Door-to-door pension service quality evaluation and monitoring system based on millimeter wave radar technology

Abstract

The invention discloses a door-to-door pension service quality evaluation and monitoring system based on a millimeter wave radar technology. The system comprises a millimeter wave MIMO radar module, a signal preprocessing and trajectory analysis unit, a service type calibration unit, a target distinguishing unit, a service quality evaluation unit, a comprehensive scoring unit, a selectable event identification module and an anomaly detection unit. The millimeter-wave radar module outputs indoor three-dimensional trajectory data to the signal preprocessing and trajectory analysis unit, and the indoor three-dimensional trajectory data is subjected to density clustering and improved Kalman filtering processing and then is output to the target distinguishing unit; the target distinguishing unit determines a reference position of the elderly by taking blood pressure measurement as a predefined calibration action, and divides a track by using a motion feature template of a caregiver; the evaluation unit generates distance and thermodynamic diagram indexes; the comprehensive scoring unit weights and outputs scores according to the types of the old people; and the abnormity detection unit gives an alarm when the indexes are abnormal, thereby realizing non-contact and high-precision door-to-door pension service monitoring and evaluation without privacy risk.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of quality evaluation and monitoring, and relates to a door-to-door old-age service quality evaluation and monitoring system based on a millimeter wave radar technology. BACKGROUND

[0002] With the acceleration of the process of social aging, the demand for home-based care and door-to-door nursing services has risen sharply, and the demand for objective and accurate quality evaluation of the nursing service process by regulatory authorities, service agencies and family members has become increasingly urgent.

[0003] Some institutions attempt to introduce video monitoring to obtain service process information, although the accuracy is high, but the video will directly expose the living environment and privacy of the old people, and it is easy to cause resistance. At the same time, there are risks in the security of video data storage and transmission, and long-term preservation will also bring compliance and data management burdens. Therefore, non-video monitoring technology has become a trend, but the existing old-age service monitoring technology has obvious defects.

[0004] Traditional GPS positioning has serious signal attenuation indoors, and the positioning accuracy is insufficient, and it is easy to appear fake or subjective deviation by relying on mobile phone clock-in, manual visit and other methods. More importantly, the existing technology lacks effective means to distinguish the indoor activity trajectories of nursing staff and old people, neither a mechanism to determine the reference position of the old people based on pre-defined calibration actions, nor a nursing staff motion feature template to realize accurate separation of the two types of trajectories, resulting in trajectory confusion and loss of reliable data basis for subsequent evaluation.

[0005] The service quality evaluation dimension is single. The existing scheme focuses on simple indicators such as service time, and does not build a distance indicator reflecting the spatial proximity of nursing staff and old people, nor does it lack a heat map indicator depicting the indoor activity distribution of nursing staff, which cannot comprehensively and objectively present the effectiveness of nursing and the spatial service coverage of nursing staff.

[0006] In addition, the comprehensive scoring link does not consider the differentiated needs of different types of old people, and does not adjust the evaluation indicators for different types of old people, resulting in a scoring result that is out of touch with the actual nursing needs, making it difficult to accurately measure the service quality. These problems make it difficult for existing monitoring technology to meet the requirements of high-quality and fine evaluation of door-to-door old-age services, and a technical solution is needed that can accurately distinguish trajectories, evaluate multiple dimensions, and conduct differentiated comprehensive scoring. SUMMARY

[0007] To solve the problems in the background art, the application provides a door-to-door old-age service quality evaluation and monitoring system based on a millimeter wave radar technology.

[0008] To achieve the above purpose, the technical solution adopted by the application is as follows: The door-to-door old-age care service quality evaluation and monitoring system based on millimeter wave radar technology, characterized in that, comprising: A target distinguishing unit determines the old person reference position based on the pre-defined calibration action, and distinguishes the caregiver trajectory and the old person trajectory by using the caregiver motion feature template; A service quality evaluation unit generates distance indicators and heat map indicators according to the distinguished trajectories; the distance indicators reflect the spatial proximity between the caregiver and the old person; and the heat map indicators reflect the spatial activity distribution of the caregiver; A comprehensive scoring unit weights and sums the distance indicators and the heat map indicators according to the old person type to output the service quality score.

[0009] Specifically, further comprising: A millimeter wave MIMO radar module arranged in the old person's residence to output indoor three-dimensional trajectory data; A signal preprocessing and trajectory analysis unit receives the trajectory data of the millimeter wave MIMO radar module, performs preprocessing and multi-target trajectory tracking, and then outputs to the target distinguishing unit; A service type calibration unit inputs the old person type to the target distinguishing unit and the comprehensive scoring unit simultaneously.

[0010] Specifically, the signal preprocessing and trajectory analysis unit performs the following before outputting the trajectory to the target distinguishing unit: Clustering the three-dimensional trajectory data based on a density clustering algorithm; Implementing multi-target trajectory tracking by using an improved Kalman filtering algorithm.

[0011] Specifically, the pre-defined calibration action is a standard service action for initially serving the old person; Receiving the trajectory data after the end of the standard service action, extracting the kinematic features of the caregiver standard service action end process to generate the caregiver motion feature template.

[0012] Specifically, the target distinguishing unit performs the following according to the old person type: When the old person type is disabled, the trajectory data that is long-term in a closed area with the old person reference position as the center and a preset radius and has a speed lower than a speed threshold is determined as the old person trajectory; When the old person type is self-care, the similarity of each trajectory to the caregiver motion feature template is calculated, and the trajectory with the highest similarity is classified as the caregiver trajectory.

[0013] Specifically, the service quality evaluation unit performs the following calculations: calculating the distance indicators and calculating the heat map indicators; The calculation of the distance indicators includes the near-distance accompanying time proportion and the effective proximity frequency; The calculation of the heat map indicators includes the spatial coverage rate and the long-time stay penalty ratio.

[0014] Specifically, the comprehensive scoring unit executes: When the old person type is disabled, a first weight value is assigned to the near-distance accompanying time proportion; When the old person type is self-care, a second weight value is assigned to the space coverage rate.

[0015] Specifically, it further comprises an event identification module for: Identifying standard service actions.

[0016] Specifically, it further comprises an anomaly detection unit for: Receiving the service quality score output by the comprehensive scoring unit; Receiving the long-time stay penalty ratio output by the service quality assessment unit; When the long-time stay penalty ratio exceeds the preset alert value or the service quality score is lower than the anomaly threshold, an alarm is generated.

[0017] Specifically, the service type designation unit, at the service start: Receives the old person type input through a physical button or an interactive interface; Transmits the old person type to the target area differentiation unit and the comprehensive scoring unit in real time.

[0018] Compared with the prior art, the present application has the following beneficial effects: the present application sets up a target area differentiation unit, accurately determines the old person reference position relying on pre-defined designation actions, and realizes effective differentiation of the trajectories of the caregiver and the old person by using the caregiver motion feature template, completely solving the problem of easy confusion of the two types of trajectories and unreliable data basis in traditional monitoring, and providing accurate trajectory data support for subsequent quality assessment.

[0019] The service quality assessment unit generates a distance index reflecting the spatial proximity of the caregiver and the old person, and a heat map index depicting the indoor activity distribution of the caregiver, breaking the limitation of single evaluation dimension in traditional evaluation, and comprehensively and objectively presenting the effectiveness of the caregiver's accompanying and the spatial service coverage, avoiding the one-sidedness of relying only on simple indicators such as time length.

[0020] The comprehensive scoring unit weights and sums the distance index and the heat map index according to the old person type, fully adapts to the differentiated care needs of different old people, makes the service quality score more in line with the actual care scene, and effectively avoids the distortion problem caused by the one-size-fits-all of traditional scoring.

[0021] At the same time, combined with the non-contact and high-precision characteristics of the millimeter wave radar, the above-mentioned accurate evaluation is realized, the privacy of the old people is protected, the risk of privacy leakage of video monitoring is avoided, and the related data can be encrypted and uploaded and stored locally, providing reliable basis for service traceability, dispute handling and quality improvement, and significantly improving the scientificity, objectivity and practicality of the quality assessment and monitoring of home-based care services. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 is the overall appearance of the equipment of the home-based elderly care service quality evaluation and monitoring system based on millimeter wave radar technology of the present application; Figure 2 is the front structure schematic diagram of the core circuit board of the present application; Figure 3 is the back structure schematic diagram of the core circuit board of the present application; Figure 4 is the installation and service scene schematic diagram of the millimeter wave radar monitoring equipment in the home-based elderly care service scene of the present application; Figure 5 is the overall working process diagram of the home-based elderly care service quality evaluation and monitoring system based on millimeter wave radar technology of the present application; Figure 6 is the two-dimensional projection and DBSCAN typical clustering result schematic diagram of the millimeter wave radar point cloud collection of the present application; Figure 7 is the target distinguishing strategy flowchart of the care worker and the old person's trajectory in the home-based elderly care service of the present application; Figure 8 is the indoor activity distribution heat map schematic diagram of the care worker in the home-based elderly care service process of the present application.

[0023] In the figure: 101, millimeter wave MIMO radar module; 102, signal preprocessing and trajectory analysis unit; 103, wireless communication unit; 104, power supply unit; 105, indicator light interaction unit; 106, service type calibration unit; 107, local storage unit; 108, GPS module; 109, recording module. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0025] As shown in Figures 1-8 The technical solutions adopted by the present application are as follows: a home-based elderly care service quality evaluation and monitoring system based on millimeter wave radar technology, comprising: A target distinguishing unit determines the old person's reference position based on a predefined calibration action, and distinguishes the care worker's trajectory and the old person's trajectory by using a care worker motion feature template.

[0026] The predefined calibration action is a standard service action for initial service of the old person; The trajectory data after the end of the standard service action is received, and the kinematic features of the end of the standard service action of the caregiver are extracted to generate a caregiver motion feature template. The standard service action is the standardized service content of the caregiver to the old person, such as blood pressure measurement, body temperature measurement, body wiping, etc.

[0027] In a specific embodiment, the target distinguishing unit takes a pre-defined calibration action (such as a blood pressure measurement action) as the core basis, receives the trajectory data of the caregiver and the old person generated by the signal preprocessing and trajectory analysis unit 102 during the blood pressure measurement action, and determines the reference position of the old person by recording the fixed position (such as the bedside, sitting or lying position) of the old person during this process.

[0028] At the same time, after the end of the blood pressure measurement, the subsequent trajectory data output by the signal preprocessing and trajectory analysis unit 102 is received, and the kinematic features (such as the speed curve, acceleration curve, and displacement mode) of the caregiver when returning the blood pressure instrument are extracted to generate a caregiver motion feature template.

[0029] In distinguishing the trajectory, the target distinguishing unit performs differentiated logic in combination with the old person type input by the service type calibration unit: if the old person type is disabled, the trajectory that has been in a closed area with the old person reference position as the center and a preset radius for a long time and has a speed lower than a speed threshold in the trajectory output by the signal preprocessing and trajectory analysis unit 102 is determined as the old person trajectory.

[0030] If the old person type is self-care, the similarity of each trajectory to the caregiver motion feature template is calculated, the trajectory with the highest similarity is classified as the caregiver trajectory, and the remaining trajectories are classified as the old person trajectory.

[0031] The service quality evaluation unit generates a distance index and a heat map index based on the distinguished trajectories.

[0032] The distance index reflects the spatial proximity of the caregiver and the old person.

[0033] The heat map index reflects the spatial activity distribution of the caregiver.

[0034] The service quality evaluation unit first receives the separated caregiver trajectory and old person trajectory output by the target distinguishing unit, and generates the distance index and the heat map index based on the two types of trajectories.

[0035] The comprehensive scoring unit weights and sums the distance index and the heat map index according to the old person type to output a service quality score.

[0036] The comprehensive scoring unit first receives the old person type input by the service type calibration unit, and receives the normalized distance index and heat map index output by the service quality evaluation unit.

[0037] Subsequently, the weighting coefficients of the two types of indicators were adjusted according to the type of elderly person: if the elderly person is disabled, the first weighting value is assigned to the proportion of close companionship time in the distance indicator, focusing on assessing the closeness of companionship.

[0038] If the elderly person is classified as self-reliant, a second weight value is assigned to the spatial coverage rate in the heat map indicators, focusing on assessing the integrity of activity coverage.

[0039] Finally, the weighting coefficients are multiplied by the corresponding normalized indicators and then summed to obtain the final service quality score.

[0040] In one specific embodiment, the in-home elderly care service quality assessment and monitoring system based on millimeter-wave radar technology further includes: The millimeter-wave MIMO radar module 101 is installed in the elderly's living area to output indoor three-dimensional trajectory data.

[0041] The millimeter-wave MIMO radar module 101 is installed in the elderly's residence and uses millimeter-wave technology to detect the indoor scene. By collecting the spatial position information of caregivers and elderly people when they are active indoors, it generates and outputs indoor three-dimensional trajectory data containing the position coordinates of the two people. This data is then directionally transmitted to the signal preprocessing and trajectory analysis unit 102 as the raw data source for subsequent data processing.

[0042] The millimeter-wave MIMO radar module 101 provides the system with real and continuous raw data of the three-dimensional trajectory of indoor personnel activities. Without the three-dimensional trajectory data output by this module, the work of the signal preprocessing and trajectory analysis unit 102 and subsequent units will lose its data foundation. It is a prerequisite for realizing subsequent trajectory differentiation and quality assessment.

[0043] like Figure 1 The diagram shows the equipment of the in-home elderly care service quality assessment and monitoring system based on millimeter-wave radar of this invention. This equipment can achieve the following functions using existing conventional technologies. The implemented millimeter-wave radar-based in-home care service quality assessment equipment adopts a portable, integrated shell structure. It is small in size and lightweight, allowing for easy carrying and rapid deployment by caregivers during in-home services. The front of the equipment shell features status indicator lights, operation buttons, and necessary ventilation openings to ensure equipment stability and ease of use.

[0044] like Figure 2 The front view of the core circuit board of the device is shown, including a millimeter-wave MIMO radar module 101, a signal preprocessing and trajectory analysis unit 102, a wireless communication unit 103, a power supply unit 104, and an indicator light interaction unit 105.

[0045] Among them, the millimeter-wave MIMO radar module 101 is used to collect high-resolution point cloud data of the activity areas of caregivers and the elderly.

[0046] The signal preprocessing and trajectory analysis unit 102 is responsible for denoising, clustering, trajectory tracking and service behavior pattern analysis of point cloud data.

[0047] The wireless communication unit 103 (supporting 4G network) can upload the analysis results to the cloud management platform in real time.

[0048] The power supply unit 104 is externally connected to a rechargeable lithium battery, combined with a DC-DC voltage stabilizing module to ensure stable operation of the power supply part of the device.

[0049] The indicator light interaction unit 105 integrates start / end / type selection buttons and multi-color status indicator lights, making it easy for caregivers to quickly complete device configuration and status confirmation, as well as battery remaining capacity prompts.

[0050] As shown in Figure 3 The back structure of the device core circuit board includes the service type calibration unit 106, the local storage unit 107, the GPS module 108, and the recording module 109.

[0051] The service type calibration unit 106 can input the old person's type (self-care / disabled) through the interface or the key before the service starts, so that the system can use differentiated index weights and analysis rules in subsequent analysis.

[0052] The local storage unit 107 contains an SD card for storing raw data and analysis results of the service process.

[0053] The GPS module 108 can provide accurate verification of location and time when needed; the recording module 109 can collect necessary voice communication data under the premise of privacy compliance to assist in judging the quality of caregiver companionship.

[0054] Specifically, it also includes an event recognition module for: Receiving the trajectory micro-Doppler features output by the signal preprocessing and trajectory analysis unit 102.

[0055] Recognizing standard service actions.

[0056] The event recognition module mainly identifies service process events, which are based on trajectory and local micro-motion features, specifically micro-Doppler components and short-time speed change values. The system can detect and record several key events, such as approaching and staying for auxiliary care (close contact and low-speed continuous), mopping / sweeping actions (continuous movement covering a large ground area), changing bed linen / arranging bedding (high-density stay around the bed with low-amplitude micro-motion), etc. Event detection is based on rule thresholds and trained classifiers, and event timestamps are recorded together with trajectories to local storage and upload packages.

[0057] The overall operation flowchart of the door-to-door old-age service quality evaluation and monitoring system based on millimeter wave radar technology is shown in Figure 4

[0058] The first step is the arrival of the caregiver, the second step is the completion of the deployment and initialization of the equipment, the third step is the start of recording, the fourth step is the measurement of blood pressure, the fifth step is the normal operation of the equipment and the reporting of data, the sixth step is the end of the service, and the seventh step is the shutdown of the equipment, thus ending the door-to-door old-age service process.

[0059] In a specific embodiment, as shown in Figure 5 , a typical device installation scene in a door-to-door old-age service is shown. After the caregiver arrives, the device is placed at the predetermined placement point in the old person's bedroom. The placement point should cover the old person's bed and the caregiver's working area, ensuring that the radar horizontal detection angle of about 120 degrees and the maximum detection distance of about 7 meters can include the old person's bed and a larger working radius. The caregiver starts the service by pressing the start button for a long time, and selects the type of the person being cared for as self-care or disabled in the system.

[0060] After selection, the caregiver clicks the start recording button of the door-to-door old-age service monitoring device, and the device starts recording.

[0061] Optionally, if necessary, the recording function can be turned on through system settings.

[0062] Calibration of the old person's position: After starting recording, the caregiver first completes the blood pressure measurement for the old person at the old person's bedside as prompted. This action is considered a pre-defined calibration action, which is used to record and fix the reference position of the person being cared for and the reference point of the room coordinate system. The pre-defined calibration action is a standard service action for the initial service of the old person, which can be blood pressure measurement, body temperature measurement, etc. The radar will determine the old person's position at this time.

[0063] The normal operation steps of the device include data acquisition and preprocessing, point cloud projection and ground processing, point cloud clustering, front and back frame data association and trajectory production, state estimation and improved Kalman filtering, target differentiation strategy, occluded trajectory loss and re-identification, and service process event identification. At the same time, real-time data including target position and service status are uploaded to the cloud at a fixed frequency.

[0064] The signal preprocessing and trajectory analysis unit 102 is a prior art. It receives the trajectory data of the millimeter wave MIMO radar module 101, performs preprocessing and multi-target trajectory tracking, and outputs to the target differentiation unit; ​The signal preprocessing and trajectory analysis unit 102 first receives the indoor three-dimensional trajectory data transmitted by the millimeter wave MIMO radar module 101, performs preprocessing (such as denoising, filtering, and removing invalid data caused by environmental interference to ensure data accuracy) and multi-target trajectory tracking (using clustering, trajectory association, etc. Algorithm, the trajectories of the two targets of the indoor caregiver and the old man are tracked separately to avoid confusion of multi-target trajectories in the original data) on the original trajectory data, and after processing is completed, the clear and distinguishable multi-target trajectory data is output to the target distinguishing unit.

[0065] The signal preprocessing and trajectory analysis unit 102 purifies and optimizes the original three-dimensional trajectory data on the one hand, and solves the interference and confusion problems that may exist in the original data. On the other hand, through multi-target trajectory tracking, the original data is converted into effective data that can be used for trajectory differentiation, ensuring that the trajectory data transmitted to the target distinguishing unit is analyzable, and providing qualified data input for the target distinguishing unit to accurately distinguish the trajectories of caregivers and the elderly.

[0066] Specifically, the signal preprocessing and trajectory analysis unit 102 performs the following before outputting the trajectory to the target distinguishing unit: Clustering three-dimensional trajectory data based on density clustering algorithm; Implementing multi-target trajectory tracking using improved Kalman filtering algorithm.

[0067] The millimeter wave MIMO radar module 101 uses a millimeter wave MIMO radar. The millimeter wave MIMO radar has a frame rate of 10 Hz. ( configurable) output each frame of detection point cloud.

[0068] Specifically, for each frame, first perform radar front-end signal processing: based on FMCW signal, do single fast time dimension FFT to get range dimension spectrum, then do FFT on multi-block along time to get range-Doppler spectrum result, then do angle dimension FFT to get angle information, combined with signal amplitude, get a group of detection points through CFAR detector.

[0069] Each detection point contains 4 dimensions of data, respectively: distance , horizontal angle , pitch angle , and velocity . Convert ( ) to Cartesian coordinates ( ) to get three-dimensional point cloud. To reduce the interference of environmental static echoes, long-term background modeling exponential weighted average algorithm and static cluster removal method are used to remove points that exist stably for a long time. At the same time, threshold filtering is performed on points with low amplitude and low SNR. The above preprocessing is all completed in the signal preprocessing and trajectory analysis unit 102 to reduce the burden of uploading and subsequent processing.

[0070] Point cloud clustering: On the 2D projection plane, DBSCAN density clustering is performed on the aggregated point cloud within each time window (here set to 5 frames). The clustering parameters (neighborhood distance eps, minimum point number minPts) can be adaptively adjusted according to the room grid size and radar point density. The clustering outputs several clusters, each of which represents a candidate target position. The cluster center position is taken as the observation position of the target in the current frame. To improve the consistency of clustering between consecutive frames, exponential smoothing based on historical trajectories can be performed on the cluster centers after clustering each frame. Figure 6 A schematic diagram of 2D projection and typical clustering is given.

[0071] Front and rear frame data association and trajectory generation: A data association algorithm is used to associate the cluster center of each frame with the trajectory of the last frame and multiple frames. First, a cost matrix is constructed , where the cost ; , where is the i-th row and j-th column element in the constructed cost matrix , representing the matching cost between the cluster center of the current frame and the historical multi-frame trajectory. The smaller the cost value, the stronger the association between the cluster center of the current frame and the historical trajectory, and the more suitable it is as a continuous trajectory segment of the same target.

[0072] represents the Euclidean distance calculation method, is the position vector of the cluster center of the current frame (i.e. the spatial position coordinates of a candidate target obtained by DBSCAN density clustering in the current frame), is the position vector of the target in the historical trajectory (i.e. the spatial position coordinates of a confirmed target trajectory in the previous frame or multiple frames); the physical meaning of the entire term is the Euclidean distance between the spatial positions of the candidate target in the current frame and the target in the historical trajectory, reflecting the positional consistency of the two, and the smaller the distance, the stronger the positional association.

[0073] is the estimated velocity of the cluster center of the current frame, is the estimated velocity of the target in the historical trajectory, i.e. the motion velocity of a confirmed target trajectory in the previous frame or multiple frames, estimated by an improved Kalman filter algorithm.

[0074] is the absolute value difference between the estimated velocity of the cluster center of the current frame and the estimated velocity of the target in the historical trajectory, reflecting the velocity consistency of the two. The smaller the difference, the more coherent the motion state of the current candidate target and the historical trajectory target, and the more likely they are the same target.

[0075] The echo amplitude difference specifically refers to the difference between the radar echo amplitude of the current frame's cluster center and the radar echo amplitude of historical trajectory targets. It reflects the consistency of their scattering characteristics; the echo amplitude of the same target is usually relatively stable. The smaller the value, the closer the scattering characteristics of the current candidate target are to those of historical trajectory targets, and the stronger the correlation.

[0076] , , These are configurable weight coefficients. A one-to-one association is achieved by applying the Hungarian algorithm to the cost matrix for matching.

[0077] To prevent mismatches, a gating strategy is introduced, allowing only matches with a cost less than a set threshold to occur. Unmatched observations trigger the creation of new trajectory candidates, while trajectories that have not been matched for a long time are terminated.

[0078] The threshold setting is adapted to actual application scenarios, including the spatial dimensions of the elderly's living space (such as room size and layout) and the detection accuracy of millimeter-wave radar (such as ranging error and velocity measurement error), ensuring that the threshold matches the reasonable fluctuation range of target movement and detection data within the scenario. It also considers the constituent elements of the cost matrix.

[0079] This threshold is set by regulatory agencies, service standards, or based on historical trajectory matching data (such as statistical analysis of a large number of normal matching scenarios). The maximum value setting essentially ensures that only matching pairs with strong correlations in position, velocity, and scattering characteristics are retained by limiting the maximum acceptable matching cost.

[0080] Track confirmation and deletion strategy example: number of consecutive confirmation frames (For example, after 5 frames) the candidate trajectory is confirmed as a valid trajectory; continuous missing trajectories Delete the trajectory after (e.g., 5 frames).

[0081] State estimation and improved Kalman filtering: For each confirmed trajectory, an extended / adaptive Kalman filter is used to estimate the target state vector. State transitions can employ either a constant-velocity model or a model with an acceleration term. To adapt to the non-uniform motions in nursing scenarios, an adaptive process noise covariance model is used. .

[0082] Increase when observations indicate that the target acceleration fluctuation is increasing. To improve the filter's response to sudden motion. Reduce the speed when the target approaches the bed and its velocity is low. Stable estimation. Observation noise covariance. The signal-to-noise ratio is dynamically set based on the echo amplitude. The filtered trajectory is also post-processed and smoothed to reduce measurement noise and compensate for missing frames caused by short-term occlusion.

[0083] The service type marking unit 106 synchronously transmits the old person type to the target distinguishing unit and the comprehensive scoring unit.

[0084] The service type marking unit 106 receives the input information of the old person type through a physical button or an interactive interface, and synchronously transmits the old person type to the target distinguishing unit and the comprehensive scoring unit after obtaining the input information, so that the two units can simultaneously obtain the key pre-parameters of the old person type.

[0085] The service type marking unit 106 provides a key judgment basis for the target distinguishing unit and the comprehensive scoring unit. The target distinguishing unit needs to determine the trajectory distinguishing logic according to the old person type (for example, the disabled old person focuses on the reference position to determine the trajectory, and the self-care old person focuses on the motion feature template matching trajectory). The comprehensive scoring unit needs to adjust the weights of the distance index and the heat map index according to the old person type. If the old person type information transmitted by the unit is lacking, the two units will not be able to realize the differentiated operation adapted to the actual nursing needs of the old person, and it is the key to ensure the accuracy of the system evaluation.

[0086] Specifically, the target distinguishing unit executes the following according to the old person type input by the service type marking unit 106: When the old person type is disabled, the trajectory that is long-term in the closed area with the old person reference position as the center and a preset radius and has a speed lower than a speed threshold in the trajectory output by the signal preprocessing and trajectory analysis unit 102 is determined as the old person trajectory. When the old person type is self-care, the similarity of each trajectory to the motion feature template of the caregiver is calculated, and the trajectory with the highest similarity is classified as the caregiver trajectory.

[0087] Specifically, the service type marking unit 106 transmits the old person type to the target distinguishing unit and the comprehensive scoring unit in real time when the service starts. The old person type is received through a physical button or an interactive interface.

[0088] The old person type is transmitted to the target distinguishing unit and the comprehensive scoring unit in real time.

[0089] In a specific embodiment, the target distinguishing unit determines the target distinguishing strategy (caregiver and old person determination): the overall flowchart of the target distinguishing strategy is shown in Figure 7 .

[0090] Before the service starts, the caregiver first selects the old person type as self-care or disabled through the service type marking unit. This information will directly affect the judgment logic of the system in the subsequent target distinguishing and trajectory analysis.

[0091] After the service officially begins, the caregiver first measures the blood pressure of the old person according to the process. The spatial behavior characteristics of this process play an important role in distinguishing the target: during the time period of measuring blood pressure, the caregiver and the old person usually coincide in the same position of the old person's bedside or the position where the old person sits or lies.

[0092] The point cloud trajectory collected by the millimeter wave MIMO radar module 101 in this stage will present the state of a single target or two targets in similar coordinates. The two targets are positioned synchronously using this time window, and the initial reference position of the person being cared for is recorded in combination with the calibration time point .

[0093] For the disabled old people, they usually have little activity throughout the service process, or even remain in a bedridden state for a long time. At this time, the point cloud data corresponding to the radar echo is sparse and the motion characteristics are not obvious, which can easily cause the automatic tracking algorithm to miss detection or break the track.

[0094] For this situation, the present application uses the reference position in the blood pressure measurement stage as the fixed position of the old person throughout the process, and defines the old person's area in combination with the small radius bed area (0.5m-1.0m, configurable). As long as the target trajectory is long-term within this area and the speed is lower than the set speed threshold , it is determined as the old person's trajectory.

[0095] If there is a temporary lack of trajectory in this area, the trajectory is completed by interpolating the positions of the previous and subsequent frames to ensure the stability of the subsequent analysis.

[0096] For self-care old people, the old people may freely move throughout the service process, and the trajectory may intersect or even coincide with the caregiver.

[0097] Therefore, the present application introduces a caregiver motion feature extraction step after the blood pressure measurement is completed: After the measurement is completed, the caregiver usually has the action of putting the blood pressure instrument back in place. After the system detects this high-speed event, the speed curve, acceleration curve, and displacement pattern of the corresponding trajectory are recorded as the kinematic characteristic template of the caregiver .

[0098] In the subsequent service process, the system calculates the cosine similarity of each trajectory segment with , and the trajectory with high similarity is classified as the caregiver's trajectory, and the remaining trajectories are classified as the old person's trajectory.

[0099] In the case of detecting more than two targets at the same time, the system adopts multi-dimensional feature comprehensive judgment, including: long-term average speed, trajectory stability, average distance from the bed, and initial calibration information. For boundary cases, such as two people moving in the room at the same time and close to each other, the system will call the decision tree classification model based on historical data to input the above feature combination for final judgment, ensuring the accuracy of the separation of the trajectories of the elderly and caregivers.

[0100] The object determined as the trajectory of the elderly will be used as a fixed reference object in subsequent service quality evaluation, and its proximity to the caregiver's trajectory, stay duration, and communication frequency will be used as the basis for calculating core evaluation indicators. The remaining trajectories are marked as the activity trajectories of caregivers or other visitors and do not participate in the elderly reference calculation.

[0101] The purpose of blood pressure measurement is limited. Blood pressure measurement here is a standard service action of the system target differentiation unit. This standard service action can be used as a pre-defined calibration action. At the beginning of the home nursing service, the caregiver performs blood pressure measurement on the elderly to help the system record the initial position of the elderly in the living place (i.e., the elderly reference position), providing a spatial reference for subsequent differentiation of the activity trajectories of caregivers and the elderly, and is not used for any other scenario (such as position calibration outside the service process, trajectory differentiation of non-system, etc.).

[0102] The blood pressure measurement performed in this scenario only obtains behavior process data related to the calibration of the elderly reference position (such as the spatial coordinates of the measurement action), and the physiological data generated by the blood pressure value does not participate in any medical judgment, does not have the diagnostic efficacy of hypertension, hypotension, and other diseases, and is not used as a basis for evaluating the health status of the elderly, but only serves the trajectory differentiation function of the system.

[0103] After the caregiver performs the blood pressure measurement behavior, the caregiver needs to put the blood pressure meter into the storage box, and at this time, the system takes the end action of the blood pressure measurement standard service action as the basis for extracting the caregiver's motion features. The blood pressure meter storage action feature is extracted to generate the caregiver motion feature template.

[0104] Blood pressure measurement can be used as a pre-defined calibration action, and it can also complete the extraction of the caregiver motion feature template. At the same time, the pre-defined calibration action and the extraction of the caregiver motion feature template can use other different standard service actions, such as temperature measurement, blood oxygen measurement, heart rate measurement, etc.

[0105] Encryption cloud uploading and saving rules of motion trajectory. The indoor activity trajectory data (i.e., motion trajectory data) of the caregiver and the old person generated by the millimeter wave MIMO radar module 101 after being collected, processed by the signal preprocessing and trajectory analysis unit 102, needs to be encrypted before being uploaded to the cloud server to ensure data security and avoid leakage of trajectory information. The saving duration of the encrypted motion trajectory data in the cloud server is strictly limited to 7 days, and if it exceeds 7 days, it will be processed according to the system preset strategy (such as automatic deletion), which not only meets the short-term data usage requirements such as service quality traceability and abnormality detection, but also reduces the privacy risk and data management cost brought by long-term storage.

[0106] Occlusion, trajectory loss and re-identification. In the presence of short-term occlusion or measurement frame loss, the improved Kalman filter provides a predicted position, and when the predicted position is close to the new observation, re-association is performed. If there is a risk of trajectory ID exchange (for example, two people walking close together), the motion direction, speed, RCS feature (radar cross section) of the trajectory history are used for reconfirmation, and the relative relationship constraint of the trajectory is introduced, such as the old person should be near the bed, to avoid misclassification.

[0107] After the service is completed, service index calculation and data uploading are performed. The service index includes two parts, specifically distance-related indicators and heat map-related indicators, and a unified comprehensive score is calculated based on the two types of indicators.

[0108] Specifically, the service quality evaluation unit performs: receiving the caregiver trajectory and the old person trajectory from the target distinguishing unit; calculating distance indicators: near-distance accompanying time proportion: the proportion of the cumulative duration of the caregiver trajectory and the old person trajectory distance less than the first distance threshold to the total service duration.

[0109] The first distance threshold is a configurable threshold, which is the core distance determination standard for calculating the near-distance accompanying time proportion and the effective approaching times, and is the critical value for distinguishing the near-distance accompanying of the caregiver and the old person and the non-near-distance accompanying.

[0110] Only when the distance between the caregiver trajectory and the old person trajectory is less than the threshold, the near-distance accompanying duration is counted; only when the distance between the two changes from greater than the threshold to less than or equal to the threshold, it is counted as 1 effective approaching.

[0111] Effective approaching times: the number of times the distance between the caregiver trajectory and the old person trajectory changes from greater than the first distance threshold to less than or equal to the first distance threshold.

[0112] calculating heat map indicators: spatial coverage: the proportion of the grid in which the caregiver trajectory stays for more than the stay threshold.

[0113] The stay threshold is a configurable threshold used as a time duration criterion for calculating the spatial coverage. It is a critical value for distinguishing between a valid stay and a brief passing of the caregiver in the ground grid. Only when the stay duration of the caregiver trajectory in a certain ground grid exceeds the threshold, the grid is determined as a visited grid, which is counted in the spatial coverage statistics.

[0114] The long-time stay penalty ratio is the cumulative time duration ratio of the caregiver trajectory exceeding the second distance threshold and staying for more than the time duration threshold from the reference location of the elderly.

[0115] The second distance threshold is a configurable threshold used as a distance criterion for calculating the long-time stay penalty ratio. It is a critical value for distinguishing between staying within a reasonable range and staying far away from the elderly. Only when the distance between the caregiver trajectory and the reference location of the elderly exceeds the threshold, and the stay duration exceeds the time duration threshold, is the long-time stay penalty duration counted.

[0116] The time duration threshold is a configurable threshold used as a time duration criterion for calculating the long-time stay penalty ratio. It is a critical value for distinguishing between a brief absence and a long absence of the caregiver. Only when the distance between the caregiver trajectory and the reference location of the elderly exceeds the second distance threshold, and the stay duration exceeds the threshold, is the long-time stay penalty duration counted, triggering the penalty logic.

[0117] In particular, it further comprises an anomaly detection unit for: Receiving the service quality score output by the comprehensive score unit.

[0118] Receiving the long-time stay penalty ratio output by the service quality assessment unit.

[0119] Generating an alarm when the long-time stay penalty ratio exceeds a preset alert value or the service quality score is lower than an anomaly threshold.

[0120] The preset alert value is a configurable threshold used as a criterion for determining whether the long-time stay penalty ratio is abnormal. It is a critical value for triggering the long-time stay alarm. When the long-time stay penalty ratio output by the service quality assessment unit exceeds the threshold, the anomaly detection unit determines that the caregiver has a risk of long-time absence from the elderly, and generates an alarm.

[0121] The anomaly threshold is a configurable threshold used as a criterion for determining whether the service quality score meets the standard. It is a critical value for triggering the service quality substandard alarm. When the service quality score output by the comprehensive score unit is lower than the threshold, the anomaly detection unit determines that the service quality is unqualified, and generates an alarm.

[0122] Distance-related indicators.

[0123] Close-contact time ratio (CCTR): defined as: where is the time point is the Euclidean distance between the caregiver and the elderly reference position, is an indicator function, which equals 1 if the condition is satisfied, otherwise equals 0, is the sampling interval, is the total service duration, is the close distance threshold (default setting is = 1.0 m, configurable).

[0124] Effective Interaction Count (EIC): defined as the number of times when the distance between the caregiver and the elderly changes from > D1 to ≤ D1, used to measure the interaction frequency.

[0125] Average Close Duration (ACD): the duration of all consecutive close events is calculated and averaged, i.e., where is the duration of the m-th close event, and M is the total number of close events.

[0126] Proximity-weighted Engagement Score (PWES): defined as: used to measure the depth of proximity during close time.

[0127] The above four indicators can be used as important basis for determining whether the caregiver is in close proximity to the elderly for multiple times and sufficiently.

[0128] Calculation of service quality indicators.

[0129] Heatmap-related indicators: to depict the spatial activity distribution of the caregiver, the floor of the room is discretized into grid cells (cells), with grid size (default = 0.5 m x 0.5 m, configurable), the cumulative residence time or the number of occurrences of each cell is counted and normalized to obtain the activity heat map . A typical example of a heat map is shown in Figure 8 .

[0130] Based on the following indicators are defined: Coverage Ratio (CR): ; where ​the number of cells visited (residence time exceeding a threshold in the heat map, the total number of valid cells in the room, CR reflects whether the caregiver touches most areas of the room.

[0131] Top-K occupancy (TKP): calculate the proportion of the total residence time of the top-K cells in the heat map, if the proportion is too high, it means that the activity concentration is too strong, and there may be a long time to stay away from the old man.

[0132] Long-stay penalty (LSP): the sum of all event times of the caregiver staying away from the bed more than (e.g. =2m) and staying more than (e.g. =300s) is calculated as a proportion of the total service time, which is a penalty term for abnormal away behavior.

[0133] Coverage uniformity (CU): CU can be represented by the standard deviation of the coefficient of variation of the heat map cell residence time, and a lower coefficient of variation means more uniform coverage. The above heat map indicators combined with distance-related indicators can more completely reflect the spatial participation of the caregiver in performing tasks such as wiping, cleaning, and organizing in the room and whether there is a long time away from the post.

[0134] Specifically, the comprehensive score unit executes: receive the type of the old person from the service type designation unit 106; receive the distance index and the heat map index from the service quality assessment unit; when the type of the old person is disabled, assign a first weight value to the near-distance accompanying time proportion; when the type of the old person is self-care, assign a second weight value to the spatial coverage rate.

[0135] The subject of determining the weight value is issued by the regulatory agency based on historical data, and the service institution sets the nursing standard.

[0136] Adjustment rules for weight values. The weight of near-distance accompanying in nursing of disabled old people is higher than that of self-care old people.

[0137] For ease of presentation and comparison, each indicator needs to be normalized to a unified scale (0-1 range). The comprehensive service quality score can be defined as a weighted linear combination: ; where is the weight coefficient,​ The weights can be adjusted according to the service type (disabled or self-care), for example, increasing the weights of CCTR and ACD for disabled old people, and increasing the weights of CR and TKP for self-care old people. The weights and thresholds can be determined by regulatory agencies, service standards, or based on historical data in the cloud model and issued to the device.

[0138] When the service is completed, the caregiver ends the service through the end service button on the single machine device. At this time, the device will realize data packaging, uploading and privacy protection strategy: after the nursing staff presses the end button, the device completes all index calculation and heat map generation locally, and uploads the matrix containing: service summary (start / end time, service time, comprehensive score , activity heat map), key event timeline and brief description, and encrypted raw trajectory summary to the cloud server.

[0139] Raw point cloud and complete audio (if turned on) can be stored in encrypted form in the SD card locally and selectively uploaded according to the strategy; if the network is unavailable, the uploading task is retained and automatically synchronized when the network is restored.

[0140] GPS position is only uploaded when enabled and necessary, used to verify the location information of the caregiver's arrival and departure to the community.

[0141] A typical use process of a home-based elderly care service quality evaluation and monitoring system based on millimeter wave radar technology based on the present application is as follows: nursing staff A places the device on the bed in the home at 09:00 and long-presses "start", selects "disabled old people" in the system interface and completes the initial calibration by measuring the blood pressure of the old people.

[0142] The device records . In the next 40 minutes, the radar continuously collects point clouds and continuously performs clustering and trajectory tracking on the device side, identifying two trajectories: one is the static trajectory of the old people, and the other is the moving trajectory of the nursing staff. The system counts that the cumulative time of the nursing staff and the old people within a distance of ≤1.0m is 18 minutes, the number of near-distance switching is 6 times, and the average near-distance duration is 3 minutes; the heat map shows that the bed coverage rate is 85% and the overall room coverage rate is 60%, and there is a 12-minute long stay (triggering LSP alert) in the area near the door. After the service is completed, the nursing staff A presses the end service button on the single machine device, the device generates a report and safely uploads it, and the cloud side marks this service as "normal but needs attention to the door stay" according to the reference model set by the agency.

[0143] In a specific embodiment, nursing staff A needs to provide 1 time of 40 minutes of home-based elderly care service for disabled old people B living in a certain community, and the specific implementation process is as follows: First step, mmWave MIMO radar module 101 deployment and start: after the caregiver A carries the system equipment on the door, the mmWave MIMO radar module 101 is placed at the bedside of the old man B in the bedroom (covering the old man's bed and the caregiver's active area, ensuring that the radar horizontal detection angle is 120 degrees and the farthest detection distance is 7 meters). After starting the equipment, the mmWave MIMO radar module 101 starts collecting indoor three-dimensional position data and outputs real-time indoor three-dimensional trajectory data to the signal preprocessing and trajectory analysis unit 102.

[0144] Second step, service type calibration: the caregiver A calibrates the service type through the physical button operation of the equipment, selects the old man type as disabled, and this unit synchronously transmits the disabled old man type information to the target distinguishing unit and the comprehensive score unit.

[0145] Third step, signal preprocessing and trajectory analysis and pre-defined calibration action execution: the signal preprocessing and trajectory analysis unit 102 receives the original three-dimensional trajectory data output by the mmWave radar module, first performs preprocessing (eliminates environmental static echo interference points and low amplitude low SNR points), then clusters through the density clustering algorithm and realizes multi-target trajectory tracking through the improved Kalman filtering algorithm.

[0146] At the same time, the caregiver A performs the pre-defined calibration action (blood pressure measurement) as prompted by the system, measures the blood pressure of the old man B at the bedside, and the signal preprocessing and trajectory analysis unit 102 block synchronously transmits the trajectory data of this process to the target distinguishing unit.

[0147] Fourth step, target distinguishing: the target distinguishing unit receives the trajectory data during blood pressure measurement transmitted by the signal preprocessing and trajectory analysis unit 102, records the position of the old man B at the bedside as the old man reference position (coordinates set as X0, Y0, Z0). After the blood pressure measurement is completed, the caregiver A puts the blood pressure instrument back on the living room storage rack, and the target distinguishing unit extracts the speed curve and displacement mode of the caregiver A during this process to generate a caregiver motion feature template.

[0148] In subsequent services, the target distinguishing unit determines the trajectory that is always within a closed area with a radius of 0.8 m centered on the old man reference position and has a speed lower than 0.2 m / s as the trajectory of the old man B according to the disabled old man type determination rule.

[0149] The trajectory with a similarity of 92% to the caregiver motion feature template is determined as the trajectory of the caregiver A, and the two types of trajectories after distinguishing are output to the service quality evaluation unit.

[0150] Fifth step, service quality index calculation: the service quality evaluation unit receives the trajectory data after distinguishing and calculates two types of indexes: Distance indicator: Set the first distance threshold to 1.0 m, and count the cumulative duration of the trajectory distance between the caregiver A and the old man B within 40 minutes of service ≤1.0 m as 18 minutes, and the "near distance accompanying time ratio" = 18 / 40 = 45%; the number of times the trajectory distance between the caregiver A and the old man B changes from >1.0 m to ≤1.0 m is 6 times, that is, the effective approaching times = 6 times.

[0151] Heat map indicator: Disperse the floor of the old man's bedroom into 0.5m x 0.5m grid units, set the stay threshold to 30 seconds, the second distance threshold to 2.0m, and the duration threshold to 300 seconds; the number of grid units visited (staying for more than 30 seconds) is 28, the total number of effective grid units in the room is 40, and the space coverage rate = 28 / 40 = 70%.

[0152] Caregiver A only stays in the door area 2.5m away from the old man's reference location for 5 minutes (300 seconds) during the 25th-30th minute of service, and the long-time stay penalty ratio = 5 / 40 = 12.5%.

[0153] Sixth, comprehensive service quality score: The comprehensive scoring unit receives the type of disabled old people transmitted by the service type labeling unit, and assigns weights according to the preset weight distribution rule (the disabled old people focus on near-distance accompanying, and the near-distance accompanying time ratio is assigned a weight of 0.4, the effective approaching times are assigned a weight of 0.2, the space coverage rate is assigned a weight of 0.2, and the long-time stay penalty ratio is assigned a weight of 0.2, and the long-time stay penalty ratio is calculated according to 1 normalization value), and the normalized indicators (near-distance accompanying time ratio normalized value 0.45, effective approaching times normalized value 0.6, space coverage rate normalized value 0.7, long-time stay penalty ratio normalized value 0.125) are weighted and summed to get: Comprehensive service quality score = 0.4 x 0.45 + 0.2 x 0.6 + 0.2 x 0.7 + 0.2 x (1-0.125) = 0.18 + 0.12 + 0.14 + 0.175 = 0.615 (full score 1.0).

[0154] Seventh, data storage and privacy protection: After the service is completed, the system will upload the differentiated motion trajectory data to the cloud server after encryption, and save it in the cloud for 7 days according to the encryption rule. The local storage module synchronously saves the complete trajectory data and analysis results. Blood pressure measurement in this service is only used to determine the old man's reference location, and the blood pressure values generated are not used for any disease diagnosis and are not included in the health assessment category.

[0155] In summary, by integrating the millimeter wave MIMO radar and edge trajectory analysis capability in the mobile device, combined with service type calibration, initial position calibration, robust point cloud clustering and improved Kalman filter tracking algorithm, and a series of quantitative indicators based on distance and heat map, the full process of home nursing service is realized. The non-contact, quantifiable and traceable monitoring and evaluation. This embodiment can protect the privacy of the elderly and effectively prevent fraud based on GPS or manual clocking, providing a reliable quality assessment means for regulatory agencies and service providers.

[0156] Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can modify the technical solutions described in the foregoing embodiments or make equivalent replacements for part of the technical features, and any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A home-based elderly care service quality assessment and monitoring system based on millimeter-wave radar technology, characterized in that, Including: The target differentiation unit determines the elderly person's reference position based on predefined calibration actions and uses a caregiver motion feature template to distinguish the caregiver's trajectory from the elderly person's trajectory. The service quality assessment unit generates distance indicators and heat map indicators based on the differentiated trajectories; the distance indicators reflect the spatial proximity between the caregiver and the elderly; the heat map indicators reflect the spatial activity distribution of the caregiver. The comprehensive scoring unit outputs a service quality score by weighting and summing distance and heat map indicators according to the type of elderly person.

2. The in-home elderly care service quality assessment and monitoring system based on millimeter-wave radar technology according to claim 1, characterized in that, It also includes: The millimeter-wave MIMO radar module (101) is installed in the elderly's living area and outputs indoor three-dimensional trajectory data. The signal preprocessing and trajectory analysis unit (102) receives the trajectory data from the millimeter-wave MIMO radar module (101), performs preprocessing and multi-target trajectory tracking, and outputs the data to the target differentiation unit. The service type labeling unit (106) simultaneously transmits the elderly type input to the target differentiation unit and the comprehensive scoring unit.

3. The in-home elderly care service quality assessment and monitoring system based on millimeter-wave radar technology according to claim 2, characterized in that, The signal preprocessing and trajectory analysis unit (102) performs the following steps before outputting the trajectory to the target differentiation unit: Clustering of 3D trajectory data based on density clustering algorithm; An improved Kalman filter algorithm is applied to achieve multi-target trajectory tracking.

4. The in-home elderly care service quality assessment and monitoring system based on millimeter-wave radar technology according to claim 2, characterized in that, The predefined calibration actions are the standard service actions for initially serving the elderly; The system receives trajectory data after the completion of standard service actions and extracts kinematic features of the process to generate a motion feature template for the caregiver.

5. The in-home elderly care service quality assessment and monitoring system based on millimeter-wave radar technology according to claim 1, characterized in that, The target differentiation unit performs the following based on the type of elderly person: When the elderly person is classified as disabled, the trajectory data that is consistently within a closed area centered on the elderly person's reference position and with a preset radius, and whose speed is below a speed threshold, will be identified as the elderly person's trajectory. When the elderly person is classified as self-reliant, the similarity between each trajectory and the caregiver's motion feature template is calculated, and the trajectory with the highest similarity is classified as the caregiver's trajectory.

6. The in-home elderly care service quality assessment and monitoring system based on millimeter-wave radar technology according to claim 1, characterized in that, The service quality assessment unit performs the following calculations: distance index and heat map index. The distance metrics include the percentage of time spent in close proximity and the number of effective approaches; The metrics used to calculate the heatmap include spatial coverage and long-stay penalty ratio.

7. The in-home elderly care service quality assessment and monitoring system based on millimeter-wave radar technology according to claim 1, characterized in that, The comprehensive scoring unit performs the following: When the elderly person is classified as disabled, the first weight value is assigned to the proportion of close care time. When the elderly person is classified as self-reliant, a second weight value is assigned to the space coverage rate.

8. The in-home elderly care service quality assessment and monitoring system based on millimeter-wave radar technology according to claim 1, characterized in that, It also includes an event recognition module, used for: Identify standard service actions.

9. The in-home elderly care service quality assessment and monitoring system based on millimeter-wave radar technology according to claim 1, characterized in that, It also includes an anomaly detection unit, used for: Receive the service quality score output by the comprehensive scoring unit; Receive the long dwell time penalty ratio output by the service quality assessment unit; An alarm is generated when the long-stay penalty ratio exceeds the preset warning value or the service quality score falls below the abnormal threshold.

10. The in-home elderly care service quality assessment and monitoring system based on millimeter-wave radar technology according to claim 2, characterized in that, The service type identification unit (106) at service startup: Receive elderly person type input via physical buttons or interactive interface; The elderly person's type is transmitted in real time to the target differentiation unit and the comprehensive scoring unit.

Images