scene of the microwave radar.

[0195] The judgment module 20 is used to determine whether the echo signal detects a peak value of the merged signal;

[0196] The execution module 30 is configured to, if so, construct a corresponding sparse reconstruction model based on the pre-acquired original radar signal as superimposed data of a sparse source, decompose the aliased signal into corresponding sparse units using a preset sparse coding, extract multi-dimensional features from the sparse units, and establish a target distribution point cloud of the multi-dimensional features through a preset three-dimensional feature space, wherein the multi-dimensional features specifically include range features, Doppler features, and angle features;

[0197] The second judgment module 40 is used to determine whether the angular direction of the target distribution point cloud forms a continuous region;

[0198] The second execution module 50 is used to identify the local point cloud density of the target distribution point cloud if no target distribution point cloud is formed, dynamically estimate the corresponding number of targets based on the local point cloud density, apply a preset Kalman filter to continuously track the target parameters of each target, generate the separated multi-target echo, and reconstruct the multi-target echo into a pseudo-color radar image. The target parameters specifically include target velocity and target velocity change.

[0199] In this embodiment, the receiving module 10 repeatedly outputs and receives echo signals from a pre-set detection scenario based on the microwave radar within a pre-defined time period. Then, the judging module 20 determines whether these echo signals detect a merged signal peak, and executes corresponding steps accordingly. For example, if the system determines that the repeatedly output and received echo signals from the detection scenario do not detect a merged signal peak, the system assumes that there is a certain distance separation between the various human targets, and that the radar echoes do not exhibit significant overlap or aliasing in time, frequency, and angle dimensions. The system will continue to maintain the current output and reception cycle, increasing the detection capability for weak or distant targets, and consider adjusting the radar's receiving gain or changing the transmission power to enhance the detection of long-range or small-target targets. The system boasts strong target detection capabilities. Since there's no signal peak merging, the targets corresponding to the echo signals are relatively dispersed. The system can directly process each independent echo, extracting multi-dimensional features (such as range, velocity, and Doppler effect) and performing target identification and localization in a three-dimensional feature space. Furthermore, if the number of targets is small, the system can identify each target individually using a simple separation method based on range and Doppler features, ensuring the accurate position and motion state of each target. Based on the target distribution, the system dynamically adjusts the radar system's operating parameters (such as scanning angle and transmission frequency) to better adapt to scene requirements. If the system detects a relatively static detection area, it can also reduce the output frequency to decrease power consumption. For example... When the system detects repeated output and receives echo signals from the detection scene, detecting a peak in the merged signal, the execution module 30 assumes that the distances between the various human targets are close, easily causing radar echo overlap or aliasing. The system uses pre-collected raw radar signals as superimposed data from sparse sources to construct a corresponding sparse reconstruction model. It then uses pre-defined sparse coding to decompose the aliased signal into corresponding sparse units, extracting multi-dimensional features from these units. These multi-dimensional features include range features, Doppler features, and angle features. A target distribution point cloud with multi-dimensional features is established using a pre-defined three-dimensional feature space. The system treats the raw radar signals as superimposed data from multiple sparse sources and employs sparse... The reconstructed model deconstructs the signal, effectively separating the merged signal into sparse units with identifiable features. This overcomes the limitation of traditional methods in detecting multiple people at close range, which cannot distinguish individuals. At the same time, it extracts three-dimensional features of distance, Doppler, and angle from the sparse units, forming richer target information. This is beneficial for distinguishing targets from multiple dimensions, especially when targets have similar shapes and directions of movement, thus improving the accuracy of target classification and recognition. Furthermore, the target distribution point cloud established through the three-dimensional feature space can intuitively show the relative position and density relationship of targets in space, providing a basic data structure for subsequent operations such as target quantity estimation, trajectory separation, and dynamic tracking, and enhancing the system's multi-target management capabilities in crowded or occluded scenarios.Then, the second judgment module 40 determines whether the angular directions of these target distribution point clouds form a continuous region, and executes the corresponding steps accordingly. For example, when the system determines that the angular directions of the target distribution point clouds form a continuous region, the system will consider that relatively complete target reflection information has been detected in this angular region, which can more accurately represent the directional expansion range of the target in space. The system will adjust the echo signal processing method to merge the echo signals in the continuous region to avoid unnecessary signal separation. At this time, the target recognition strategy should focus on the analysis of group behavior rather than individual target separation in order to improve processing efficiency. At the same time, based on density analysis, motion patterns, and multi-dimensional features, the number of targets is dynamically estimated, and appropriate classification algorithms (such as cluster-based classification methods) are used to classify and identify these targets. Through the dynamic positional changes of the targets, the relationship between targets can be judged more accurately, avoiding... To avoid misidentification in multi-target scenarios, and for occasional continuous angular regions, the system should carefully eliminate noise interference. By setting reasonable thresholds and signal enhancement mechanisms, it can ensure that the continuity of target distribution is not due to misjudgment caused by interference signals or environmental noise. For example, when the system determines that the angular direction of the target distribution point cloud does not form a continuous region, the second execution module 50 will consider that no complete target reflection information has been detected in this angular region, and the directional expansion range of the target in space cannot be represented. The system will identify the local point cloud density of the target distribution point cloud, dynamically estimate the number of human targets based on different local point cloud densities, apply a pre-set Kalman filter, continuously track the target parameters of each human target, including target velocity and target velocity changes, generate separated multi-target echoes, and reconstruct these multi-target echoes into a pseudo-color radar image.By further analyzing the local point cloud density of the target distribution point cloud, the system can capture multiple high-density reflective sub-regions even when there is a lack of extension in the angular direction, thereby dynamically estimating the actual number of targets. This density-based supplementation mechanism significantly improves the system's robustness in weak signal areas or under non-ideal reflection conditions. At the same time, when multiple targets are close to each other or move at similar speeds, their radar echoes often overlap in time, frequency, and angular dimensions, making it difficult to distinguish instantaneous targets. In this case, a Kalman filter is introduced, based on the historical motion state of the target and combined with new observation data, to continuously estimate and correct the velocity and velocity change parameters of each target. This not only effectively avoids the interruption of target trajectory due to short-term aliasing, but also maintains the consistency of target trajectory when partially obscured or missing signals. After completing the sparse reconstruction, feature extraction, point cloud generation, and target tracking of multiple targets, the system synthesizes the separated target echo signals into a pseudo-color radar image. Each target is assigned a different color code according to its characteristic parameters. This image-based approach improves the readability of the system results and the efficiency of human-computer interaction, allowing end users to intuitively identify the number, location, and direction of movement of targets. ;

[0200] In this embodiment, the execution module further includes:

[0201] The computing unit is used to calculate the amplitude squared integral of the sparse unit, normalize the amplitude squared integral to a preset standard scale, apply a preset continuous wavelet transform to the sparse unit, and generate a corresponding time-frequency spectrum, wherein the time-frequency spectrum specifically includes the dominant frequency, bandwidth and frequency change rate.

[0202] The judgment unit is used to determine whether the time-frequency spectrum detects a preset low-frequency periodic motion feature, wherein the low-frequency periodic motion feature specifically includes breathing and heartbeat;

[0203] An execution unit is configured to, if so, acquire the target distance corresponding to the echo delay based on the echo delay of the sparse unit, statistically analyze the internal point distribution pattern of the sparse unit according to the target distance, and track the time trajectory evolution of the sparse unit within a preset time window, wherein the internal point distribution pattern specifically includes the principal axis length, flatness, and density distribution characteristics.

[0204] In this embodiment, the system calculates the amplitude squared integral of the sparse units, normalizes these amplitude squared integrals to a pre-set standard scale, applies a pre-set continuous wavelet transform to these sparse units, and generates a corresponding time-frequency spectrum. The time-frequency spectrum specifically includes the dominant frequency, bandwidth, and rate of change of frequency. Then, the system determines whether the time-frequency spectrum detects pre-set low-frequency periodic motion features, specifically including breathing and heartbeat, to execute corresponding steps. For example, if the system determines that the time-frequency spectrum does not detect the pre-set low-frequency periodic motion features, the system will consider that multiple human micro-motion signals are superimposed in the sparse units, leading to... If energy diffusion or waveform distortion occurs in the frequency band, making it impossible to clearly identify low-frequency features, the system will temporarily classify the reflective target corresponding to the current sparse cell as a "non-biological target" or "inactive target." It will then combine position and velocity characteristics for target classification and filtering to avoid misidentifying them as interactive or alarm-requiring human objects. If the target is located in the region of interest (such as a bed, sofa, or monitoring area), the analysis period will be automatically extended (e.g., from 5 seconds to 10 seconds) for a second time-frequency analysis to improve weak signal detection capabilities. Furthermore, if no low-frequency periodic features are detected in multiple sparse cells, the system may attempt to adjust the radar gain, antenna beam, or activate auxiliary channels (such as vertical beamforming) to improve detection sensitivity. If such "targets without low-frequency characteristics" appear repeatedly, they should be marked and uploaded to the cloud or backend database as part of behavioral risk identification (e.g., sudden fainting or abnormal stillness). For example, when the system determines that a pre-set low-frequency periodic motion feature has been detected in the time-frequency spectrum, the system will consider the radar signal quality to be good, the micro-motion components (such as breathing, heartbeat, etc.) not masked by noise or interference, the system can clearly distinguish the target's vital signs, and the signal processing effect is good. The system will collect the target distance corresponding to the echo delay based on the sparse unit echo delay, and statistically analyze the internal point distribution pattern of the sparse unit according to different target distances. The morphology specifically includes the principal axis length, flatness, and density distribution characteristics. Within a pre-set time window, the system tracks the temporal trajectory evolution of these sparse units. Based on the echo delay of the sparse units, the system can collect the delay information of the echo signal, thereby accurately calculating the target distance. Through echo delay, the system can determine the target's position. Especially when the target is at different distances, the system can dynamically adjust its detection range to further improve tracking accuracy. At the same time, by statistically analyzing the internal point distribution morphology of the sparse units (including principal axis length, flatness, density distribution characteristics, etc.), the system can further reveal the target's morphological characteristics. For example, the target's density distribution characteristics help to infer the target's size, shape, and motion state.Changes in the spindle length and flatness can be used to determine the target's trajectory and attitude. Within a set time window, the system can track the temporal evolution of sparse units. By continuously tracking the target's trajectory, the system can acquire the target's motion trend (such as changes in speed and direction) in real time, enabling dynamic monitoring and prediction of the target. This is particularly important for applications such as smart homes, health monitoring, and security systems.

[0205] In this embodiment, it also includes:

[0206] The identification module is used to identify the number of neighboring points of each point cloud within a preset radius based on the point cloud density of the target distribution point cloud.

[0207] The third judgment module is used to determine whether the number of neighbor points is less than a preset value;

[0208] The third execution module is used to detect extreme outliers of the multidimensional features if the condition is met, evaluate the dispersion of each point cloud based on the extreme outliers, statistically analyze the topological characteristics of the target distribution point cloud, and add additional features to each point cloud. Specifically, the extreme outliers include reflection energy supermassive and frequency jumps, the topological characteristics include the main direction of the point cloud and the overall shape of the point cloud, and the additional features include local curvature and local velocity gradient.

[0209] In this embodiment, the system identifies the number of neighboring points within a pre-defined radius for each point cloud based on the point cloud density of the target distribution point cloud. Then, the system determines whether the number of these neighboring points is less than a pre-defined value to execute corresponding steps. For example, if the system determines that the number of neighboring points within the pre-defined radius for each point cloud is not less than the pre-defined value, the system considers the current point cloud to have high local density and stability, and is not a discrete noise point. Therefore, there is no need to remove noise points. The system uses a clustering algorithm to cluster these dense point cloud regions, distinguishing different targets. The clustering results help the system identify different targets and simultaneously... Clustered point clouds, by extracting features such as geometric shape, boundaries, and centroid location, can be further analyzed to determine the target's orientation, volume, and morphology. Furthermore, if time-series data is available, the system can calculate the target's motion parameters, such as velocity, acceleration, and direction, based on its trajectory. Kalman filtering and other algorithms can then be used to track the target and predict its future position. Based on the target's geometric features and dynamic changes, classification algorithms (such as deep learning models and support vector machines) can be used for target recognition, identifying whether the target is a pedestrian, a stationary object, or another moving object. For example, when the system determines that each point cloud point is within a pre-defined radius... If the number of neighboring points is less than a preset value, the system considers the current point cloud to be unstable and may belong to discrete noise points, requiring noise point removal. The system detects extreme outliers in multi-dimensional features, specifically including excessive reflection energy and frequency jumps. Based on these extreme outliers, the system assesses the dispersion of each point cloud and statistically analyzes the topological characteristics of the target distribution point cloud, including the main orientation and overall shape of the point cloud. Additional features are added to each point cloud, including local curvature and local velocity gradient. By determining the number of neighboring points within a preset radius and combining this with extreme outliers in reflection energy and frequency, the system can more effectively... The system accurately distinguishes between valid target points and discrete noise points, thereby avoiding the accidental deletion of edge targets or weakly reflective targets, improving the purity and reliability of point cloud data. At the same time, it further analyzes the topological characteristics of the point cloud (such as main direction and overall shape) to extract additional features such as local curvature and velocity gradient, which helps to accurately model the geometry and motion morphology of the target, improve the multi-dimensional feature recognition capability of the target and the stability of continuous tracking. Furthermore, with the help of multi-dimensional outlier detection and topological analysis, the system can not only remove invalid point clouds, but also identify anomalous point clouds with characteristics such as severe reflection and sudden velocity changes, which helps to warn of potential security risks or atypical behaviors and enhance the system's intelligent monitoring and response capabilities.

[0210] In this embodiment, the second execution module further includes:

[0211] The generation unit is used to identify continuous frame point clouds from the target distribution point cloud, perform temporal fusion of the continuous frame point clouds, and generate a corresponding high-density point cloud region.

[0212] The second judgment unit is used to determine whether the high-density region of the point cloud recurs within a preset frequency.

[0213] The second execution unit is configured to, if so, acquire the position information of the high-density region of the point cloud, extract preset micro-motion spectrum features from the high-density region of the point cloud, and dynamically classify the target type of the continuous frame point cloud based on the position information and the micro-motion spectrum features. Specifically, the micro-motion spectrum features include breathing frequency and arm swing frequency, and the target type specifically includes dynamic human body, interference object and non-human body object.

[0214] In this embodiment, the system identifies continuous frame point clouds from the target distribution point cloud, performs temporal fusion on these continuous frame point clouds to generate corresponding high-density point cloud regions, and then determines whether the high-density point cloud region recurs within a pre-set frequency to execute corresponding steps. For example, when the system determines that the high-density point cloud region does not recur within the pre-set frequency, the system considers that the point cloud changes in this region may not have obvious stability or periodicity, and there may be no continuous target present. The system will identify whether there are large changes or sudden fluctuations in the point cloud in this region. If the high-density region does not recur within the predetermined frequency, it indicates that the target may have moved out of the current detection range, or the target may be masked by noise. At the same time, for high-density regions that do not recur, the system needs to update the target state, perform target tracking or state update for leaving the current region, which involves adjusting the target's motion model. The system predicts paths to avoid mistakenly tracking a non-existent or inactive target. If the high-density area does not recur, the system needs to further filter out potential environmental noise and eliminate false alarms. By dynamically adjusting the sensor sensitivity and optimizing the radar signal processing algorithm, the system can reduce interference from irrelevant factors on target detection and improve the system's sensitivity and accuracy for effective targets. For example, when the system determines that a high-density area in the point cloud recurs within a pre-set frequency, the system will consider that the point cloud changes in that area have obvious stability or periodicity and that a continuous target exists. The system will obtain the location information of the high-density area in the point cloud and extract pre-set micro-motion spectrum features from the high-density area in the point cloud. The micro-motion spectrum features specifically include breathing frequency and arm swing frequency. Based on the location information and micro-motion spectrum features, the system dynamically classifies the target types of continuous frame point clouds. The target types specifically include dynamic human bodies, interference objects, and non-human objects.By extracting micro-motion spectral features, the system can effectively distinguish dynamic human beings from interference objects or non-human objects. For example, human breathing frequency has specific low-frequency stable characteristics, while arm swing frequency reflects a rhythmic movement pattern. These features are clearly distinguishable in the spectrum. Compared to methods that rely solely on static morphology or positional point clouds, this method of fusing micro-motion features significantly improves the accuracy of identifying dynamic human beings, avoiding misidentification of objects such as leaves blowing in the wind or curtains swaying as human targets. Furthermore, when high-density point cloud regions repeatedly appear in the frequency spectrum, the system can determine that the target is a long-term active or stationary object, rather than a transient disturbance. This discrimination based on frequency stability is helpful in… The system prioritizes processing resources for real, continuous human targets, improving overall monitoring efficiency and real-time response capabilities. This makes it particularly suitable for applications such as personnel lingering monitoring and security deployment. Furthermore, by integrating location information and micro-motion spectrum data, the system can dynamically classify targets in continuous frame point clouds, such as identifying moving human figures, periodic interference sources (e.g., fans), and stationary non-human objects (e.g., pillars, sofas). This classification method based on "target behavior + micro-motion features" gives the system a stronger target understanding capability, propelling it from "point cloud detection" to "semantic understanding," providing a solid foundation for subsequent behavior recognition, intelligent decision-making, and early warning control.

[0215] In this embodiment, the determination module further includes:

[0216] The acquisition unit is used to acquire the full width at half maximum (FWHM) of the main peak of the echo signal;

[0217] The third judgment unit is used to determine whether the full width at half maximum (FWHM) of the main peak exceeds a preset multiple of the width of the historical single target waveform.

[0218] The third execution unit is used to detect several small peaks of the full width at half maximum (FWHM) of the main peak if the condition is met, identify the integrated energy on the left and right sides of the small peaks, and dynamically collect the superimposed waveforms caused by the reflection amplitudes of each target based on the symmetry of the integrated energy. Specifically, the small peaks include shoulder peak sub-valleys, plateau-shaped waveform sub-valleys, and inter-peak sub-valleys.

[0219] In this embodiment, the system acquires the full width at half maximum (FWHM) of the echo signal and then determines whether the FWHM exceeds a preset multiple of the historical single-target waveform width to execute corresponding steps. For example, when the system determines that the FWHM of the echo signal does not exceed a preset multiple of the historical single-target waveform width, the system considers that the current echo signal may originate from a relatively concentrated single target, whose signal waveform is relatively stable and has not undergone excessive expansion or overlap. The system confirms that the current echo signal originates from a single target and performs single-target tracking and analysis, extracting relevant information about the target, such as position, velocity, and trajectory, to further predict the target's motion trend and possible behavioral changes. Simultaneously, since the echo signal indicates that the target is single and clear... The system selects and optimizes the target detection algorithm to improve its tracking accuracy. For example, it uses a Kalman filter or other tracking algorithms to accurately track the target, avoiding misjudgments caused by multi-target interference. The system can reduce the need for complex calculations, focusing on the tracking and analysis of a single target, thus improving efficiency. Furthermore, when the system confirms that the echo signal originates from a single target, it can appropriately increase its tolerance to signal noise and interference, ignoring the impact of background noise interference on the signal, thereby further improving the accuracy and stability of detection. In cases with less interference, the system can more clearly analyze the behavioral characteristics of the target; for example, when the system determines that the full width at half maximum (FWHM) of the echo signal exceeds a preset multiple of the historical single-target waveform width, the system will consider the current echo... The signal originates from multiple dispersed targets. The system detects several smaller peaks at half-width (FWHM) of the main peak. These smaller peaks include secondary valleys of shoulder peaks, plateau-shaped waveforms, and inter-peak valleys. The system identifies the integrated energy on both sides of these smaller peaks. Based on the symmetry of different integrated energies, it dynamically acquires the superimposed waveforms caused by the reflection amplitudes of each target. By finely identifying the smaller peaks in the main peak (such as secondary valleys of shoulder peaks and plateau-shaped waveforms) and analyzing the integrated energy on both sides, the system can effectively distinguish multiple reflection sources in the superimposed signal. Compared to directly identifying a wide main peak as a "fuzzy target," this method can more precisely distinguish multiple approaching targets, significantly improving the accuracy of multi-target analysis and avoiding the omission or misidentification of individual targets due to waveform aliasing. Furthermore, based on the various... By analyzing the symmetry of the integrated energy on both sides of the small peak, the system can dynamically infer the reflection amplitude of each target in the composite waveform. This approach helps to reconstruct the independent echo characteristics of each target, enabling the system to extract the true reflection intensity, possible material differences, or body shape distribution characteristics of each target. This improves the ability to judge the target category, shape, and behavior in subsequent identification. Furthermore, in complex scenarios where multiple targets are close together, overlap, or frequently occlude (such as crowds approaching, walking through corridors, or multiple people converging), the system can not only maintain the continuity of echo recognition by refining the processing after the full width at half maximum (FWHM) of the main peak exceeds the limit, but also dynamically adjust the number and distribution information of targets in the perception model, thereby improving the stability and robustness of the system for multi-target tracking and classification in dynamic environments.

[0220] In this embodiment, the second determination module further includes:

[0221] The second calculation unit is used to sort the target distribution point cloud according to a preset angle value, construct a corresponding angle sequence by sorting, and calculate the angle increment of the angle sequence.

[0222] The fourth judgment unit is used to determine whether the angle increment has increased;

[0223] The fourth execution unit is used to identify the break point after the angle increment increases if the break point is identified, perform clustering and cutting at the break point as the separation boundary of the point cloud target, divide the point cloud target into several continuous segments based on the position of the break point, and use the several continuous segments as the angle region of a single candidate target.

[0224] In this embodiment, the system sorts the target distribution point cloud according to preset angle values, constructs corresponding angle sequences, calculates the angle increments of these angle sequences, and then determines whether these angle increments have increased to execute corresponding steps. For example, when the system determines that the angle increments of the angle sequences have not increased, the system considers the target distribution to be relatively uniform or to have a large stable region. The system sorts the target distribution point cloud according to the angle values ​​to obtain a corresponding angle sequence. Assuming that at a certain moment, the target point cloud distribution is relatively uniform and the interval between each point in the angle sequence is small, the system calculates the angle increments of these angle sequences, i.e., the differences between adjacent angle values. If the target distribution is relatively uniform, the angle increments will be relatively consistent with little difference. The calculated angle increments reflect the target's spatial distribution. If these angle increments remain consistent or change little, it indicates a relatively uniform target distribution, and the system can accurately determine the target's distribution characteristics. For example, within a relatively uniform area, the target may exhibit continuous reflected signals with small changes in angle increments. Furthermore, if the angle increments remain within a certain range without increasing, it indicates a relatively stable target distribution. The system will not identify significant fluctuations or discrete point clouds in the target distribution; in other words, the relative positions of the targets will change little, without dense areas or scattered variations. In such cases, the system assumes the targets are in a relatively stable or uniformly distributed state, potentially eliminating the need for further complex processing such as target separation or noise removal. For instance, when the system detects a significant increase in the angle increment of the angle sequence, it assumes the target distribution is concentrated, exhibiting dense areas or scattered changes. The system identifies the breakpoints where the angle increment increases, performs clustering at these breakpoints as separation boundaries for the point cloud targets, and divides the point cloud targets into several continuous segments based on the locations of these breakpoints. These continuous segments are then used as the angle regions of individual candidate targets. By identifying breakpoints in the angle sequence where the angle increment significantly increases, the system can accurately locate the targets within the point cloud. The transition boundary effectively separates multiple targets that are close to or partially overlap each other, avoiding target confusion and improving the accuracy of subsequent classification and recognition. At the same time, it uses the abrupt change characteristics of angle increment as the basis for clustering and segmentation, so that the system can still flexibly divide the point cloud data into regions when there are multiple targets, especially when there are dense distributions and uneven spacing at angles. This enhances the adaptability of point cloud processing to complex distributions. Furthermore, by dividing the entire point cloud into several continuous segments as candidate target regions, the system can process each segment more independently and in a more targeted manner, such as dynamically setting filtering parameters and assigning tracking numbers, thereby reducing the computational burden on irrelevant regions and improving the overall operating efficiency and real-time performance of the system.

[0225] In this embodiment, the receiving module further includes:

[0226] The switching unit is used to adaptively switch the beam coverage strategy of the microwave radar based on the preset scene type of the detection scene. The scene type specifically includes corridor, room and doorway, and the beam coverage strategy specifically includes omnidirectional, directional and variable beam.

[0227] The fifth judgment unit is used to determine whether the beam coverage strategy can meet the preset target dynamic detection.

[0228] The fifth execution unit is used to dynamically activate a preset static shielding zone according to the preset detection parameters of the microwave radar if no, and to block the obstacle echoes pre-collected by the microwave radar through the static shielding zone. The detection parameters specifically include the horizontal scanning angle range, the vertical coverage range, and the effective distance.

[0229] In this embodiment, the system adaptively switches the beam coverage strategy of the microwave radar based on a pre-defined scene type, specifically including corridors, rooms, and doorways. The beam coverage strategy includes omnidirectional, directional, and variable beams. The system then determines whether these beam coverage strategies can meet the pre-defined target dynamic detection requirements and executes the corresponding steps. For example, when the system determines that the microwave radar's beam coverage strategy can meet the pre-defined target dynamic detection requirements, the system considers that the current radar beam configuration has effectively covered the area where the target may appear or move, and has sufficient detection sensitivity and spatial resolution to stably detect the target in the field. In the context of dynamic behavior perception, the system maintains its current omnidirectional, directional, or variable beam coverage to ensure continuous and stable detection performance. Simultaneously, if both detection requirements and energy conservation are needed, it can switch to low-frequency or intermittent scanning modes to reduce energy consumption. Furthermore, provided the beam coverage requirements are met, the system activates subsequent fine-grained tracking functions, such as human posture estimation and micro-motion recognition (e.g., breathing, gait), to further uncover dynamic behavioral characteristics. For example, if the system determines that the microwave radar's beam coverage strategy cannot meet the pre-set target dynamic detection requirements, it will conclude that the current radar beam configuration cannot effectively cover the area where the target may appear. The system dynamically activates pre-set static shielding zones based on pre-defined detection parameters of the microwave radar, including the horizontal scanning angle range, vertical coverage range, and effective distance. These static shielding zones block obstacle echoes pre-collected by the microwave radar. By activating the static shielding zones, the system blocks echo signals from non-target areas, avoiding interference from obstacles. By setting the horizontal scanning angle range, vertical coverage range, and effective distance, the system can accurately locate the radar detection area, reducing unnecessary interference areas and making the radar's detection range more focused. This allows for better coverage of the area required for dynamic target detection. Simultaneously, the introduction of static shielding zones effectively reduces the impact of unnecessary obstacle echo signals on the radar system. By blocking obstacle echoes collected by the radar, these echo signals are prevented from being confused with target echoes, ensuring the system can focus on target detection. This is especially important when target detection conditions change or the environment is complex. Dynamically activating static shielding zones means the radar can adaptively adjust to real-time environmental changes, thereby improving the system's stability and accuracy in complex environments. By shielding interference signals, the system can more accurately capture target signals, especially in situations with multiple targets or complex backgrounds, contributing to improved detection reliability.

[0230] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A real-time human detection method based on microwave radar, characterized in that, Includes the following steps: Based on the preset detection scenario of the microwave radar, the echo signal of the detection scenario is repeatedly output and received within a preset time period. Determine whether the echo signal detects a peak value of the merged signal; If so, based on the pre-collected original radar signal, the original radar signal is used as superimposed data of sparse source to construct a corresponding sparse reconstruction model. The aliased signal is decomposed into corresponding sparse units using a preset sparse coding. Multidimensional features are extracted from the sparse units. The target distribution point cloud of the multidimensional features is established through a preset three-dimensional feature space. The multidimensional features specifically include distance features, Doppler features, and angle features. Determine whether the angular direction of the target distribution point cloud forms a continuous region; If no target distribution point cloud is formed, the local point cloud density of the target distribution point cloud is identified. Based on the local point cloud density, the corresponding number of targets is dynamically estimated. A preset Kalman filter is applied to continuously track the target parameters of each target, generate the separated multi-target echo, and reconstruct the multi-target echo into a pseudo-color radar image. The target parameters specifically include target velocity and target velocity change.

2. The real-time human detection method based on microwave radar according to claim 1, characterized in that, The step of decomposing the aliased signal into corresponding sparse units using a preset sparse coding method and extracting multidimensional features from the sparse units further includes: Calculate the squared amplitude integral of the sparse unit, normalize the squared amplitude integral to a preset standard scale, apply a preset continuous wavelet transform to the sparse unit, and generate a corresponding time-frequency spectrum, wherein the time-frequency spectrum specifically includes the dominant frequency, bandwidth, and frequency change rate. Determine whether the time-frequency spectrum detects a preset low-frequency periodic motion feature, wherein the low-frequency periodic motion feature specifically includes breathing and heartbeat; If so, based on the echo delay of the sparse unit, the target distance corresponding to the echo delay is collected, and according to the target distance, the internal point distribution pattern of the sparse unit is statistically analyzed. Within a preset time window, the time trajectory evolution of the sparse unit is tracked. The internal point distribution pattern specifically includes the principal axis length, flatness, and density distribution characteristics.

3. The real-time human detection method based on microwave radar according to claim 1, characterized in that, After the step of establishing the target distribution point cloud of the multidimensional features through a preset three-dimensional feature space, the method further includes: Based on the point cloud density of the target distribution point cloud, identify the number of neighboring points of each point cloud within a preset radius; Determine whether the number of neighboring points is less than a preset value; If so, then detect extreme outliers of the multidimensional features, evaluate the dispersion of each point cloud based on the extreme outliers, statistically analyze the topological characteristics of the target distribution point cloud, and add additional features to each point cloud. Specifically, the extreme outliers include reflection energy supermassive and frequency jumps, the topological characteristics include the main direction of the point cloud and the overall shape of the point cloud, and the additional features include local curvature and local velocity gradient.

4. The real-time human detection method based on microwave radar according to claim 1, characterized in that, The step of identifying the local point cloud density of the target distribution point cloud and dynamically estimating the corresponding number of targets based on the local point cloud density further includes: Identify continuous frame point clouds from the target distribution point cloud, perform temporal fusion of the continuous frame point clouds, and generate corresponding high-density point cloud regions. Determine whether the high-density region of the point cloud recurs within a preset frequency; If so, the location information of the high-density region of the point cloud is obtained, and a preset micro-motion spectrum feature is extracted from the high-density region of the point cloud. Based on the location information and the micro-motion spectrum feature, the target type of the continuous frame point cloud is dynamically divided. The micro-motion spectrum feature specifically includes breathing frequency and arm swing frequency, and the target type specifically includes dynamic human body, interference object and non-human body object.

5. The real-time human detection method based on microwave radar according to claim 1, characterized in that, The step of determining whether the echo signal detects a peak value of the merged signal further includes: Obtain the full width at half maximum (FWHM) of the main peak of the echo signal; Determine whether the full width at half maximum (FWHM) of the main peak exceeds a preset multiple of the historical single-target waveform width; If so, then detect several small peaks of the full width at half maximum (FWHM) of the main peak, identify the integrated energy on the left and right sides of the small peaks, and dynamically collect the superimposed waveforms caused by the reflection amplitudes of each target based on the symmetry of the integrated energy. Specifically, the small peaks include shoulder peak sub-valleys, plateau-shaped waveform sub-valleys, and inter-peak sub-valleys.

6. The real-time human detection method based on microwave radar according to claim 1, characterized in that, The step of determining whether the angular direction of the target distribution point cloud forms a continuous region further includes: The target distribution point cloud is sorted according to a preset angle value, and the sorting constructs a corresponding angle sequence. The angle increment of the angle sequence is then calculated. Determine whether the angle increment has increased; If so, identify the break point after the angle increment increases, perform clustering and cutting at the break point as the separation boundary of the point cloud target, divide the point cloud target into several continuous segments based on the position of the break point, and use the several continuous segments as the angle region of a single candidate target.

7. The real-time human detection method based on microwave radar according to claim 1, characterized in that, The step of repeatedly outputting and receiving the echo signal of the detection scenario based on the microwave radar preset detection scenario within a preset time period further includes: Based on the preset scene type of the detection scene, the beam coverage strategy of the microwave radar is adaptively switched. The scene type specifically includes corridor, room and doorway, and the beam coverage strategy specifically includes omnidirectional, directional and variable beam. Determine whether the beam coverage strategy can meet the preset target dynamic detection; If not, then according to the preset detection parameters of the microwave radar, the preset static shielding area is dynamically activated to block the obstacle echoes pre-collected by the microwave radar through the static shielding area. The detection parameters specifically include the horizontal scanning angle range, the vertical coverage range, and the effective distance.

8. A real-time human body detection system based on microwave radar, characterized in that, include: The receiving module is used to repeatedly output and receive the echo signal of the detection scene within a preset time period based on the preset detection scene of the microwave radar. The judgment module is used to determine whether the echo signal detects a peak value of the merged signal; The execution module is used to, if so, construct a corresponding sparse reconstruction model based on the pre-acquired original radar signal as superimposed data of a sparse source, decompose the aliased signal into corresponding sparse units using a preset sparse coding, extract multi-dimensional features from the sparse units, and establish a target distribution point cloud of the multi-dimensional features through a preset three-dimensional feature space, wherein the multi-dimensional features specifically include distance features, Doppler features, and angle features; The second judgment module is used to determine whether the angular direction of the target distribution point cloud forms a continuous region; The second execution module is used to identify the local point cloud density of the target distribution point cloud if no target distribution point cloud is formed, dynamically estimate the corresponding number of targets based on the local point cloud density, apply a preset Kalman filter to continuously track the target parameters of each target, generate the separated multi-target echo, and reconstruct the multi-target echo into a pseudo-color radar image. The target parameters specifically include target velocity and target velocity change.

9. The real-time human body detection system based on microwave radar according to claim 8, characterized in that, The execution module further includes: The computing unit is used to calculate the amplitude squared integral of the sparse unit, normalize the amplitude squared integral to a preset standard scale, apply a preset continuous wavelet transform to the sparse unit, and generate a corresponding time-frequency spectrum, wherein the time-frequency spectrum specifically includes the dominant frequency, bandwidth and frequency change rate. The judgment unit is used to determine whether the time-frequency spectrum detects a preset low-frequency periodic motion feature, wherein the low-frequency periodic motion feature specifically includes breathing and heartbeat; An execution unit is configured to, if so, acquire the target distance corresponding to the echo delay based on the echo delay of the sparse unit, statistically analyze the internal point distribution pattern of the sparse unit according to the target distance, and track the time trajectory evolution of the sparse unit within a preset time window, wherein the internal point distribution pattern specifically includes the principal axis length, flatness, and density distribution characteristics.

10. The real-time human body detection system based on microwave radar according to claim 8, characterized in that, Also includes: The identification module is used to identify the number of neighboring points of each point cloud within a preset radius based on the point cloud density of the target distribution point cloud. The third judgment module is used to determine whether the number of neighbor points is less than a preset value; The third execution module is used to detect extreme outliers of the multidimensional features if the condition is met, evaluate the dispersion of each point cloud based on the extreme outliers, statistically analyze the topological characteristics of the target distribution point cloud, and add additional features to each point cloud. Specifically, the extreme outliers include reflection energy supermassive and frequency jumps, the topological characteristics include the main direction of the point cloud and the overall shape of the point cloud, and the additional features include local curvature and local velocity gradient.

Images