An AI human body sensor integrating security, care and linkage functions

Abstract

The application discloses an AI human body sensor integrating security, care and linkage functions, comprising: a radar module for emitting radar signals and receiving reflected signals to obtain signal collection information; a security module for configuring a security strategy, wherein the security strategy is used for detecting illegal intrusion and border crossing behavior based on the signal collection information and triggering a first alarm mechanism; a care module for configuring a care strategy, wherein the care strategy is used for detecting the activity state of a person under care based on the signal collection information and triggering a second alarm mechanism; a linkage module for constructing a monitoring scene based on security information transmitted by the security module and care information transmitted by the care module, determining a linkage strategy based on the monitoring scene, and sending the linkage strategy to corresponding execution equipment; and a power module. The AI human body sensor realizes high-precision human body detection under the premise of protecting privacy, integrates security and care functions, reduces false positives and false negatives, constructs a multi-scene linkage mechanism, and realizes device collaborative response.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of sensors, in particular to an AI human sensor integrating security, care and linkage functions. BACKGROUND

[0002] With the rapid growth of intelligent security and smart pension demands, traditional sensors face many technical bottlenecks in human body detection and scene linkage: in the security field, cameras need to directly collect images, which may easily cause privacy disputes in home, nursing home and other scenes; the infrared sensor is significantly affected by light and temperature, and has a high false alarm rate in low light or complex environments; traditional sensors are difficult to distinguish human activities from pet or object movements, and are easy to produce false triggering. In the care field, wearable devices are inconvenient to use or may be missed; traditional sensors can only detect presence and are difficult to analyze complex activity states such as falling and long-time stillness in real time. Traditional security and care systems are mostly independently deployed, lack data fusion and coordination mechanisms, and security alarms and care abnormal events cannot be associated, resulting in low response efficiency; and the device compatibility is poor. SUMMARY

[0003] The present application aims to at least partly solve one of the above technical problems. To this end, the present application aims to provide an AI human sensor integrating security, care and linkage functions, which realizes high-precision human body detection under the premise of protecting privacy, integrates security and care functions, reduces false positives and false negatives, builds a multi-scene linkage mechanism, and realizes device collaborative response.

[0004] To achieve the above-mentioned purpose, the present application provides an AI human sensor integrating security, care and linkage functions, comprising:

[0005] a radar module for transmitting radar signals and receiving reflected signals to obtain signal acquisition information;

[0006] a security module for configuring a security strategy, the security strategy being based on signal acquisition information to detect illegal intrusion and boundary crossing behavior and trigger a first alarm mechanism;

[0007] a care module for configuring a care strategy, the care strategy being based on signal acquisition information to detect the activity state of the person being cared for and trigger a second alarm mechanism;

[0008] a linkage module for constructing a monitoring scene based on security information transmitted by the security module and care information transmitted by the care module, determining a linkage strategy based on the monitoring scene, and sending the linkage strategy to the corresponding execution device;

[0009] a power module for powering the radar module, the security module, the care module and the linkage module.

[0010] According to some embodiments of the present application, the security module comprises:

[0011] The noise reduction module is configured to extract echo signals in the signal acquisition information, demodulate the echo signals into intermediate frequency signals, input the intermediate frequency signals into a band-pass filter, and obtain noise reduction signals;

[0012] The first generation module is configured to perform distance estimation and Doppler frequency estimation on the noise reduction signals, obtain a distance gate and a Doppler frequency, generate a range-Doppler map based on the distance gate and the Doppler frequency, and perform joint probability data association based on the range-Doppler map to obtain a tracking trajectory of the target;

[0013] The second generation module is configured to perform short-time Fourier transform on the noise reduction signals to generate a time-frequency map, input the time-frequency map into a pre-trained CNN classifier, and output a type of the target.

[0014] The detection module is configured to set a virtual boundary, detect illegal intrusion and border-crossing behavior based on the tracking trajectory of the target, the type of the target, and the virtual boundary, and trigger a first alarm mechanism.

[0015] According to some embodiments of the present application, the first generation module comprises:

[0016] The distance estimation module is configured to perform pulse compression on the noise reduction signals to obtain a distance gate.

[0017] The Doppler frequency estimation module is configured to:

[0018] convert the noise reduction signals into a first Hankel matrix, perform SVD decomposition on the first Hankel matrix to obtain a singular value vector and left and right singular matrices, construct an optimization problem based on the sparsity of the singular value vector, and solve the optimization problem based on an iterative hard threshold to obtain a denoised singular value vector.

[0019]

[0020] construct a diagonal matrix based on the denoised singular value vector, construct a signal matrix based on the diagonal matrix and the left and right singular matrices, recover the Hankel structure through anti-diagonal average on the signal matrix, and obtain a second Hankel matrix.

[0021] calculate a covariance matrix of the second Hankel matrix, construct a MUSIC spectrum based on the covariance matrix, find a peak value, and determine a Doppler frequency.

[0022] generate a range-Doppler map based on the distance gate and the Doppler frequency.

[0023] According to some embodiments of the present application, the first generation module further comprises:

[0024] The first determination module is configured to: ​

[0025] extracting candidate measurements from the range-Doppler map;

[0026] determining a current tracking target set, a state of each target in the target set including a range, a radial velocity, a Doppler frequency and a Doppler frequency rate of change;

[0027] a constructing module configured to construct an association matrix, and split the association matrix into all possible association combinations, each measurement being associated with at most one target or clutter;

[0028] a calculating module configured to perform probability calculation on the split results, and filter out an association matrix corresponding to a maximum probability value to obtain a tracking trajectory of the target.

[0029] According to some embodiments of the present application, the detecting module comprises:

[0030] a second determining module configured to acquire a security area range, define a closed polygon area based on the security area range, and determine a virtual boundary;

[0031] a third determining module configured to determine an illegal intrusion and a crossing behavior based on whether the tracking trajectory of the target enters the virtual boundary and the type of the target, and trigger a first alarm mechanism.

[0032] According to some embodiments of the present application, the monitoring module comprises:

[0033] a posture detecting module configured to determine reflection characteristics and relative position relationships of each part of the person being monitored in the radar signal based on signal acquisition information, and judge a posture of the person being monitored; the posture includes standing, sitting, lying and falling;

[0034] a region detecting module configured to divide the monitoring area into a plurality of preset sub-areas, acquire reflection intensity and time delay of the radar signal based on the signal acquisition information, and judge a sub-area in which the person being monitored is located;

[0035] a direction detecting module configured to determine Doppler shift change of the radar signal and phase change of the reflection signal based on the signal acquisition information, and judge a moving direction of the person being monitored in the monitoring area;

[0036] a fourth determining module configured to determine an activity state of the person being monitored according to the posture of the person being monitored, the sub-area in which the person being monitored is located, and the moving direction of the person being monitored in the monitoring area.

[0037] According to some embodiments of the present application, the linkage module comprises:

[0038] The fifth determining module is used for determining a scene grid map according to a security area included in the security information transmitted by the security module and a care area included in the care information transmitted by the care module; dividing the scene grid map to obtain a plurality of grid blocks; and constructing a scene according to the plurality of grid blocks to obtain a scene terrain map;

[0039] The sixth determining module is used for determining first description language information corresponding to illegal intrusion and border crossing behavior based on the security information; and determining second description language information corresponding to an activity state of a person being cared for based on the care information.

[0040] The adding module is used for determining scene element information corresponding to the first description language information and the second description language information respectively, and adding the scene element information to the scene terrain map to obtain a monitoring scene.

[0041] According to some embodiments of the present application, the power supply state detecting module is further used for obtaining a current signal when the power supply module supplies power for the radar module, the security module, the care module and the linkage module, comparing the current signal with a preset current signal, and determining a power supply state of the power supply module according to a comparison result.

[0042] According to some embodiments of the present application, the solar charging panel is further used for charging the power supply module.

[0043] The trickle control module is used for controlling the solar charging panel to trickle charge the power supply module when a voltage value of the power supply module is less than a first voltage threshold or greater than a second voltage threshold.

[0044] According to some embodiments of the present application, the alarm module is further used for:

[0045] Obtaining charging data of N times of trickle control;

[0046] Calculating a trickle control parameter according to the charging data of N times of trickle control; calculating an absolute value of a difference between the trickle control parameter and 1; and when the absolute value of the difference is greater than a preset threshold, it is indicated that the trickle control is abnormal, and an alarm prompt is sent.

[0047] The application provides an AI human sensor integrating security, care and linkage functions, adopts a radar module to replace a camera, analyzes human characteristics through electromagnetic wave reflection signals, avoids direct image or video collection, eliminates privacy leakage risks from the source, realizes non-contact detection without the need of user wearing equipment, can realize long-term non-inductive monitoring, and improves the usability and acceptance of the care system. In the security scene, based on the micro-motion characteristic analysis of the radar signal, the illegal intrusion and pet and object movement are accurately distinguished, and the false alarm rate is reduced; through dynamic modeling of the cross-border area, the human position change is monitored in real time, and a hierarchical alarm mechanism (such as pre-alarm, sound and light alarm and remote notification) is triggered. In the care scene, the activity state of the person being cared for is detected in real time, the care efficiency is improved through the second alarm mechanism (such as sending a message to the family members and starting an emergency call), and the monitoring blind area of the traditional equipment is filled. The linkage module integrates the security information and the care information, constructs a dynamic monitoring scene, and automatically generates a linkage strategy: compatible with mainstream intelligent devices (such as intelligent lamps, door locks, cameras and alarms), realizes cross-brand device cooperation through a standardized interface, and breaks the system barriers.

[0048] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the application. The objects and other advantages of the present application can be achieved and obtained by the structure particularly pointed out in the written description and the accompanying drawings.

[0049] The technical solutions of the present application will be further described in detail below with the help of the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0050] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation on the present application. In the drawings:

[0051] Figure 1 is a block diagram of an AI human sensor integrating security, care and linkage functions according to an embodiment of the present application;

[0052] Figure 2 is a block diagram of a security module according to an embodiment of the present application;

[0053] Figure 3 is a block diagram of a first generation module according to an embodiment of the present application. DETAILED DESCRIPTION

[0054] The preferred embodiments of the present application are described below in combination with the accompanying drawings, and it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and do not limit the present application.

[0055] As Figure 1As shown in the figure, this invention proposes an AI human body sensor that integrates security, monitoring, and linkage functions, including:

[0056] The radar module is used to transmit radar signals and receive reflected signals to obtain signal acquisition information.

[0057] The security module is used to configure security policies. The security policies are based on signal acquisition information to detect illegal intrusion and boundary crossing behaviors, and trigger the first alarm mechanism.

[0058] The care module is used to configure care strategies. The care strategy is to detect the activity status of the person being cared for based on signal acquisition information and trigger a second alarm mechanism.

[0059] The linkage module is used to construct a monitoring scenario based on the security information transmitted by the security module and the care information transmitted by the care module, determine the linkage strategy based on the monitoring scenario, and send the linkage strategy to the corresponding execution device.

[0060] The power supply module is used to supply power to the radar module, security module, monitoring module, and linkage module.

[0061] The working principle and beneficial effects of the above technical solution are as follows: Replacing cameras with radar modules, it analyzes human characteristics through electromagnetic wave reflection signals, avoiding direct image or video acquisition and eliminating the risk of privacy leaks at the source. Non-contact detection eliminates the need for users to wear devices, enabling long-term, unobtrusive monitoring and improving the usability and acceptance of the caregiving system. In security scenarios, based on the micro-motion characteristic analysis of radar signals, it accurately distinguishes between illegal intrusions and the movement of pets or objects, reducing false alarm rates. Through dynamic modeling of boundary-crossing areas, it monitors changes in human position in real time, triggering the first alarm mechanism (such as pre-alarm, audible and visual alarm, and remote notification). In caregiving scenarios, it monitors the activity status of the person being cared for in real time, improving caregiving efficiency through a second alarm mechanism (such as sending text messages to family members or initiating emergency calls), filling the monitoring blind spots of traditional equipment. The linkage module integrates security and caregiving information to construct dynamic monitoring scenarios and automatically generates linkage strategies: it is compatible with mainstream smart devices (such as smart lights, door locks, cameras, and alarms), achieving cross-brand device collaboration through standardized interfaces, breaking down system barriers.

[0062] like Figure 2 As shown, according to some embodiments of the present invention, a security module includes:

[0063] The noise reduction module is used to extract the echo signal from the signal acquisition information and demodulate it into an intermediate frequency signal; the intermediate frequency signal is then input into a bandpass filter to obtain the noise-reduced signal.

[0064] The first generation module is configured to perform distance estimation and Doppler frequency estimation on the noise-reduced signal to obtain a range gate and a Doppler frequency, and generate a range-Doppler map based on the range gate and the Doppler frequency; and perform joint probability data association based on the range-Doppler map to obtain a tracking trajectory of the target.

[0065] The second generation module is configured to perform short-time Fourier transform on the noise-reduced signal to generate a time-frequency map, and input the time-frequency map into a pre-trained CNN classifier to output a type of the target.

[0066] The detection module is configured to set a virtual boundary, detect illegal intrusion and border-crossing behavior based on the tracking trajectory of the target, the type of the target and the virtual boundary, and trigger a first alarm mechanism.

[0067] The working principle of the above technical solution is as follows: The noise reduction module extracts an echo signal, suppresses noise and interference, and outputs a high-quality intermediate frequency signal, and further reduces noise through a band-pass filter. The intermediate frequency signal is subjected to matched filtering, i.e., pulse compression of the LFM signal, to obtain a range gate. The noise-reduced signal is subjected to Doppler frequency estimation to obtain a Doppler frequency. A range-Doppler map is generated based on the range gate and the Doppler frequency, wherein the horizontal axis is the Doppler frequency and the vertical axis is the range. The target state is updated based on the association probability of the signal point (measurement) in the range-Doppler map and the existing track to obtain a tracking trajectory of the target. The association probability is calculated based on the Mahalanobis distance or the Bayesian method. The noise-reduced signal is subjected to short-time Fourier transform to generate a time-frequency map, which is input into the input layer of the CNN classifier. The time-frequency features are extracted based on the convolution layer, and the type of the target, such as a person, a vehicle, an animal, etc., is output based on the fully connected layer. A virtual boundary is set. If a target (of a specified type, such as a vehicle or certain animals) enters from outside the boundary to inside the boundary, it is determined as illegal intrusion. If a target (of a specified type, corresponding to a person) enters from inside the boundary to outside the boundary, it is determined as border-crossing behavior.

[0068] The above technical solution has the following beneficial effects: illegal intrusion and border-crossing behavior can be accurately identified, and a first alarm mechanism is triggered, thereby improving the accuracy of security.

[0069] As shown in FIG. Figure 3 According to some embodiments of the present application, the first generation module comprises:

[0070] The distance estimation module is configured to perform pulse compression on the noise-reduced signal to obtain a range gate.

[0071] The Doppler frequency estimation module is configured to:

[0072] convert the noise-reduced signal into a first Hankel matrix; and perform SVD decomposition on the first Hankel matrix to obtain a singular value vector and left and right singular matrices.

[0073] An optimization problem is constructed based on the sparsity of the singular value vector; the optimization problem is solved based on an iterative hard threshold value to obtain a denoised singular value vector;

[0074] A diagonal matrix is constructed based on the denoised singular value vector; a signal matrix is constructed according to the diagonal matrix and left and right singular matrices; a second Hankel matrix is obtained by recovering a Hankel structure through anti-diagonal line averaging of the signal matrix;

[0075] A covariance matrix of the second Hankel matrix is calculated, a MUSIC spectrum is constructed according to the covariance matrix, and a peak value is searched to determine a Doppler frequency;

[0076] A range-Doppler map is generated based on the range gate and the Doppler frequency.

[0077] The working principle of the above technical solution is as follows: when the denoised signal is pulse compressed, matching filtering is performed, the received linear frequency modulation (LFM) signal is convolved with the conjugate time delay version of the transmitted signal, the signal pulse width is compressed, and the range resolution is improved. The range is 1 / 2 of the product of the speed of light and the time delay.

[0078] The denoised signal s(n) is converted into a first Hankel matrix, and each row is a continuous subsequence of the signal.

[0079]

[0080] Wherein, H1 is the first Hankel matrix; L is the number of rows, M is the number of columns, and N=L+M-1;

[0081] SVD decomposition is:

[0082] H1=U∑V H

[0083] Wherein, U, V are left and right singular matrices; Σ is a singular value diagonal matrix; V H is the conjugate transpose of V;

[0084] The singular value vector of the signal is sparse, and the singular value of the noise is small. The optimization problem is:

[0085]

[0086] Wherein, is the estimated value of Σ; is the square of the Frobenius norm of the difference between the two matrices, and the objective function is to minimize this square error; K is a given non-negative integer constant;

[0087] The optimization problem is solved based on an iterative hard threshold value to obtain a denoised singular value vector; A diagonal matrix is constructed based on the singular value vector after denoising. A signal matrix is constructed according to the diagonal matrix and left and right singular matrices; the signal matrix is restored to a one-dimensional signal through anti-diagonal line averaging, and repeated processing is performed to obtain a second Hankel matrix. A covariance matrix of the second Hankel matrix is calculated, the covariance matrix is subjected to eigenvalue decomposition to determine a signal subspace and a noise subspace, a MUSIC spectrum is constructed, and a peak value is searched to determine a Doppler frequency.

[0088] The above technical solution has the beneficial effects that: the distance gate is extracted by pulse compression; the noise is suppressed by Hankel matrix decomposition and SVD denoising; the Doppler frequency is extracted from the reconstructed signal by using the MUSIC algorithm, the range and velocity resolution is improved, and an accurate range-Doppler map is obtained.

[0089] According to some embodiments of the application, the first generation module further comprises:

[0090] The first determination module is configured to:

[0091] extract candidate measurements from the range-Doppler map;

[0092] determine a current target set for tracking, and the state of each target in the target set comprises a range, a radial velocity, a Doppler frequency and a Doppler frequency change rate;

[0093] The construction module is configured to construct a correlation matrix, and split the correlation matrix into all possible correlation combinations, each measurement is associated with at most one target or clutter.

[0094] The calculation module is configured to calculate the probability of the split result, filter out the correlation matrix corresponding to the maximum probability value, and obtain the tracking trajectory of the target.

[0095] The working principle of the above technical solution is that: the dimension of the correlation matrix A is MxN, A ij =1 indicates that the measurement i is associated with the target j, otherwise 0. The correlation combination is split, for example, for M=2 measurements and N=2 targets, the possible correlation combinations include: measurement 1→target 1, measurement 2→target 2. Measurement 1→target 1, measurement 2→clutter. Measurement 1→clutter, measurement 2→target 2. Calculate the probability of each correlation combination, filter out the correlation matrix corresponding to the maximum probability, and update the target tracking trajectory. For the measurement associated with the target, the target state is updated using Kalman filtering.

[0096] The above technical solution has the beneficial effects that: candidate measurements are extracted from the range-Doppler map, and the range, velocity, Doppler frequency and its change rate of the target are maintained. All possible measurement-target correlation combinations are exhausted. The maximum probability correlation matrix is selected to update the target state, which facilitates improving the accuracy of multi-target association and obtaining an accurate tracking trajectory.

[0097] According to some embodiments of the present application, the detection module comprises:

[0098] The second determination module is configured to obtain a security area range, define a closed polygon area based on the security area range, and determine a virtual boundary.

[0099] The third determination module is configured to determine illegal intrusion and boundary-crossing behavior based on whether the tracking trajectory of the target enters the virtual boundary and the type of the target, and trigger the first alarm mechanism.

[0100] The working principle of the above technical solution is as follows: the security area is defined through user interaction (such as map interface drawing) or configuration file (such as GeoJSON format). A closed polygon area is defined based on the security area range, and a virtual boundary is determined. The ray method or cross product method is used to determine whether the current position of the target is in the virtual boundary. After determining whether the current position of the target is in the virtual boundary, the type of the target is also used to determine the abnormal behavior.

[0101] The beneficial effects of the above technical solution are as follows: the security area range is obtained and converted into a closed polygon. Based on the target position and type, it is determined whether illegal intrusion or boundary-crossing behavior occurs. According to the detection result, the alarm mechanism is triggered. Combined with geometric algorithms and target tracking, the accuracy of behavior judgment is improved.

[0102] According to some embodiments of the present application, the care module comprises:

[0103] The posture detection module is configured to determine the reflection characteristics and relative position relationship of each part of the person being cared for in the radar signal based on the signal acquisition information, and determine the posture of the person being cared for. The posture includes standing, sitting, lying down, and falling down.

[0104] The area detection module is configured to divide the care area into a plurality of preset sub-areas, obtain the reflection intensity and time delay of the radar signal based on the signal acquisition information, and determine the sub-area in which the person being cared for is located.

[0105] The direction detection module is configured to determine the Doppler frequency shift change of the radar signal and the phase change of the reflection signal based on the signal acquisition information, and determine the moving direction of the person being cared for in the care area.

[0106] The fourth determination module is configured to determine the activity state of the person being cared for according to the posture of the person being cared for, the sub-area in which the person being cared for is located, and the moving direction of the person being cared for in the care area.

[0107] The working principle of the technical solution is as follows: the reflection points of different parts of the body (such as the head, torso, and limbs) are identified through the amplitude, phase, and Doppler shift of the radar echo signal. The ranging capability (based on time delay) and angle resolution of the radar are used to construct the relative spatial distribution of the body parts. The extracted features are compared with predefined posture templates (such as standing with a high head and dispersed limbs, and falling with a flat body and dramatic changes in Doppler shift). A classification model (such as SVM or CNN) is used to train the posture labels to obtain the posture classification results. The regional detection module divides the care area into multiple sub-areas, and judges the current sub-area of the person being cared for based on the reflection intensity and time delay of the radar signal. The direction detection module judges the moving direction of the person being cared for in the care area based on the Doppler shift and phase change. The activity state of the person being cared for is determined by integrating the posture, sub-area, and moving direction.

[0108] The beneficial effects of the technical solution are as follows: the activity state of the person being cared for can be accurately determined.

[0109] According to some embodiments of the present application, the linkage module comprises:

[0110] The fifth determination module is configured to determine a scene grid map according to the security area included in the security information transmitted by the security module and the care area included in the care information transmitted by the care module; divide the scene grid map to obtain a plurality of grid blocks; and construct a scene based on the plurality of grid blocks to obtain a scene terrain map.

[0111] The sixth determination module is configured to determine first description language information corresponding to illegal intrusion and border crossing behavior based on the security information, and determine second description language information corresponding to the activity state of the person being cared for based on the care information.

[0112] The adding module is configured to determine scene element information corresponding to the first description language information and the second description language information, respectively, and add the scene element information to the scene terrain map to obtain a monitoring scene.

[0113] The working principle of the technical solution is as follows: the coordinates of the security area (such as a virtual boundary or a restricted area) of the security module and the care area (such as a bedroom or a bathroom) of the care module are integrated. The combined area is divided into regular grids (such as 1m x 1m squares), and each grid block records its spatial attributes (such as whether it is a security restricted area or a care focus area). A two-dimensional or three-dimensional terrain map is generated based on the grid block attributes. Security behaviors (illegal intrusion, border crossing) and care states (posture, activity trajectory) are converted into structured description languages. The description language information is mapped to scene elements and embedded into the terrain map to generate an interactive monitoring scene. Among them, the security elements are intrusion point markers (red flashing icons) and border crossing paths (dashed arrows). The care elements are fall position heat maps and activity trajectory dynamic lines. The elements are superimposed on the terrain map in the form of independent layers.

[0114] The beneficial effects of the above technical solution are that the linkage system can realize deep integration of security and care information, build a precise and interactive monitoring scene, and improve emergency response efficiency.

[0115] According to some embodiments of the present application, further comprising: a power supply state detection module, configured to acquire a current signal when the power supply module supplies power to the radar module, the security module, the care module and the linkage module, compare the current signal with a preset current signal, and determine the power supply state of the power supply module according to the comparison result.

[0116] The working principle of the above technical solution is that a current sensor (such as a Hall current sensor) is connected in series on the line through which the power supply module supplies power to each module, and the current signal is collected in real time. According to the rated power, working characteristics and actual use scene of each module, the current range of each module when normally working is preset as the preset current signal. The real-time collected current signal of each module is compared with the corresponding preset current signal. If the real-time current signal is within the preset current range, it is determined that the power supply state of the power supply module for the module is normal. Otherwise, it indicates that the power supply state of the power supply module for the module is not normal.

[0117] The beneficial effects of the above technical solution are that through the effective operation of the power supply state detection module, the power supply problem of the power supply module can be found in time, the normal work of the radar, security, care and linkage modules is ensured, and the reliability and stability of the whole system are improved.

[0118] According to some embodiments of the present application, further comprising: a solar charging panel, configured to charge the power supply module;

[0119] A trickle control module is configured to control the solar charging panel to trickle charge the power supply module when the voltage value of the power supply module is less than a first voltage threshold or greater than a second voltage threshold.

[0120] The working principle and beneficial effects of the above technical solution are that the AI human body sensor integrating security, care and linkage functions has high integration, has multiple modules, and each module needs to work continuously, so there is a large power demand. Solar energy is converted into electrical energy to charge the power supply module, realizing the sustainability of system power supply. When the voltage of the power supply module is abnormal (too low or too high), the trickle charging mode is started to avoid damage to the battery caused by overcharging or deep discharge, prolonging the battery life.

[0121] According to some embodiments of the present application, further comprising: an alarm module, configured to:

[0122] Acquire charging data of N times of trickle control;

[0123] The trickle control parameters are calculated according to the charging data of N times of trickle control; the absolute value of the difference between the trickle control parameters and 1 is calculated, and when it is determined that the absolute value of the difference is greater than a preset threshold, it indicates that the trickle control is abnormal, and an alarm prompt is issued.

[0124] The working principle of the above technical solution is: the charging data when the solar charging panel performs trickle charging for the power module is obtained based on the alarm module, the trickle control parameters are calculated according to the charging data of N times of trickle control; the absolute value of the difference between the trickle control parameters and 1 is calculated, and when it is determined that the absolute value of the difference is greater than a preset threshold, it indicates that the trickle control is abnormal, and an alarm prompt is issued, which is convenient for timely adjusting the control parameters of the trickle control module, facilitating safe and reliable charging of the power module, and protecting the power module.

[0125] The trickle control parameters are calculated according to the charging data of N times of trickle control, including:

[0126]

[0127] Wherein, K is the trickle control parameter; β is the current difference parameter;

[0128]

[0129] Wherein, A i is the theoretical current of the i-th time of trickle control; a i is the actual current of the i-th time of trickle control; is the integral of , and the integral parameter is x, the upper limit of integration is υ, and the lower limit of integration is -υ, and υ is a preset fault tolerance parameter.

[0130] K represents the deviation distribution characteristics of the actual current and the theoretical current, and is used to evaluate the stability of the charging process. β reflects the discrete degree of the actual current around the theoretical value. υ is related to the setting accuracy. is the integral of the probability density function of the Gaussian distribution in the interval [-υ, υ]. For example, N = 3, A i = [50, 50, 50], a i = [48, 52, 49], then is 0.333 mA; then β is 1.68 mA; calculate Numerical integration (such as trapezoidal method or Simpson method) is adopted. The approximation is the cumulative distribution function (CDF) difference value of the Gaussian distribution, Wherein, erf() is the error function, which can be determined by table lookup or library function. For example, υ = 2β = 3.36 mA, then K=0.436. K is close to 1, and the deviation distribution of the actual current from the theoretical current is concentrated near 1, that is, the absolute value of the difference between the trickle control parameter and 1 is less than or equal to a preset threshold. K is far from 1, that is, the absolute value of the difference between the trickle control parameter and 1 is greater than the preset threshold, indicating that there is abnormal fluctuation. The preset threshold is a threshold for judging whether there is fluctuation, and is set according to detection requirements, and the value range is (0.05, 0.3). The higher the detection requirement, the smaller the preset threshold.

[0131] The beneficial effects of the above technical solutions are: combining statistical average and dispersion degree, comprehensively reflecting the fluctuation characteristics of the charging process, effectively evaluating the stability of the trickle charging process, and providing a quantitative index for abnormal detection. Precise monitoring and abnormal early warning of the trickle charging process are realized, and the safety and reliability of the power module charging are improved.

[0132] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. An AI human body sensor integrating security, monitoring, and linkage functions, characterized in that, include: The radar module is used to transmit radar signals and receive reflected signals to obtain signal acquisition information. The security module is used to configure security policies. The security policies are based on signal acquisition information to detect illegal intrusion and boundary crossing behaviors, and trigger the first alarm mechanism. The care module is used to configure care strategies. The care strategy is to detect the activity status of the person being cared for based on signal acquisition information and trigger a second alarm mechanism. The linkage module is used to construct a monitoring scenario based on the security information transmitted by the security module and the care information transmitted by the care module, determine the linkage strategy based on the monitoring scenario, and send the linkage strategy to the corresponding execution device. The power supply module is used to supply power to the radar module, security module, monitoring module and linkage module; Security module, including: The noise reduction module is used to extract the echo signal from the signal acquisition information and demodulate it into an intermediate frequency signal; the intermediate frequency signal is then input into a bandpass filter to obtain the noise-reduced signal. The first generation module is used to perform range estimation and Doppler frequency estimation on the denoised signal to obtain the range gate and Doppler frequency, and generate a range-Doppler map based on the range gate and Doppler frequency; and perform joint probability data association based on the range-Doppler map to obtain the target tracking trajectory. The second generation module is used to perform a short-time Fourier transform on the denoised signal to generate a time-frequency graph. The time-frequency graph is then input into a pre-trained CNN classifier to output the type of the target. The detection module is used to set virtual boundaries; it detects illegal intrusion and boundary crossing behaviors based on the target's tracking trajectory, target type, and virtual boundary, and triggers the first alarm mechanism. The first generation module includes: The distance estimation module is used to perform pulse compression on the noise-reduced signal to obtain the distance gate; The Doppler frequency estimation module is used for: The denoised signal is converted into a first Hankel matrix; the first Hankel matrix is ​​decomposed by SVD to obtain the singular value vector and the left and right singular matrices; Based on the sparsity of singular value vectors, an optimization problem is constructed; the optimization problem is solved by iterative hard thresholding to obtain the denoised singular value vectors. Construct a diagonal matrix based on the denoised singular value vectors; construct a signal matrix based on the diagonal matrix and the left and right singular matrices; recover the Hankel structure from the signal matrix by anti-diagonal averaging to obtain the second Hankel matrix; Calculate the covariance matrix of the second Hankel matrix, construct the MUSIC spectrum based on the covariance matrix and find the peak value to determine the Doppler frequency; Range-Doppler maps are generated based on range gates and Doppler frequencies; The first generation module also includes: The first determining module is used for: Extracting candidate measurements from distance-Doppler maps; Determine the current set of tracked targets. The state of each target in the target set includes distance, radial velocity, Doppler frequency, and rate of change of Doppler frequency. The building module is used to construct the correlation matrix, which is broken down into all possible correlation combinations. Each measurement is associated with at most one target or clutter. The calculation module is used to perform probability calculations on the splitting results, filter out the correlation matrix corresponding to the highest probability value, and obtain the target's tracking trajectory.

2. The AI ​​human body sensor integrating security, monitoring, and linkage functions as described in claim 1, characterized in that, The detection module includes: The second determination module is used to obtain the security area range, define a closed polygonal region based on the security area range, and determine the virtual boundary; The third determination module is used to determine illegal intrusion and boundary crossing behaviors based on whether the target's tracking trajectory enters the virtual boundary and the type of the target, and to trigger the first alarm mechanism.

3. The AI ​​human body sensor integrating security, monitoring, and linkage functions as described in claim 1, characterized in that, The care module includes: The posture detection module is used to determine the reflection characteristics and relative positional relationships of various parts of the person being cared for in radar signals based on signal acquisition information, and to determine the posture of the person being cared for; the posture includes standing, sitting, lying down and falling. The area detection module is used to divide the care area into multiple preset sub-areas, and to determine the sub-area where the person being cared for is located based on the radar signal reflection intensity and time delay obtained from the signal acquisition information. The direction detection module is used to determine the Doppler frequency shift of the radar signal and the phase change of the reflected signal based on the signal acquisition information, and to determine the direction of movement of the person being cared for within the care area; The fourth determination module is used to determine the activity status of the person being cared for based on the person's posture, the sub-area where the person is located, and the direction of movement of the person within the care area.

4. The AI ​​human body sensor integrating security, monitoring, and linkage functions as described in claim 1, characterized in that, The linkage module includes: The fifth determining module is used to determine the scene grid map based on the security area included in the security information transmitted by the security module and the care area included in the care information transmitted by the care module; divide the scene grid map into several grid blocks; and construct the scene based on the several grid blocks to obtain the scene terrain map. The sixth determination module is used to determine the first descriptive language information corresponding to illegal intrusion and boundary crossing behavior based on security information; and to determine the second descriptive language information corresponding to the activity status of the person being cared for based on care information. An addition module is used to determine the scene element information corresponding to the first description language information and the second description language information respectively, and add them to the scene topographic map to obtain the monitoring scene.

5. The AI ​​human body sensor integrating security, monitoring, and linkage functions as described in claim 1, characterized in that, Also includes: The power supply status detection module is used to acquire the current signal when the power supply module supplies power to the radar module, security module, monitoring module and linkage module, compare the current signal with the preset current signal, and determine the power supply status of the power supply module based on the comparison result.

6. The AI ​​human body sensor integrating security, monitoring, and linkage functions as described in claim 5, characterized in that, Also includes: Solar charging panels are used to charge the power module; The trickle charge control module is used to control the solar charging panel to trickle charge the power module when the voltage value of the power module is less than a first voltage threshold or greater than a second voltage threshold.

7. The AI ​​human body sensor integrating security, monitoring, and linkage functions as described in claim 6, characterized in that, Also includes: Alarm module, used for: Obtain charging data from N trickle-feed control cycles; Based on the charging data from N trickle control cycles, trickle control parameters are calculated. The absolute value of the difference between the trickle control parameters and 1 is calculated. When the absolute value of the difference is greater than a preset threshold, it indicates that the trickle control is abnormal and an alarm is issued.

Images