Millimeter wave radar treadmill personnel existence detection method, device and equipment and medium

Abstract

The invention relates to a target person detection method and device based on millimeter wave radar, equipment and a medium. According to the method, radar detection configuration is adjusted, a gridding coordinate system is established in a target treadmill monitoring area, millimeter wave radar is used for collecting feature data related to movement of a target person, and movement track changes of the target person are extracted to generate person position information; acquiring breathing and heartbeat signals of a target person, and performing signal filtering and feature extraction to obtain vital sign information; comprehensively evaluating the existence state of the target person according to a preset threshold value and a discrimination standard in combination with the person position information and the vital sign information; and when the evaluation result shows that the target person leaves, the vital signs are abnormal or a cheating behavior exists, judging that the target person is in an abnormal state, and generating a corresponding early warning instruction. By adopting the method, abnormal conditions can be found in time, early warning can be generated, and the monitoring accuracy and response speed are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of radar detection, and particularly relates to a treadmill personnel presence detection method, device, equipment and medium of a millimeter wave radar. BACKGROUND

[0002] With the surge in demand for intelligent fitness and home health management, millimeter wave radars are introduced into the treadmill scene. By utilizing the characteristics of 60GHz or 77GHz frequency band electromagnetic waves that can still penetrate clothing, dust and capture chest micro-motions under strong shielding and low illumination conditions, the millimeter wave radars can realize the synchronous perception of the presence of personnel in the running area and the breathing and heartbeat signals.

[0003] The prior art usually adopts a single millimeter wave radar fixedly installed at the front end or the side of the treadmill to make threshold value judgment on the echo energy in a preset monitoring area and with a fixed detection threshold. Then, the heart rate and the breathing frequency are extracted by using a sliding average or a simple band-pass filter, and are compared with an empirical threshold value. If the energy is lower than the threshold, it is determined that the personnel leave. If the heart rate interval exceeds the empirical range, an abnormal alarm is triggered. The entire process can be completed on an embedded MCU, and the modification cost is low and does not infringe privacy.

[0004] However, the prior art directly grafts the millimeter wave radar into the treadmill environment without making special optimization for the complex electromagnetic scene composed of the metal frame, high-speed movement of the running belt and ground multipath reflection, so that the fixed monitoring area and the fixed threshold are prone to misjudging the echo of the running belt as the human body, resulting in "false targets". When running, the body fluctuates greatly and the posture changes quickly, so it is difficult for the simple energy threshold to distinguish the real chest micro-motion from the mechanical vibration. The extraction of vital signs is seriously polluted by the running belt vibration and motor noise, and the accuracy of abnormal state judgment is low. In addition, the system lacks continuous grid tracking of the position of the personnel, which cannot distinguish between normal end of exercise and accidental fall, nor can it identify the cheating behavior of deliberately shielding the radar, and the user experience and safety guarantee are both insufficient. SUMMARY

[0005] Therefore, it is necessary to provide a millimeter wave radar treadmill personnel presence detection method, device, equipment and medium capable of accurately detecting the presence of personnel to solve the above technical problems. In a first aspect, the application provides a millimeter wave radar treadmill personnel presence detection method, comprising:

[0006] S1, based on the personnel position tracking grid, according to the original detection data related to the target personnel motion at each grid coordinate point, the motion trajectory change of the target personnel is extracted, and the personnel position information is generated; the personnel position tracking grid is a grid coordinate system established based on the spatial features of the target treadmill monitoring area covered by the adjusted radar detection configuration; the original detection data is the feature data related to the target personnel motion obtained by processing the echo signal received after the millimeter wave radar emits electromagnetic waves according to the adjusted radar detection configuration;

[0007] S2, obtaining the vital sign signal of the target personnel, and obtaining the vital sign information by signal filtering and feature extraction on the vital sign signal; the vital sign signal includes the breathing signal and the heartbeat signal;

[0008] S3, according to the personnel position information and the vital sign information, combining the corresponding preset threshold and discrimination standard, the existence state of the target personnel is comprehensively evaluated, and the evaluation result is obtained;

[0009] S4, when the evaluation result shows that the target personnel leaves, the vital sign is abnormal or there is a cheating behavior, it is determined as an abnormal state; the abnormal state is used to indicate the generation of a corresponding warning instruction.

[0010] In one of the embodiments, S2 includes:

[0011] S21, the original detection data and the region identification information from each radar node are fused to obtain fused detection data with space-time identification; the region identification information is used to mark the grid region to which the original detection data belongs, and the space-time identification includes the grid coordinate and the time stamp;

[0012] S22, the fused detection data is matched to the corresponding coordinate point of the personnel position tracking grid to generate an association mapping result;

[0013] S23, based on the association mapping result, the Doppler shift feature of the target personnel is converted into the velocity vector of the corresponding grid point; the Doppler shift feature is extracted from the target personnel at the corresponding grid coordinate point in the fused detection data;

[0014] S24, according to the time stamp sequence and the continuity of each grid coordinate point, the grid coordinate points of the same target personnel at different time, the associated fused detection data and the corresponding velocity vector are time-sequentially connected to generate preliminary motion trajectory data;

[0015] S25, based on the trajectory points corresponding to the continuous time stamps in the preliminary motion trajectory data, the spatial distance deviation of adjacent trajectory points is calculated through a preset distance threshold; the spatial distance deviation is used for comparison with the threshold to identify abnormal trajectory points;

[0016] S26, remove abnormal trajectory points from the preliminary motion trajectory data, and complete the trajectory according to the spatio-temporal continuity of the remaining trajectory points to obtain the corrected motion trajectory data;

[0017] S27, based on the corrected motion trajectory data, capture and analyze the spatio-temporal variation characteristics of the continuous trajectory points of the same target personnel to obtain the velocity vector direction change rate of adjacent trajectory points;

[0018] S28, compare the velocity vector direction change rate with the preset threshold, extract the corresponding trajectory points that exceed the preset threshold, and generate personnel location information containing the grid coordinates and time stamp thereof.

[0019] In one of the embodiments, S21 includes:

[0020] S211, perform global time reference conversion and grid coordinate alignment processing on the original detection data and region identification information from each radar node to obtain aligned detection data with unified spatio-temporal stamp;

[0021] S212, calculate the weight coefficient based on the region overlap degree of adjacent nodes in the aligned detection data; the calculation formula of the weight coefficient is:

[0022]

[0023] wherein, W(x, y) is the weight coefficient, O i (x,y) is the region overlap degree of node i at grid coordinates (x,y), R i is the reliability factor of node i, d c，i is the distance from the center of the detection region of node i to the grid coordinates (x,y), and σ is the scale parameter of the distance attenuation factor;

[0024] S213, screen the original detection data of each radar node, remove the repeatedly appearing detection data at the same grid coordinates, and obtain the screened detection data;

[0025] S214, based on the weight coefficient and the screened detection data, perform weighted fusion calculation on the multi-node screened detection data at the same grid coordinates to obtain the initial fusion result; the initial fusion result is calculated by the following formula:

[0026]

[0027] wherein, D f (x,y) is the initial fusion result at grid coordinates (x,y), N is the number of nodes participating in fusion, d i (x,y) is the original detection data of node i at grid coordinates (x,y);

[0028] S215, retaining the initial fusion result with a confidence higher than the preset confidence threshold to obtain fusion detection data.

[0029] In one of the embodiments, after S28, further comprising:

[0030] S281, performing angle information analysis of the reflected signal of the target person positioned by the personnel position information to obtain a spatial angle distribution result of the target person;

[0031] S282, based on the spatial angle distribution result, combining the radar cross-section differences of different parts of the human body, matching the spatial angle distribution of the target person with each part of the human body to obtain a matching relationship;

[0032] S283, based on the matching relationship, identifying the body orientation and motion posture of the target person through matching association to obtain posture information;

[0033] S284, classifying and identifying the posture information based on a preset posture feature library to obtain a posture classification result; the posture feature library is updated based on a large amount of historical data of normal and abnormal postures of people on the treadmill;

[0034] S285, based on the posture classification result, judging whether the target person is facing forward to the front of the treadmill for normal movement to obtain a movement state judgment result;

[0035] S286, analyzing the movement state judgment result, and when detecting that the person is not facing forward more than a set threshold, determining an abnormal presence state.

[0036] In one of the embodiments, S3 comprises:

[0037] S31, performing time sequence continuity verification on the personnel position information, and when detecting that the personnel position information suddenly disappears or the trajectory data is interrupted, displaying an evaluation result that the target person leaves;

[0038] S32, based on the vital sign information, combining a preset physiological signal mode library to classify and identify different physiological signal modes, and when detecting that the vital sign signal is abnormal or disappears, displaying an evaluation result that the vital sign is abnormal.

[0039] In one of the embodiments, S4 comprises:

[0040] S41, according to the personnel position information and the vital sign information, performing multi-source information fusion on multi-dimensional information of the target person to obtain fusion information; the multi-dimensional information includes radar scattering characteristics, motion characteristics, physiological signals, and environmental information;

[0041] S42, based on the fusion information, combining the real person existence discrimination standard established based on the human electromagnetic wave reflection characteristic parameter and the motion physiology characteristic, analyzing the authenticity of the detected target person, and obtaining a comprehensive score;

[0042] S43, when the comprehensive score is lower than a set threshold, determining that there is a cheating behavior.

[0043] In one of the embodiments, S1 is further preceded by:

[0044] S11, based on the adjusted radar detection configuration, dynamically switching the transmission frequency of the millimeter wave radar according to the interference frequency distribution characteristics in the real-time electromagnetic environment monitoring result through a frequency hopping sequence generation rule, to generate a frequency-adjusted signal; the adjusted radar detection configuration is obtained by adjusting the monitoring area range and detection parameters of the millimeter wave radar in the target treadmill environment;

[0045] S12, adjusting the pulse width and cycle duty cycle of the frequency-adjusted signal through time-frequency waveform parameter reconstruction to obtain an optimized waveform signal;

[0046] S13, quadrature code multiplexing coding the optimized waveform signal to generate a coded signal; the coded signal is used for the transmission waveform parameter in the adjusted radar detection configuration.

[0047] In a second aspect, the application also provides a millimeter wave radar treadmill personnel existence detection device, comprising:

[0048] A trajectory monitoring module is configured to extract the motion trajectory change of the target person based on a personnel position tracking grid and according to the original detection data related to the motion of the target person at each grid coordinate point, and generate personnel position information; the personnel position tracking grid is a grid coordinate system established based on the spatial characteristics of the target treadmill monitoring area covered by the adjusted radar detection configuration; the original detection data is feature data related to the motion of the target person obtained by processing the echo signal received after the millimeter wave radar transmits electromagnetic waves according to the adjusted radar detection configuration;

[0049] A vital sign signal monitoring module is configured to obtain the vital sign signal of the target person, and obtain vital sign information by signal filtering and feature extraction on the vital sign signal; the vital sign signal includes a breathing signal and a heartbeat signal;

[0050] An existence state evaluation module is configured to evaluate the existence state of the target person according to the personnel position information, the vital sign information and the authenticity judgment result, in combination with the corresponding preset threshold and discrimination standard, and obtain an evaluation result; the authenticity judgment result is obtained by matching degree analysis of the human electromagnetic wave reflection characteristic parameter and the motion physiology characteristic;

[0051] An evaluation result determination module is configured to determine an abnormal state when the evaluation result shows that the target person leaves, has abnormal vital signs, or has fraudulent behavior, and the abnormal state is used to generate a corresponding early warning instruction.

[0052] In a third aspect, the present application also provides a computer device, comprising a memory and a processor, the memory stores a computer program, and the processor implements the millimeter wave radar treadmill person existence detection method of the first aspect when executing the computer program.

[0053] In a fourth aspect, the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the millimeter wave radar treadmill person existence detection method of the first aspect.

[0054] The millimeter wave radar treadmill person existence detection method, device, computer device and storage medium described above extract the motion trajectory by constructing a person position tracking grid, synchronously collect vital sign signals such as respiration / heartbeat, and comprehensively evaluate the person state in combination with a preset threshold, and finally trigger an early warning for off-site, abnormal vital signs or fraudulent behavior, thereby realizing non-contact and high-precision person existence state monitoring and abnormal early warning, and significantly improving the intelligent level and response efficiency of security monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related art, the drawings needed to be used in the embodiment or related art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0056] Figure 1 A flowchart of a millimeter wave radar treadmill person existence detection method provided by the present application is shown in the figure.

[0057] Figure 2 A flowchart of a method for generating an encoded signal in an optional embodiment of the present application is shown in the figure.

[0058] Figure 3 A structural diagram of a millimeter wave radar treadmill person existence detection device provided by the present application is shown in the figure. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.

[0060] The millimeter wave radar treadmill personnel presence detection method provided by the embodiments of the present application can be applied in the following implementation environment. The control terminal communicates with the monitoring device deployed around the treadmill through a network. The data storage system can store the data required to be processed by the monitoring device. The data storage system can be integrated on the control terminal.

[0061] The device deployed around the treadmill maintains an adaptive spatial positional relationship with the treadmill, and the monitoring range thereof can cover the movement area of the treadmill and a reasonable range around the treadmill. The relevant information of the personnel moving on the treadmill can be collected without affecting the normal use of the treadmill, such as the position and movement trajectory of the personnel through the emission and reflection of specific signals, and the capture of vital sign signals such as respiration and heartbeat. The control terminal processes and analyzes the collected data to realize the detection of the presence state of the personnel on the treadmill.

[0062] In an exemplary embodiment, as shown in Figure 1 A millimeter wave radar treadmill personnel presence detection method is provided. The method is described by taking the control terminal as an example. The method comprises the following steps:

[0063] S1, based on the personnel position tracking grid, the movement trajectory change of the target personnel is extracted according to the original detection data related to the movement of the target personnel at each grid coordinate point, and the personnel position information is generated. The personnel position tracking grid is a grid coordinate system established based on the spatial features of the target treadmill monitoring area covered by the adjusted radar detection configuration. The original detection data is the feature data related to the movement of the target personnel obtained by processing the echo signal received by the millimeter wave radar after emitting electromagnetic waves according to the adjusted radar detection configuration.

[0064] Optionally, first, the personnel position tracking grid adaptive to the radar detection range is constructed. The treadmill and the surrounding monitoring area are divided into three-dimensional grid units according to a predetermined accuracy to form a spatial positioning reference. The millimeter wave radar emits electromagnetic waves according to the adjusted detection parameters (such as transmission frequency, scanning angle, and sampling rate). After receiving the echo signal reflected by the target personnel, the feature data related to the movement is extracted through pre-processing such as clutter suppression and pulse compression, including distance information calculated through echo time delay, azimuth angle information analyzed through array antenna phase difference, and radial movement speed reflected through Doppler shift. Based on the above data, combined with the grid coordinate system, the personnel position information containing the grid coordinates of the target personnel at different times is generated through multi-frame data association and trajectory fitting, and the dynamic tracking of the movement trajectory is realized.

[0065] S2, obtain the vital sign signal of the target personnel, and obtain the vital sign information by performing signal filtering and feature extraction on the vital sign signal. The vital sign signal includes a respiration signal and a heartbeat signal.

[0066] Exemplarily, the micro-Doppler effect of the millimeter wave radar is used to capture the vital sign signals of the target person. After the electromagnetic waves emitted by the radar are reflected by the fluctuation of the human chest (breathing) and the beating of the heart (heartbeat), weak frequency modulation signals are generated, which are mixed with motion interference signals. By Kalman filtering, the motion noise is separated, and then wavelet transform or Fourier analysis is used to extract features from the filtered signals to obtain breathing frequency (usually 0.1-0.5Hz), heartbeat frequency (usually 1-2Hz) and other vital sign information, which provides a quantitative basis for judging the physiological state of the person.

[0067] S3, according to the personnel position information and the vital sign information, combining the corresponding preset threshold and discrimination criterion, the existence state of the target person is comprehensively evaluated, and an evaluation result is obtained.

[0068] Specifically, state evaluation is achieved through multi-dimensional feature fusion. Based on the personnel position information, the trajectory continuity (such as whether there is a sudden interruption) and the spatial distribution (such as whether it is located in the treadmill area) are analyzed, and the preset position threshold such as the distance threshold for leaving the monitoring area is combined to determine whether the person leaves; based on the vital sign information, the extracted breathing and heartbeat frequencies are compared with the normal physiological range threshold to identify vital sign abnormalities; at the same time, through the matching analysis of the human electromagnetic wave reflection characteristics (such as the dynamic change mode of the radar cross section) and the motion physiology characteristics (such as the correlation between the motion posture and the vital signs), it is determined whether there is a cheating behavior, such as placing a fake person. Finally, the evaluation results including "person exists, person leaves, vital sign abnormality, and cheating behavior" are output.

[0069] S4, when the evaluation result shows that the target person leaves, the vital sign is abnormal, or there is a cheating behavior, it is determined as an abnormal state; the abnormal state is used to indicate the generation of a corresponding warning instruction.

[0070] Among them, for the evaluation result output by S3, if the abnormal state determination condition is met, the person leaves, the vital sign is abnormal, or there is a cheating behavior, the warning mechanism is triggered. According to the type of abnormality, a differentiated warning instruction is generated: such as triggering a device pause instruction when the person leaves, triggering a sound and light alarm and pushing emergency information when the vital sign is abnormal, and triggering a device locking instruction when a cheating behavior is detected, to ensure the safety and standardization of the use of the treadmill.

[0071] The millimeter wave radar personnel presence detection method can realize accurate monitoring of the movement track and vital signs of the treadmill personnel by combining multi-source radar signal processing technology; through multi-dimensional feature fusion and threshold discrimination, the personnel presence state can be comprehensively evaluated, and scenes such as normal use, personnel leaving, abnormal vital signs and cheating behaviors can be effectively distinguished. The method solves the problems of poor environmental adaptability, high false judgment rate and single function in the traditional detection technology, improves the accuracy, comprehensiveness and intelligent level of the treadmill personnel presence detection, and provides reliable technical support for protecting user safety and standardizing equipment use.

[0072] In one of the embodiments, S2 comprises:

[0073] S21, fusion processing of the original detection data and regional identification information from each radar node to obtain fusion detection data with space-time identification; the regional identification information is used to mark the grid area to which the original detection data belongs, and the space-time identification includes grid coordinates and a time stamp.

[0074] Optionally, for the distributed detection scene of multiple radar nodes, the original detection data including distance, azimuth angle, Doppler shift and other features output by each node and the corresponding regional identification information are collected, which is used to locate the grid partition to which the data belongs. Through a global time synchronization algorithm, the time references of different nodes are unified, and each piece of data is given a unique time stamp; at the same time, based on a preset grid coordinate system, the data of each node is mapped to a unified spatial coordinate, completing the space-time alignment. A weighted fusion algorithm is used to integrate the multi-source data of the same grid coordinate, considering factors such as node reliability and regional overlap, to finally generate fusion detection data containing both grid coordinates with spatial identification and time stamps with time identification, solving the space-time inconsistency problem of multi-node data.

[0075] S22, matching the fusion detection data to the corresponding coordinate points of the personnel position tracking grid to generate an association mapping result.

[0076] Specifically, the personnel position tracking grid is a three-dimensional grid coordinate system covering the treadmill monitoring area, and each grid point corresponds to a unique spatial coordinate. Through coordinate matching, the grid coordinates in the fusion detection data are accurately associated with the preset coordinate points of the tracking grid, and a mapping relationship between the data and the physical space position is established. In the association process, a neighbor search algorithm is used to optimize the matching efficiency, so that each piece of fusion detection data is accurately bound to the corresponding coordinate point in the tracking grid to form an association mapping result, providing a spatial reference for subsequent track construction.

[0077] S23, based on the association mapping result, converting the Doppler shift feature of the target personnel into a velocity vector of the corresponding grid point; the Doppler shift feature is extracted from the target personnel at the corresponding grid coordinate point in the fusion detection data.

[0078] Exemplarily, the Doppler shift feature is the difference between the radar echo signal frequency and the transmitted signal frequency, which is caused by the movement of the target person relative to the radar, and reflects the radial velocity of the target person along the radar line-of-sight direction. The Doppler shift feature of the grid point where the target person is located is extracted from the association mapping result, and combined with the spatial azimuth angle of the grid point, i.e., the angle relative to the radar node, to decompose the radial velocity into a velocity vector in three-dimensional space, including the velocity magnitude and direction. The conversion of the velocity vector realizes the conversion from one-dimensional radial velocity to three-dimensional space motion state, and more comprehensively describes the target motion feature.

[0079] S24, according to the time stamp order and the continuity of each grid coordinate point, the grid coordinate points of the same target person at different times, the associated fusion detection data and the corresponding velocity vectors are time-sequentially connected to generate preliminary motion trajectory data.

[0080] Optionally, the time stamp order is to sort the fusion detection data according to the time sequence of data acquisition; the continuity of the grid coordinate point is that the spatial distance of the grid points where the target is located at adjacent time points does not exceed a preset threshold, and the preset threshold is set based on the normal motion speed of the human body. The detection data of the same target person at different times is identified through Kalman filtering prediction, and the discrete grid coordinate points, fusion detection data and velocity vectors are connected in time axis according to the time stamp order and spatial continuity, to form preliminary trajectory data describing the motion process of the target, and realize the construction from single point detection to continuous trajectory.

[0081] S25, based on the trajectory points corresponding to the continuous time stamps in the preliminary motion trajectory data, the spatial distance deviation of adjacent trajectory points is calculated through a preset distance threshold value; the spatial distance deviation is used for comparison with the threshold value to identify abnormal trajectory points.

[0082] Specifically, the distance threshold value is the maximum reasonable distance of the trajectory points at adjacent time points set according to the physical limit of human motion, such as the maximum moving distance within 0.5 seconds in the running state. By calculating the spatial distance deviation between the trajectory points corresponding to two continuous time stamps in the preliminary trajectory, the deviation is compared with the distance threshold value: when the deviation exceeds the threshold value, the latter trajectory point is determined as an abnormal trajectory point, which may be caused by noise interference or false detection, providing a judgment basis for trajectory correction.

[0083] S26, the abnormal trajectory points are removed from the preliminary motion trajectory data, and the trajectory is completed according to the space-time continuity of the remaining trajectory points to obtain the corrected motion trajectory data.

[0084] Wherein, for the abnormal trajectory points identified in S25, they are removed from the preliminary trajectory through data cleaning; for the trajectory breaks after the abnormal points are removed, linear interpolation based on velocity vector is adopted to combine the time stamp interval and spatial distribution characteristics of the remaining trajectory points for trajectory completion, so that the modified trajectory data maintains continuity in time and space dimensions and more truly reflects the actual movement process of the target person.

[0085] S27, based on the modified motion trajectory data, the spatio-temporal variation characteristics of the consecutive trajectory points of the same target person are captured and analyzed to obtain the velocity vector direction change rate of adjacent trajectory points.

[0086] Exemplarily, the velocity vector direction change rate is the ratio of the included angle between the velocity vectors of two consecutive trajectory points to the time interval, reflecting the speed of change of the target movement direction. The included angle between the velocity vectors of adjacent trajectory points is calculated by the cosine theorem, and the direction change rate per unit time is obtained by combining the time interval of the two points, so as to quantify the degree of sudden change of the target movement direction.

[0087] S28, the velocity vector direction change rate is compared with a preset threshold, and the corresponding trajectory points exceeding the preset threshold are extracted to generate personnel location information containing their grid coordinates and time stamps.

[0088] Specifically, the preset threshold is set according to the direction change range of human body normal movement in the treadmill use scene, such as the direction change rate threshold of sudden turning and sudden stopping. By comparing the direction change rate calculated in S27 with the preset threshold, when the change rate exceeds the threshold, the corresponding trajectory point is marked as a key position point, reflecting the significant change of the movement state. Finally, the marked key position points and their corresponding grid coordinates and time stamps are extracted to form the location information containing the key state of the target person movement, providing core data for subsequent state evaluation.

[0089] In the above embodiment, through the conversion from Doppler shift characteristics to velocity vector, three-dimensional quantification of the target movement state is realized; with the trajectory concatenation, abnormal point elimination and completion technology, the continuity and accuracy of the motion trajectory are improved; through the analysis of the velocity vector direction change rate, the key state of the target movement is accurately captured.

[0090] In one exemplary embodiment, S21 includes:

[0091] S211, the original detection data and region identification information from each radar node are subjected to global time reference conversion and grid coordinate alignment processing to obtain aligned detection data with unified space-time stamp.

[0092] Exemplarily, the original detection data is collected by multiple millimeter wave radar nodes, each node may have a time deviation due to clock drift, and the area identification information marks the grid partition to which the data belongs. Through a network time protocol or a hardware synchronization trigger mechanism, the timestamps of all nodes are unified to a global time reference, eliminating the time difference; at the same time, based on a preset personnel position tracking grid coordinate system, the original spatial coordinates (such as distance, azimuth) output by each node are converted into grid unified coordinates, realizing the spatial alignment of data from different nodes.

[0093] S212, a weight coefficient is calculated based on the area overlap degree of adjacent nodes in the aligned detection data; the calculation formula of the weight coefficient is:

[0094]

[0095] wherein W(x, y) is the weight coefficient, O i (x, y) is the area overlap degree of node i at grid coordinate (x, y), R i is the reliability factor of node i, d c，i is the distance from the center of the detection area of node i to the grid coordinate (x, y), and σ is the scale parameter of the distance attenuation factor.

[0096] Specifically, the area overlap degree refers to the proportion of the overlapping area of the detection range of the node and the area where the grid coordinate is located, reflecting the coverage effectiveness of the node on the grid point; the reliability factor is a coefficient (range 0-1) preset based on the historical detection accuracy of the node, the higher the accuracy, the greater the reliability factor; d c，i is the straight-line distance from the center of the detection area of node i to the grid coordinate, and σ is the distance attenuation factor, controlling the influence degree of distance on the weight. By fusing the area overlap degree, the node reliability and the spatial distance factor, the weight coefficient W(x, y) of node i on the grid (x, y) is calculated, so that the nodes with more effective coverage, higher reliability and closer distance occupy higher weights in the fusion.

[0097] S213, the original detection data of each radar node is screened to eliminate repeated detection data at the same grid coordinate, obtaining the screened detection data.

[0098] Exemplarily, the same grid coordinate may be detected by multiple radar nodes at the same time, resulting in repeated data, such as when adjacent nodes cover the overlapping area. Through hash mapping, repeated data at the same grid coordinate is identified, and detection data with the highest signal strength or the largest signal-to-noise ratio is retained, and redundant items are eliminated. De-duplication screening reduces the amount of data and avoids interference of repeated information on the fusion result.

[0099] S214, based on the weight coefficient and the screened detection data, the screened detection data of multiple nodes at the same grid coordinate are weighted and fused to obtain an initial fusion result; the initial fusion result is calculated by the following formula:

[0100]

[0101] wherein, D f (x,y) is the initial fusion result at the grid coordinate (x,y), N is the number of nodes participating in fusion, d i (x,y) is the original detection data of node i at the grid coordinate (x,y).

[0102] Specifically, for the detection data of multiple radar nodes at the same grid coordinate after deduplication screening, such as the measurement values of distance, speed and other parameters of the same position by different nodes, a weighted fusion algorithm is used for data integration. First, the weight coefficient W(x,y) calculated in S212 is given to the detection data of each node, and the size of the weight value directly reflects the contribution of the node data in the fusion-the higher the reliability, the closer the distance to the grid point and the higher the regional coverage overlap, the higher the weight of the data. The sum of the products of the weights and detection data of all nodes is calculated by the numerator, that is, the accumulation of the data of each node in proportion to the weight; the sum of the weights of all nodes is calculated in the denominator, which is used to normalize the numerator result to eliminate the influence of the absolute value difference of the weight. The initial fusion result obtained finally is the optimal estimation value after integrating the information of multiple nodes, which not only retains the dominant role of high-quality data, but also offsets the random error of a single node through the complement of multiple data sources, so that the fusion result is closer to the true value.

[0103] For example, if the weight of node A to the grid (x,y) is 0.6 (high detection accuracy), and the detection value is 2.3 m; the weight of node B is 0.4 (lower accuracy), and the detection value is 2.5 m;

[0104] The fusion result is (0.6x2.3+0.4x2.5) / (0.6+0.4)=2.38 m, which is more biased towards the measurement value of the high-weight node, and also absorbs the effective information of other nodes.

[0105] S215, retaining the initial fusion result with a confidence higher than the preset confidence threshold to obtain the fused detection data.

[0106] The confidence degree is used to quantify the reliability of the initial fusion result, and is calculated by the consistency of the fusion data (such as the variance of the multi-node detection value) or the matching degree with historical data. The preset confidence threshold is set according to the application scenario (such as 0.7), and when the confidence degree of the initial fusion result is higher than the threshold, it is determined as valid data and retained; otherwise, it is considered as low-reliability data and eliminated. Through the confidence degree screening, the final fusion detection data with high precision is obtained.

[0107] In the above embodiments, the weighted fusion based on the weight coefficient fully gives play to the complementary advantages of multi-source data, and the de-duplication screening and confidence filtering effectively eliminate redundant and low-quality data, so that the finally generated fusion detection data has higher precision and reliability.

[0108] In one of the embodiments, after S28, it further includes:

[0109] S281, angle information analysis of the reflected signal of the target person positioned by the personnel position information, to obtain a spatial angle distribution result of the target person.

[0110] Specifically, the personnel position information has determined the coordinates of the target person in the grid, based on which, the angle information of the radar reflected signal at the position is extracted, including the azimuth angle and the elevation angle. Through the beamforming technology, the angle of the reflected signal is estimated to determine the angle distribution of the electromagnetic wave reflected back to the radar from different parts of the target person, to form a spatial angle distribution result. The result reflects the angle position of each part of the target person's body in space.

[0111] S282, based on the spatial angle distribution result, combining the radar cross section area difference of different parts of the human body, the spatial angle distribution of the target person is matched with each part of the human body to obtain a matching relationship.

[0112] Exemplarily, the radar cross section area (RCS) is a physical quantity describing the reflection ability of a target to radar electromagnetic waves, and different parts of the human body (such as the torso, limbs, and head) have significant differences in RCS due to differences in shape and material, for example, the torso RCS is usually greater than the limbs). The reflection intensity (positively correlated with RCS) corresponding to different angles in the spatial angle distribution result is compared with the known RCS feature library of each part of the human body, and the human body part corresponding to the reflection signal in each angle interval is determined through a pattern matching algorithm (such as template matching), to establish a matching relationship between the spatial angle distribution and the human body part, and realize the mapping from angle information to body part.

[0113] S283, based on the matching relationship, the body orientation and motion posture of the target person are recognized through matching association to obtain posture information.

[0114] Wherein, on the basis of matching relationship, the spatial distribution of human body parts at different time is analyzed through time sequence analysis: the body orientation can be determined by the main direction of the angle distribution of the torso and head, such as the head angle pointing to the orientation; the movement posture is identified by the angle change trend of the limbs (such as the arm swing amplitude and the leg flexion angle) (such as the periodic change of the leg angle when running). By integrating the spatial position and dynamic change of each part of the body, the posture information including the body orientation and the limb movement state is generated, and the real-time posture of the target person is quantitatively described.

[0115] S284, classifying and identifying the posture information based on a preset posture feature library to obtain a posture classification result; the posture feature library is updated based on a large number of historical data of normal and abnormal postures of people on a treadmill.

[0116] Optionally, the posture feature library includes feature templates of typical normal postures (such as facing forward to the front of the treadmill, uniform running posture) and abnormal postures (such as facing backward to the treadmill, lateral inclination, and pre-falling) in the running field. The above templates are constructed by clustering and feature extraction on a large number of historical posture data, and are continuously updated and optimized with new data. By using support vector machine or deep learning classification algorithm, the current posture information is compared with the templates in the feature library, and the posture classification result is output, such as normal running, standing backward, lateral inclination, etc., to realize the standardized classification of the posture.

[0117] S285, judging whether the target person is facing forward to the front of the treadmill for normal movement based on the posture classification result to obtain a movement state judgment result.

[0118] Specifically, in combination with the body orientation and movement features in the posture classification result, the judgment standard is set: if the body orientation is facing forward to the front of the treadmill, and the movement posture conforms to the normal running / walking features (such as the limb movement amplitude being within the normal range), it is determined as a normal movement state; if the body orientation is backward or lateral, or the movement posture deviates from the normal range, such as excessive inclination, it is determined as a non-frontal movement state. The movement state judgment result is obtained by this judgment, which clearly indicates whether the current movement of the person conforms to the safety specification.

[0119] S286, analyzing the movement state judgment result, and determining an abnormal presence state when it is detected that the person is not facing forward for more than a set threshold.

[0120] Exemplarily, the threshold is set as the duration of non-frontal orientation, such as 10 seconds. When the movement state judgment result is a non-frontal movement state and the duration exceeds the threshold, the system determines an abnormal presence state. This determination mechanism can identify postures that may lead to danger in time, such as backward running that is prone to imbalance and lateral inclination that is prone to falling, providing a triggering condition for subsequent warning.

[0121] In the above embodiments, the target personnel posture is accurately identified by matching the reflection signal angle analysis with the human body part, and the normal and abnormal motion states can be effectively distinguished by combining the dynamically updated posture feature library and threshold judgment mechanism, especially for high sensitivity to dangerous postures such as non-front facing, which provides more comprehensive technical support for ensuring the safety of treadmill use.

[0122] In one of the embodiments, S3 comprises:

[0123] S31, time sequence continuity verification is performed on the personnel position information, and when it is detected that the personnel position information suddenly disappears or the trajectory data is interrupted, an evaluation result of the target personnel leaving is displayed.

[0124] Specifically, by setting a time window (such as 5 seconds) and a position continuity threshold (based on the maximum distance of normal human movement speed), it is judged whether the time interval and spatial distance of adjacent position points in the window conform to the normal motion rule. If in continuous multiple windows, the position information suddenly disappears (no new coordinate data) or the distance between trajectory points exceeds the threshold and there is no reasonable completion possibility, such as the coordinates from the monitoring area directly changing to null, it is determined that the position information suddenly disappears or the trajectory data is interrupted, triggering the evaluation result of the target personnel leaving. This process adjusts to different motion intensities through dynamic threshold adjustment, such as the continuity standards difference between fast walking and slow running, reducing false positives.

[0125] S32, based on the vital sign information, different physiological signal patterns are classified and identified in combination with a preset physiological signal pattern library, and when it is detected that the vital sign signal is abnormal or disappears, an evaluation result of vital sign abnormality is displayed.

[0126] Specifically, the vital sign information includes parameters such as respiratory rate, heart rate, and signal strength, and the preset physiological signal pattern library contains the normal physiological signal range of healthy adults in motion state (such as respiratory rate 12-20 times / minute, heart rate 60-180 times / minute) and typical abnormal patterns (such as rapid breathing, sudden cardiac arrest, and sudden signal amplitude drop). The pattern library is constructed and dynamically optimized through clinical data and exercise physiology research. Adaptive threshold method and pattern matching algorithm are used to compare real-time vital sign information with the pattern library: if the parameter exceeds the normal range (such as heart rate > 100 times / minute at rest) or matches the abnormal pattern (such as periodic disappearance of heart rate signal), it is determined that the vital sign signal is abnormal; if the signal strength is below the detection threshold and lasts for a preset time (such as 3 seconds), it is determined that the signal disappears. Both cases trigger the evaluation result of "vital sign abnormality", and the abnormality degree classification (such as mild and severe) can be quantified by the degree of deviation from the normal range.

[0127] In the above embodiments, the precise judgment of the personnel leaving state is realized through the time sequence continuity verification, and the rapid identification of the vital sign abnormality is completed in combination with the physiological signal mode library, and the two together constitute a multi-dimensional existence state evaluation mechanism, which effectively reduces the safety risks caused by accidental leaving of personnel or physiological sudden conditions.

[0128] In one of the embodiments, S4 comprises:

[0129] S41, multi-source information fusion is performed on multi-dimensional information of the target personnel according to the personnel position information and the vital sign information to obtain fusion information; the multi-dimensional information includes radar scattering characteristics, motion characteristics, physiological signals and environmental information.

[0130] Exemplarily, the personnel position information provides the motion trajectory and the spatial distribution, and the vital sign information includes physiological parameters such as respiration and heartbeat, on the basis of which multi-dimensional features are further extracted: the radar scattering characteristics refer to the change law of the reflection intensity and polarization mode of the target personnel against electromagnetic waves with angle and frequency, such as the dynamic change mode of the radar cross section of different parts of the human body; the motion characteristics include speed vector, attitude change rate and other parameters reflecting the motion state; the physiological signals are the respiration rate and heartbeat waveform in the vital sign information; and the environmental information covers the background noise and electromagnetic interference intensity of the monitoring area. Through the attention mechanism network, the multi-source information is integrated to eliminate information redundancy and conflicts, and the fusion information containing key features of each dimension is generated to realize the conversion from scattered data to a unified feature vector.

[0131] S42, based on the fusion information, the authenticity of the detected target personnel is analyzed by combining the real personnel existence judgment standard established based on the human electromagnetic wave reflection characteristic parameters and the motion physiology characteristics to obtain a comprehensive score.

[0132] Optionally, the human electromagnetic wave reflection characteristic parameters include typical radar cross section range and micro-Doppler feature of reflected signals, such as frequency modulation mode caused by respiration; the motion physiology characteristics refer to the correlation between human motion and physiological signals, such as the rule that the heartbeat frequency increases with the increase of speed when running. The judgment standard established based on the above characteristics includes multiple sub-criteria, such as whether the reflection characteristics conform to the human RCS range and whether the motion and physiological signals have reasonable correlation. Through the analytic hierarchy process or the fuzzy comprehensive evaluation model, the matching degree of the fusion information and each sub-criterion is quantitatively scored, and then the comprehensive score (range 0-100) is obtained by weighting according to the weight. The higher the score is, the higher the degree of the target conforming to the real personnel characteristics is.

[0133] S43, when the comprehensive score is lower than a set threshold, it is determined that there is a cheating behavior.

[0134] The threshold is determined based on a large amount of test data of real people and fraud scenes (such as placing a fake person, using a motion simulation device) (such as 60 points), representing the critical value of distinguishing real and fraud. If the comprehensive score is lower than the threshold, it indicates that the multi-dimensional features of the target are significantly different from the features of real people (such as no physiological signal, large difference in reflection characteristics and human body), and the system determines that there is fraud. The result can be directly used to trigger subsequent warning or device control instructions.

[0135] In the above embodiment, the multi-dimensional features of the target are integrated through multi-source information fusion, and the discrimination standard is constructed in combination with the inherent electromagnetic wave reflection and motion physiology characteristics of the human body, thereby realizing accurate identification of fraud behavior and effectively preventing the situation of evading monitoring by forging the target. It provides more stringent technical support for the standard use and safety management of treadmills and other devices.

[0136] In an optional embodiment, as shown in Figure 2 Before S1, it further includes:

[0137] S11, based on the adjusted radar detection configuration, according to the interference frequency distribution characteristics in the real-time electromagnetic environment monitoring result, the transmission frequency of the millimeter wave radar is dynamically switched through the frequency hopping sequence generation rule, and the signal after frequency adjustment is generated. The adjusted radar detection configuration is obtained by adjusting the monitoring area range and detection parameters of the millimeter wave radar in the target treadmill environment.

[0138] Specifically, the adjusted radar detection configuration needs to be adapted to the treadmill scene first: according to the size of the treadmill, the placement position and the surrounding environment, a specific monitoring area range is delimited, such as covering the running belt and the surrounding 1 meter range, and the detection parameters are optimized, such as the scanning angle and the sampling rate. On this basis, the electromagnetic environment data is collected in real time through the spectrum monitoring module, and the interference frequency distribution characteristics are analyzed, such as the working frequency of other electronic devices in the gym and the electromagnetic noise concentration frequency band. The frequency hopping sequence generation rule is based on the pseudo-random sequence algorithm, combined with the interference frequency avoidance strategy, avoids the strong interference frequency band, dynamically switches the transmission frequency of the millimeter wave radar, such as sequentially hopping in the preset frequency band set, so that the transmission frequency is always in the low interference interval, generates the signal after frequency adjustment, and preliminarily reduces the influence of electromagnetic interference.

[0139] S12, the pulse width and cycle duty ratio of the signal after frequency adjustment are adjusted through time-frequency waveform parameter reconstruction, and an optimized waveform signal is obtained.

[0140] Optionally, the time-frequency waveform parameter reconstruction is a technology for adjusting the time-domain parameters of the waveform based on signal detection performance requirements. The pulse width affects the range resolution, and the smaller the width, the higher the resolution. The duty cycle (ratio of pulse duration to period) affects the average transmit power and anti-interference capability, and a reasonable reduction in the duty cycle can reduce the probability of being interfered. For the speed range of personnel motion in the running field (such as 0-10 km / h), through adaptive algorithm adjustment: when detecting fast motion at close range, narrow pulses (such as 100 ns) are used to improve resolution; when environmental interference is enhanced, the duty cycle is reduced (such as from 20% to 10%) to reduce the superposition opportunity of interference signals. Through the above adjustment, the signal after frequency adjustment reaches a balance between range resolution and anti-interference performance, generating an optimized waveform signal.

[0141] S13, orthogonal code division multiplexing encoding is performed on the optimized waveform signal to generate an encoded signal; the encoded signal is used as the transmit waveform parameter in the adjusted radar detection configuration.

[0142] Specifically, the orthogonal code division multiplexing encoding is a technology for multiplying the optimized waveform signal by an orthogonal spread code (such as Walsh code, Gold code) to make the signal orthogonal in the code domain. Signals corresponding to different orthogonal codes can be separated by correlation demodulation at the receiving end, and will not interfere with each other even if transmitted at the same frequency and at the same time. According to the number of radar channels and anti-interference requirements, a suitable orthogonal code group is selected to encode the optimized waveform signal, so that the transmitted signal has code division multiplexing characteristics. The encoded signal is used as the transmit waveform parameter in the adjusted radar detection configuration, which can improve the multi-target resolution capability, such as distinguishing between personnel on the running machine and surrounding interference targets, and further enhance the ability to resist narrowband interference and multiple access interference.

[0143] In the above embodiments, the dynamic frequency hopping avoids interference frequencies, the time-frequency parameter reconstruction optimizes the waveform performance, and the orthogonal code division multiplexing enhances the signal anti-interference and resolution capability, forming a multi-level radar signal anti-interference scheme. Through dynamic adjustment suitable for the running field, the radar signal significantly improves the anti-electromagnetic interference capability while maintaining high resolution and detection sensitivity.

[0144] In the above method for detecting the presence of personnel on a running machine using a millimeter wave radar, the radar detection configuration is dynamically adjusted to optimize the transmitted signal, and the precise position information of the personnel is generated by fusing the data of multiple radar nodes. Combined with vital sign monitoring and posture recognition, the state of the personnel is comprehensively evaluated, and a reality discrimination mechanism is introduced to prevent fraudulent behavior, forming a complete detection process from signal optimization to state evaluation. The accuracy, anti-interference capability and safety of personnel presence detection in the running field are significantly improved, effectively solving the problems of poor environmental adaptability, high false positive rate and vulnerability to fraud in traditional detection methods.

[0145] It should be understood that although the steps in the flowcharts involved in the embodiments described above are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately executed with at least part of other steps or steps or stages in other steps.

[0146] Based on the same inventive concept, the embodiments of the present application also provide a device for implementing the above-mentioned millimeter wave radar treadmill personnel presence detection method. The problem-solving implementation scheme provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more millimeter wave radar treadmill personnel presence detection device embodiments provided below can refer to the limitations of the millimeter wave radar treadmill personnel presence detection method described above, and will not be repeated here.

[0147] In one exemplary embodiment, as shown in Figure 3 a millimeter wave radar treadmill personnel presence detection device 110 is provided, comprising:

[0148] a trajectory monitoring module 111, configured to extract the motion trajectory change of the target personnel based on a personnel position tracking grid and the original detection data related to the motion of the target personnel at each grid coordinate point, and generate personnel position information; the personnel position tracking grid is a grid coordinate system established based on the spatial features of the target treadmill monitoring area covered by the adjusted radar detection configuration; the original detection data is feature data related to the motion of the target personnel obtained by processing the echo signal received after the millimeter wave radar transmits electromagnetic waves according to the adjusted radar detection configuration;

[0149] a vital sign signal monitoring module 112, configured to obtain the vital sign signal of the target personnel, and obtain vital sign information by performing signal filtering and feature extraction on the vital sign signal; the vital sign signal includes a breathing signal and a heartbeat signal;

[0150] a presence state evaluation module 113, configured to evaluate the presence state of the target personnel according to the personnel position information, the vital sign information and the authenticity judgment result, in combination with the corresponding preset threshold and discrimination standard, and obtain an evaluation result; the authenticity judgment result is obtained by analyzing the matching degree of the human electromagnetic wave reflection characteristic parameter and the motion physiology characteristic;

[0151] The evaluation result determination module 114 is configured to determine an abnormal state when the evaluation result shows that the target person leaves, has an abnormal vital sign, or has a cheating behavior, and the abnormal state is used to instruct generation of a corresponding early warning instruction.

[0152] In one of the embodiments, the vital sign signal monitoring module comprises:

[0153] The data fusion unit is configured to fuse raw detection data and region identification information from each radar node to obtain fused detection data with space-time identification, the region identification information is used to mark a grid region to which the raw detection data belongs, and the space-time identification comprises a grid coordinate and a time stamp;

[0154] The data matching unit is configured to match the fused detection data to a corresponding coordinate point of a person position tracking grid to generate a correlation mapping result.

[0155] The velocity vector conversion unit is configured to convert Doppler shift characteristics of the target person into velocity vectors of corresponding grid points based on the correlation mapping result, the Doppler shift characteristics being extracted from the target person at the corresponding grid coordinate point in the fused detection data.

[0156] The trajectory generation unit is configured to sequentially connect, in time sequence, grid coordinate points of the same target person at different time points, associated fused detection data, and corresponding velocity vectors to generate preliminary motion trajectory data according to time stamp order and continuity of each grid coordinate point.

[0157] The distance deviation calculation unit is configured to calculate spatial distance deviations of adjacent trajectory points by a preset distance threshold value based on trajectory points corresponding to continuous time stamps in the preliminary motion trajectory data, and the spatial distance deviations are used for comparison with the threshold value to identify abnormal trajectory points.

[0158] The trajectory correction unit is configured to remove abnormal trajectory points from the preliminary motion trajectory data and complete the trajectory according to space-time continuity of remaining trajectory points to obtain corrected motion trajectory data.

[0159] The direction change analysis unit is configured to obtain a velocity vector direction change rate of adjacent trajectory points by capturing and analyzing space-time change characteristics of continuous trajectory points of the same target person based on the corrected motion trajectory data.

[0160] The abnormal trajectory extraction unit is configured to compare the velocity vector direction change rate with a preset threshold value, extract corresponding trajectory points that exceed the preset threshold value, and generate person position information comprising grid coordinates and time stamps of the trajectory points.

[0161] In one of the embodiments, S21 comprises:

[0162] A space-time alignment subunit is configured to perform global time reference conversion and grid coordinate alignment processing on raw detection data and region identification information from each radar node to obtain aligned detection data with unified space-time stamps;

[0163] A weight calculation subunit is configured to calculate a weight coefficient based on region overlap of adjacent nodes in the aligned detection data; the calculation formula of the weight coefficient is as follows:

[0164]

[0165] wherein, W(x, y) is the weight coefficient, O i (x, y) is the region overlap of node i at grid coordinate (x, y), R i is the reliability factor of node i, d c，i is the distance from the detection region center of node i to grid coordinate (x, y), and σ is the scale parameter of the distance attenuation factor;

[0166] A data screening subunit is configured to screen the raw detection data of each radar node, eliminate repeated detection data at the same grid coordinate, and obtain screened detection data.

[0167] A weighted fusion subunit is configured to perform weighted fusion calculation on the screened detection data of multiple nodes at the same grid coordinate based on the weight coefficient and the screened detection data to obtain an initial fusion result; the initial fusion result is calculated by the following formula:

[0168]

[0169] wherein, D f (x, y) is the initial fusion result at grid coordinate (x, y), N is the number of nodes participating in fusion, d i (x, y) is the raw detection data of node i at grid coordinate (x, y);

[0170] A confidence filtering subunit is configured to retain the initial fusion result with a confidence higher than a preset confidence threshold to obtain fusion detection data.

[0171] In one embodiment, the abnormal trajectory extraction unit further comprises:

[0172] An angle analysis subunit is configured to analyze angle information of the reflected signal of the target person positioned by the personnel position information to obtain a spatial angle distribution result of the target person.

[0173] A part matching subunit is configured to match the spatial angle distribution of the target person with each part of the human body based on the spatial angle distribution result and the difference in radar cross-section area of different parts of the human body to obtain a matching relationship.

[0174] The posture recognition subunit is configured to recognize the body orientation and motion posture of the target person based on the matching relationship and obtain posture information.

[0175] The posture classification subunit is configured to classify and identify the posture information based on a preset posture feature library and obtain a posture classification result. The posture feature library is updated based on historical data of normal and abnormal postures of a large number of people on a treadmill.

[0176] The motion state judgment subunit is configured to judge whether the target person is facing forward and performing normal motion based on the posture classification result and obtain a motion state judgment result.

[0177] The abnormality detection subunit is configured to analyze the motion state judgment result, and when it is detected that the person is not facing forward by more than a set threshold, it is determined that there is an abnormal presence state.

[0178] In one embodiment, the presence state evaluation module includes:

[0179] The trajectory interruption detection unit is configured to perform time sequence continuity verification on the position information of the person, and when it is detected that the position information of the person suddenly disappears or the trajectory data is interrupted, an evaluation result that the target person has left is displayed.

[0180] The vital sign analysis unit is configured to classify and identify different physiological signal patterns based on vital sign information and a preset physiological signal pattern library, and when it is detected that the vital sign signal is abnormal or missing, an evaluation result that the vital sign is abnormal is displayed.

[0181] In one embodiment, the evaluation result determination module includes:

[0182] The multi-source information fusion unit is configured to perform multi-source information fusion on multi-dimensional information of the target person based on the position information and the vital sign information, and obtain fusion information. The multi-dimensional information includes radar scattering characteristics, motion characteristics, physiological signals, and environmental information.

[0183] The authenticity analysis unit is configured to analyze the authenticity of the target person based on the fusion information, a real person existence discrimination standard established based on human electromagnetic wave reflection characteristic parameters and motion physiology characteristics, and obtain a comprehensive score.

[0184] The fraud behavior determination unit is configured to determine that there is a fraud behavior when the comprehensive score is lower than a set threshold.

[0185] In one embodiment, the trajectory monitoring module further includes:

[0186] The dynamic frequency adjustment unit is configured to dynamically switch the transmission frequency of the millimeter wave radar according to the interference frequency distribution characteristics in the real-time electromagnetic environment monitoring result based on the adjusted radar detection configuration, and generate a signal after frequency adjustment through a frequency hopping sequence generation rule.

[0187] The waveform optimization unit is configured to adjust the pulse width and cycle duty ratio of the signal after frequency adjustment through time-frequency waveform parameter reconstruction, and obtain an optimized waveform signal.

[0188] The signal coding unit is configured to perform orthogonal code division multiplexing coding on the optimized waveform signal, and generate a coded signal. The coded signal is used for the transmission waveform parameter in the adjusted radar detection configuration.

[0189] In an embodiment, a computer device is provided, including a memory and a processor, the memory stores a computer program, and the processor implements the steps of the millimeter wave radar treadmill personnel presence detection method as described above when executing the computer program.

[0190] In an embodiment, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the steps of the above method embodiments.

[0191] For the device embodiment, since it basically corresponds to the method embodiment, the related parts are described in the method embodiment. The above described device embodiment is only schematic, and the components described as separate components can or can not be physically separate, and the components displayed as a unit can or can not be a physical unit, that is, they can be located in one place, or distributed on multiple network units. According to the actual needs, some or all of the modules can be selected to achieve the purpose of the present disclosure. Those skilled in the art can understand and implement it without creative labor.

[0192] The above described embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the patent scope of the application. It should be pointed out that for those skilled in the art, without departing from the concept of the present application, some modifications and improvements can be made, which are all within the protection scope of the present application.

Claims

1. A method for detecting the presence of a person on a treadmill using a millimeter-wave radar, characterized in that: The method comprises: S1. Based on a personnel location tracking grid, extract the target person's motion trajectory changes based on the original detection data related to the target person's motion at each grid coordinate point to generate person location information; the personnel location tracking grid is established based on a gridded coordinate system divided based on the spatial characteristics of the target treadmill monitoring area covered by the adjusted radar detection configuration; the original detection data is feature data related to the target person's motion obtained by processing the echo signal received after the millimeter-wave radar transmits electromagnetic waves according to the adjusted radar detection configuration; S2. Acquire vital sign signals of the target person, and obtain vital sign information by performing signal filtering and feature extraction on the vital sign signals; the vital sign signals include a respiratory signal and a heartbeat signal; S3. Based on the personnel location information and the vital sign information, combined with corresponding preset thresholds and discrimination criteria, comprehensively assess the presence status of the target personnel to obtain an assessment result; S4. When the assessment result shows that the target person has left, has abnormal vital signs, or has engaged in deceptive behavior, it is determined to be an abnormal state; the abnormal state is used to instruct the generation of a corresponding early warning instruction.

2. The method according to claim 1, characterized in that The S2 includes: S21, fusing the raw detection data and region identification information from each radar node to obtain fused detection data with a spatiotemporal identification; the region identification information is used to mark the grid region to which the raw detection data belongs, and the spatiotemporal identification includes grid coordinates and a timestamp; S22, matching the fused detection data to corresponding coordinate points of the personnel location tracking grid to generate an association mapping result; S23. Based on the correlation mapping result, convert the Doppler shift feature of the target person into a velocity vector of a corresponding grid point; the Doppler shift feature is extracted from the corresponding grid coordinate point of the fused detection data; S24, according to the timestamp sequence and the continuity of the grid coordinate points, the grid coordinate points of the same target person at different times, the associated fusion detection data, and the corresponding velocity vector are connected in time series to generate preliminary motion trajectory data; S25. Based on the trajectory points corresponding to consecutive time stamps in the preliminary motion trajectory data, calculate the spatial distance deviation of adjacent trajectory points using a preset distance threshold; the spatial distance deviation is used to compare with the threshold to identify abnormal trajectory points; S26, removing the abnormal trajectory points from the preliminary motion trajectory data, and completing the trajectory according to the spatiotemporal continuity of the remaining trajectory points to obtain corrected motion trajectory data; S27. Based on the corrected motion trajectory data, the temporal and spatial variation characteristics of consecutive trajectory points of the target person are captured and analyzed to obtain the velocity vector direction change rate of adjacent trajectory points; S28. Compare the velocity vector direction change rate with a preset threshold, extract corresponding trajectory points that exceed the preset threshold, and generate personnel location information including its grid coordinates and timestamp.

3. The method according to claim 2, characterized in that The S21 includes: S211, performing global time reference conversion and grid coordinate alignment processing on the original detection data and the area identification information from each radar node to obtain aligned detection data with a unified time and space stamp; S212: Calculate a weight coefficient based on the area overlap of adjacent nodes in the alignment detection data; the calculation formula of the weight coefficient is: Among them, W(x, y) is the weight coefficient, O i (x,y) is the area overlap of node i at the grid coordinate (x,y), R i is the reliability factor of node i, d c，i is the distance from the center of the detection area of ​​node i to the grid coordinate (x, y), and σ is the scale parameter of the distance attenuation factor; S213, screening the original detection data of each radar node, eliminating repeated detection data at the same grid coordinate, and obtaining screened detection data; S214: Based on the weight coefficient and the filtered detection data, a weighted fusion calculation is performed on the detection data after filtering multiple nodes of the same grid coordinate to obtain an initial fusion result; the initial fusion result is calculated using the following formula: Among them, D f (x, y) is the initial fusion result at the grid coordinate (x, y), N is the number of nodes involved in the fusion, d i (x,y) is the original detection data of node i at the grid coordinate (x,y); S215: retain the initial fusion results whose confidence levels are higher than a preset confidence threshold to obtain the fusion detection data.

4. The method according to claim 2, characterized in that After S28, the following steps are also included: S281, analyzing the angle information of the reflected signal of the target person located by the personnel position information to obtain a spatial angle distribution result of the target person; S282. Based on the spatial angle distribution result and in combination with the differences in radar cross-sectional areas of different parts of the human body, match the spatial angle distribution of the target person with various parts of the human body to obtain a matching relationship. S283: Based on the matching relationship, identifying the body orientation and movement posture of the target person through matching association to obtain posture information; S284, classifying and identifying the posture information based on a preset posture feature library to obtain a posture classification result; the posture feature library is updated based on historical data of normal and abnormal postures of a large number of people on the treadmill; S285. Based on the posture classification result, determine whether the target person is exercising normally facing forward on the treadmill, and obtain a motion state determination result. S286: Analyze the motion state judgment result, and when it is detected that the non-frontal orientation of the person exceeds a set threshold, determine it as an abnormal existence state.

5. The method according to claim 1, wherein The S3 includes: S31, performing a time sequence continuity check on the personnel location information, and when it is detected that the personnel location information suddenly disappears or the trajectory data is interrupted, displaying an assessment result that the target person has left; S32. Based on the vital sign information, different physiological signal patterns are classified and identified in combination with a preset physiological signal pattern library. When an abnormal or disappeared vital sign signal is detected, an assessment result of the vital sign abnormality is displayed.

6. The method according to claim 1, characterized in that The S4 includes: S41. Performing multi-source information fusion on the multi-dimensional information of the target person based on the person's location information and the vital sign information to obtain fused information; the multi-dimensional information includes radar scattering characteristics, motion characteristics, physiological signals, and environmental information; S42. Based on the fused information, a real person presence discrimination standard established by combining the human body electromagnetic wave reflection characteristic parameters and sports physiological characteristics is used to analyze the authenticity of the detected target person and obtain a comprehensive score; S43: When the comprehensive score is lower than a set threshold, it is determined that there is fraudulent behavior.

7. The method according to claim 1, characterized in that Before S1, it also includes: S11. Based on the adjusted radar detection configuration, dynamically switch the transmission frequency of the millimeter-wave radar using a frequency hopping sequence generation rule according to interference frequency distribution characteristics in the real-time electromagnetic environment monitoring results to generate a frequency-adjusted signal; the adjusted radar detection configuration is obtained by adjusting the monitoring area range and detection parameters of the millimeter-wave radar in the target treadmill environment; S12, adjusting the pulse width and cycle duty cycle of the frequency-adjusted signal by reconstructing time-frequency waveform parameters to obtain an optimized waveform signal; S13. Perform orthogonal code division multiplexing encoding on the optimized waveform signal to generate an encoded signal; the encoded signal is used for the transmission waveform parameters in the adjusted radar detection configuration.

8. A millimeter-wave radar treadmill person presence detection device, characterized in that: The device comprises: A trajectory monitoring module is configured to extract changes in the target person's motion trajectory and generate person location information based on the raw detection data related to the target person's motion at each grid coordinate point based on a person location tracking grid. The person location tracking grid is established as a gridded coordinate system based on the spatial characteristics of the target treadmill monitoring area covered by the adjusted radar detection configuration. The raw detection data is characteristic data related to the target person's motion obtained by processing the echo signal received after the millimeter-wave radar transmits electromagnetic waves according to the adjusted radar detection configuration. A vital sign signal monitoring module is used to obtain the vital sign signals of the target person and obtain vital sign information by performing signal filtering and feature extraction on the vital sign signals; the vital sign signals include respiratory signals and heartbeat signals; a presence status assessment module for comprehensively assessing the presence status of the target person based on the person's location information, the vital sign information, and the authenticity judgment result, in combination with corresponding preset thresholds and discrimination criteria, to obtain an assessment result; the authenticity judgment result is obtained by analyzing the matching degree between the human body's electromagnetic wave reflection characteristic parameters and the movement physiological characteristics; The evaluation result determination module is used to determine that it is an abnormal state when the evaluation result shows that the target person has left, has abnormal vital signs or has engaged in deceptive behavior; the abnormal state is used to indicate the generation of a corresponding early warning instruction.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images