Robot noninvasive health physical examination and monitoring platform

Abstract

The invention relates to the technical field of robot health physical examination, and discloses a robot noninvasive health physical examination and monitoring platform. A health data acquisition module of the system integrates physiological parameters acquired by a multi-modal sensor to generate a standardized health data set; the physical sign load analysis module calculates parameter fluctuation difference to generate a physical sign load distribution state value; the physical examination parameter optimization module screens optimal physical examination parameters to generate an operation parameter set; the environment adaptability adjusting module generates an environment response physical examination scheme in combination with the environment and body position data; a health threshold value dynamic setting module adjusts an abnormal judgment threshold value to generate a grading early warning threshold value; the health trend prediction module generates a health risk prediction value through fuzzy reasoning; and the physical examination feedback control module adjusts operation parameters and scanning frequency to generate a self-adaptive regulation and control strategy. According to the platform, the accuracy and the intelligent level of health physical examination are improved, and the non-invasive health monitoring requirement of the user can be better met.

Description

Technical Field

[0001] This invention relates to the field of robotic health check technology, specifically to a robotic non-invasive health check and monitoring platform. Background Technology

[0002] In the field of health checkups and monitoring, traditional non-invasive testing methods often rely on single sensors to collect physiological data, acquiring only isolated parameters such as body temperature and heart rate. This lack of systematic data integration and outlier removal results in low standardization of the generated health datasets, making it difficult to meet the needs of subsequent precise analysis. Furthermore, existing technologies, when analyzing vital sign load, often only assess changes in a single physiological parameter, failing to effectively calculate the fluctuation differences between different physiological parameters and accurately identify abnormal fluctuation areas. This makes the generation of vital sign load distribution values ​​lack scientific rigor and comprehensiveness.

[0003] In optimizing the operational parameters of robotic physical examinations, traditional methods fail to comprehensively consider the distribution of vital signs and their effects, relying solely on experience to set the angle, duration, and contact force of the examination movements. This lack of experience prevents the selection of the optimal examination path, potentially leading to unreasonable operational parameters during the examination and affecting the accuracy of the results and the user experience. Furthermore, the impact of environmental factors on examination parameters is often overlooked. Current technologies lack the ability to collect and analyze data on environmental temperature, humidity, and user posture, making it impossible to dynamically adjust the examination plan according to environmental changes. This makes the examination results susceptible to external environmental interference, further reducing the reliability of the detection.

[0004] In the health threshold setting and risk prediction stages, existing technologies mostly use fixed thresholds for anomaly detection, failing to dynamically adjust them by combining real-time physiological parameter change rates with historical health data. This results in a lack of flexibility and adaptability in threshold setting, making it difficult to accurately reflect the differences in health status among different users. Furthermore, in health trend prediction, traditional methods fail to effectively integrate multi-parameter change trends for joint analysis, relying mostly on single-parameter or simple linear analysis methods. This fails to comprehensively capture health risk signals, leading to insufficient accuracy in health risk prediction values. In addition, existing physical examination systems lack effective feedback control mechanisms, unable to adjust examination operation parameters and scanning frequency based on the error between health risk prediction values ​​and real-time data. This hinders adaptive regulation and further limits the intelligence and precision of the physical examination system. Summary of the Invention

[0005] The purpose of this invention is to provide a robotic non-invasive health check and monitoring platform to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides a robotic non-invasive health check and monitoring platform, the platform comprising: The health data acquisition module acquires user physiological parameters based on multimodal sensors, integrates body temperature, heart rate, and blood oxygen data, removes outliers, aligns time series data, and generates a standardized health dataset. The vital signs load analysis module calculates the fluctuation differences of different physiological parameters based on the standardized health dataset, marks abnormal fluctuation areas, and generates vital signs load distribution state values. Based on the vital sign load distribution state value, the physical examination parameter optimization module extracts the combination of angle and duration of the robot's physical examination action, filters the optimal physical examination path and contact force parameters, and generates a set of physical examination operation parameters. The environmental adaptability adjustment module collects environmental temperature and humidity data and user posture data based on the physical examination operation parameter set, analyzes the degree of influence of environmental factors on the physical examination parameters, and generates an environmental response physical examination plan. The health threshold dynamic setting module extracts the real-time physiological parameter change rate based on the environmental response physical examination plan, and dynamically adjusts the abnormal judgment threshold by combining historical health data to generate a graded health early warning threshold. Based on the graded health early warning threshold, the health trend prediction module uses fuzzy inference to jointly analyze the changing trends of multiple parameters and generate a health risk prediction value. The physical examination feedback control module compares the error between real-time data and predicted values ​​based on the predicted health risk values, adjusts the physical examination operation parameters and optimizes the scanning frequency, and generates an adaptive physical examination control strategy.

[0007] Preferably, the vital sign load distribution status values ​​include a set of physiological parameter fluctuations, a set of regional load levels, and a set of abnormal markers; The physical examination operation parameter set includes path planning parameters and force control parameters; The environmental response physical examination plan includes temperature and humidity adaptation parameters and body position compensation parameters; The graded health early warning threshold includes real-time change rate parameters and historical benchmark parameters; The predicted health risk values ​​include trend joint analysis parameters and pattern classification parameters; The adaptive physical examination and control strategy includes error compensation parameters and frequency optimization parameters.

[0008] Preferably, the health data acquisition module includes: The multi-source data synchronization submodule aligns the physiological parameter acquisition timestamps based on the multimodal sensor timing, removes time-off data, and generates a time-aligned original dataset. The anomaly filtering submodule identifies and replaces outliers based on a preset reasonable range of physiological parameters, and uses adjacent data to impute missing values ​​to generate a cleaned health data sequence. The standardization processing submodule normalizes the cleaned health data sequence, unifies the units and sampling frequency, and generates a standardized health dataset.

[0009] Preferably, the physical examination parameter optimization module includes: The motion parameter extraction submodule analyzes the historical data of robot joint angles and execution time based on the physical sign load distribution state value, stores them according to load level, and generates a physical examination motion parameter library. The path optimization submodule filters physical examination paths that cover abnormal fluctuation areas from the physical examination action parameter library, calculates the path efficiency score, and generates a candidate physical examination path set. The force adaptation submodule adjusts the contact force parameters according to the user's body surface characteristics, and combines the pressure gradient with the load level to generate a set of physical examination operation parameters.

[0010] Preferably, the environmental adaptability adjustment module includes: The environmental factor acquisition submodule acquires real-time data on ambient temperature and humidity and user body position deviation, performs noise filtering and moving average processing, and generates environmental feature vectors. The parameter impact analysis submodule calculates the correlation coefficient between the environmental feature vector and the physical examination operation parameter set, marks the parameter items with significant impact, and generates a parameter sensitivity list; The dynamic matching submodule adjusts the examination path and intensity parameters based on the parameter sensitivity list, iteratively verifies environmental adaptability, and generates an environmental response examination plan.

[0011] Preferably, the health threshold dynamic setting module includes: The rate of change extraction submodule captures the first derivative of real-time physiological parameters, calculates the extreme values ​​of change within the sliding window, and generates a dynamic rate of change sequence. The threshold stratification submodule classifies abnormal levels based on the percentiles of historical health data distribution, and adjusts the threshold boundaries by combining the dynamic rate of change sequence to generate tiered health early warning thresholds.

[0012] Preferably, the health trend prediction module includes: The multi-parameter fusion submodule assigns dynamic weights to heart rate, blood oxygen, and body temperature data, calculates the joint trend index, and generates a multimodal health feature set. The pattern inference submodule matches the current health feature set with the historical abnormal pattern library, corrects trend prediction bias, and generates health risk prediction values.

[0013] Preferably, the pattern inference submodule performs the following operations: Extract the top K patterns with the highest similarity to the current health feature set from the historical abnormal pattern library; Calculate the deviation between the current feature and each pattern, and output the risk probability after weighted fusion; The initial trend prediction results are adjusted based on the risk probability to generate a health risk prediction value.

[0014] Preferably, the physical examination feedback control module includes: The error calculation submodule compares the absolute deviation between real-time physiological parameters and health risk prediction values, classifies error levels by region, and generates an error distribution map. The parameter iteration submodule adjusts the coverage of the physical examination path based on the error distribution map, optimizes the duration of a single physical examination and the frequency of repetition, and generates an adaptive physical examination control strategy.

[0015] Preferably, the parameter iteration submodule performs the following operations: Increase the density of examination path nodes in high error areas, dynamically adjust the movement speed of the examination robot arm according to the error level, record the error change rate after parameter adjustment, and iterate until the convergence threshold is reached.

[0016] Compared with the prior art, the beneficial effects of the present invention are: The vital signs load analysis module calculates the fluctuation differences of different physiological parameters based on a standardized health dataset, marks abnormal fluctuation areas, and generates vital signs load distribution state values. This breaks through the limitations of traditional technologies that only analyze a single parameter. It can comprehensively grasp the correlation changes between various physiological parameters, more accurately identify abnormalities in vital signs load, and provide a precise basis for subsequent optimization of physical examination parameters. This helps to develop a physical examination plan that is more in line with the user's actual vital signs condition.

[0017] The physical examination parameter optimization module extracts the angle and duration combination of the robot's physical examination actions based on the distribution state value of vital signs, filters the optimal physical examination path and contact force parameters, and generates a set of physical examination operation parameters. This changes the traditional method of setting parameters based on experience, making the setting of physical examination operation parameters more scientific and targeted. It can reduce the deviation of physical examination results caused by unreasonable operation parameters, while improving the user's comfort during the physical examination process and avoiding discomfort caused to the user due to improper contact force or unreasonable path.

[0018] The environmental adaptability adjustment module collects environmental temperature and humidity data and user posture data based on the physical examination operation parameter set, analyzes the degree of influence of environmental factors on the physical examination parameters, and generates an environmental response physical examination plan. This makes up for the shortcomings of traditional technology that ignores the influence of the environment. It can dynamically adjust the physical examination plan according to different environmental conditions and user posture, reduce the interference of environmental factors on the physical examination results, and ensure that accurate and reliable physical examination data can be obtained in different environments, thereby improving the applicability of the platform in different scenarios.

[0019] The dynamic health threshold setting module extracts the rate of change of real-time physiological parameters based on the environmental response physical examination plan, and dynamically adjusts the abnormal judgment threshold and generates graded health warning thresholds by combining historical health data. It abandons the traditional fixed threshold setting method, fully considers the difference between the user's real-time physiological changes and historical health status, and makes the abnormal judgment threshold more in line with the individual user's situation. It can more accurately identify health abnormal signals, reduce misjudgment or missed judgment, and provide more targeted warning standards for users with different health conditions.

[0020] The health trend prediction module uses fuzzy reasoning based on the graded health early warning threshold to jointly analyze the changing trends of multiple parameters and generate health risk prediction values. This overcomes the shortcomings of traditional prediction methods that rely on a single parameter or simple linear analysis. It can comprehensively capture health risk information by integrating the changing trends of multiple parameters, more accurately predict users' future health risks, help users know potential health problems in advance, and facilitate timely intervention measures.

[0021] The physical examination feedback control module adjusts the examination operation parameters and optimizes the scanning frequency based on the error between the real-time data and the predicted health risk values, generating an adaptive examination control strategy. This forms a complete closed-loop control mechanism that can continuously optimize the examination process according to the actual test results, ensuring that the examination operation is always in the optimal state. This further improves the accuracy of the examination results and the intelligence level of the platform. At the same time, by optimizing the scanning frequency, unnecessary testing operations can be reduced while ensuring the testing effect, thus improving the platform's operating efficiency. Attached Figure Description

[0022] Figure 1 This is a schematic diagram illustrating the working principle of the robotic non-invasive health check and monitoring platform described in this invention. Figure 2 A schematic diagram illustrating the working principle of the distribution state values ​​of vital signs and load. Figure 3 This is a schematic diagram illustrating the working principle of the health data acquisition module. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] Please see Figure 1This invention provides a robotic non-invasive health check and monitoring platform. The system includes: a multi-module collaborative system for automated health monitoring and dynamic adjustment; a health data acquisition module that uses multimodal sensors to simultaneously collect user body temperature, heart rate, and blood oxygen data, and performs timestamp alignment and outlier cleaning on the raw data to generate a standardized health dataset; a vital sign load analysis module that calculates the fluctuation differences of various physiological parameters based on the dataset, identifies abnormal fluctuation areas, and generates vital sign load distribution state values; a check-up parameter optimization module that extracts historical robot action parameters based on the load state values, optimizes the check-up path and contact intensity, and generates a set of check-up operation parameters; an environmental adaptability adjustment module that combines environmental temperature and humidity data with user position data to analyze the impact of environmental factors on check-up parameters and generate an environmental response check-up plan; a health threshold dynamic setting module that dynamically adjusts the abnormal judgment boundary based on real-time physiological parameter change rate and historical data to generate graded health warning thresholds; a health trend prediction module that uses fuzzy inference to fuse multi-parameter change trends and outputs a health risk prediction value; and a check-up feedback control module that dynamically adjusts the check-up operation parameters and scanning frequency by comparing the error between real-time data and predicted values, ultimately generating an adaptive check-up control strategy.

[0025] Example 1: See Figure 2 In the process of generating the vital sign load distribution status values ​​in the vital sign load analysis module, the construction of the physiological parameter fluctuation set relies on the deep calculation of the standardized health dataset. The heart rate coefficient of variation is obtained by analyzing the standard deviation of the continuous RR interval, the blood oxygen saturation dispersion is calculated by calculating the variance of the sampled data per minute, and the body temperature gradient change is extracted from the absolute value of the temperature difference between adjacent sampling points in the time series. The regional load level set divides the calculated fluctuation amount into three intervals according to the preset fluctuation amplitude threshold. The low load interval corresponds to the situation where all parameter fluctuations are below the lower limit of the normal range, the medium load interval includes the state where a single parameter briefly exceeds the threshold but recovers quickly, and the high load interval marks the period when multiple parameters are continuously abnormal and fluctuate violently. The abnormality mark set uses a sliding time window to scan each parameter sequence. When a parameter exceeds its reasonable upper or lower limit for a duration that reaches the set threshold within the window period, the system records the start timestamp, duration, and maximum value of the exceedance of the abnormality. These marks are associated and mapped with specific load level intervals.

[0026] The generation of the physical examination operation parameter set requires the integration of multi-dimensional information. The path planning parameters first locate the body areas that need to be examined based on the abnormality marker set. The optimal trajectory of the robotic arm to each target point is calculated based on the robot's kinematic model. This trajectory consists of a series of three-dimensional spatial coordinate points and Bézier curves connecting these points. The movement speed vector is dynamically adjusted according to the regional load level. High load areas correspond to lower scanning speeds and higher path point densities. The force control parameters are set by combining pressure sensor feedback with the human tissue elasticity model. The contact pressure threshold is initialized based on the tolerance database of different body parts. During the actual physical examination, it is fine-tuned based on real-time feedback electromyographic signals or micro-expression recognition results. The force adjustment step size is defined as the minimum increment value of a single pressure adjustment, and the magnitude of this value is positively correlated with the load level.

[0027] The temperature and humidity adaptation parameters of the environmental response physical examination solution mainly address the impact of environmental factors on measurement accuracy. The correction coefficients for sensor readings based on ambient temperature and humidity are obtained through laboratory calibration experiments, and a correspondence table between temperature and humidity changes and sensor output value drift is established. The response delay compensation value of the robotic arm actuator is calculated based on the response characteristic curves of the servo motor under different temperatures and humidity conditions. The processing of body position compensation parameters involves the fusion of computer vision and inertial measurement units. The relative position of the user's torso and limbs is captured by a depth camera, and the body position offset angle is calculated by combining IMU data. The physical examination path offset correction matrix converts this angle into the coordinate correction amount of the robotic arm end effector. The update frequency of this matrix is ​​consistent with the speed of change in the user's body position.

[0028] The dynamism of the graded health warning threshold is reflected in the synergistic effect of the real-time rate of change parameter and the historical benchmark parameter. The real-time rate of change parameter is calculated by the first derivative of the physiological parameter within a sliding window. The window width is adaptively adjusted according to the parameter characteristics. For example, a 5-second window is used for heart rate and a 30-second window is used for blood oxygen. The extreme value of change is selected as the absolute maximum value of the derivative within the window. The historical benchmark parameter is extracted from the user's long-term health data. The percentile distribution of each parameter is calculated according to different time periods (such as morning, afternoon, and night). The normal range baseline is taken as the interval between the 5th percentile and the 95th percentile. This interval is updated every 24 hours based on new data.

[0029] The generation of health risk prediction values ​​is a multi-stage reasoning process. The trend joint analysis parameters first perform weighted fusion on standardized data of heart rate, blood oxygen, and body temperature. The weight coefficients are dynamically allocated based on the measurement reliability of each parameter in the current environment. The comprehensive health index is calculated using a linear weighted model but retains a non-linear correction interface. The operation of the pattern classification parameters depends on the completeness of the historical abnormal pattern library, which contains multi-parameter change feature templates under common abnormal conditions. The matching process uses similarity calculation based on dynamic time warping. The risk category identifier is output as a discrete risk level label and corresponding confidence score.

[0030] The error compensation parameters of the adaptive physical examination control strategy mainly handle the deviation between the predicted and measured values. The adjustment of the coverage density of the physical examination path is achieved by increasing or decreasing the number of path points. The path point densification mechanism is automatically triggered in high error areas. The limit value of the robotic arm's motion acceleration is adjusted in steps according to the error level. The frequency optimization parameters are based on the spatial and temporal characteristics of the error distribution to formulate a strategy. In spatially concentrated error areas, the number of physical examinations will be increased. In temporally, the period of continuous increase in error will shorten the physical examination interval. The determination of the number of repetitions needs to comprehensively consider the balance between physical examination efficiency and user comfort.

[0031] Example 2: See Figure 3 The health data acquisition module coordinates the acquisition timing of multimodal sensors through a multi-source data synchronization submodule. This submodule uses a high-precision clock signal to synchronize the time of all sensor nodes and adds nanosecond-level timestamps to the data packets. For time-series offsets caused by network latency or differences in device response, the system applies an interpolation algorithm to generate compensation values ​​between adjacent data points, thus forming a time-aligned original dataset. The anomaly filtering submodule detects outliers based on a preset reasonable range library of physiological parameters. This library contains parameter thresholds under different age, gender, and physical conditions. When a data point is detected to exceed the dynamic threshold range, the system uses linear interpolation of adjacent valid data to replace it. For continuous missing data segments, predictive interpolation based on time series trends is used to imputate the data. Finally, the cleaned health data sequence is output. The standardization processing submodule performs Z-score normalization on the cleaned data. First, it calculates the arithmetic mean and standard deviation of each parameter sequence, and then converts each data point into a multiple of the standard deviation from the mean. At the same time, it uses resampling technology to unify sensor data of different frequencies to a baseline sampling rate, generating a standardized health dataset.

[0032] The physical examination parameter optimization module analyzes the robot's historical operation data through the motion parameter extraction submodule. This submodule extracts the angle change sequence and dwell time data of each joint of the robotic arm from the execution log, and stores them according to the load level output by the vital sign load analysis module. Higher load levels correspond to larger joint angle change rates and longer local dwell times. These parameters are indexed by body region to form a physical examination motion parameter library. The path optimization submodule performs multi-objective optimization search in this parameter library. First, it generates a set of candidate physical examination paths based on the spatial distribution of abnormal fluctuation areas. Each path contains a series of spatial coordinate points of the robotic arm end effector. The path efficiency score comprehensively considers three dimensions: total path length, area coverage, and time cost. A weighted scoring method is used to select the path scheme with the highest comprehensive score. The force adaptation submodule adjusts the contact parameters according to the user's body surface characteristics. It obtains body surface curvature and tissue thickness data through 3D vision sensors and establishes a force mapping model by combining pressure feedback information. Different pressure gradients are assigned to different load level areas. High load areas use a smaller pressure increment step for fine detection. The final output physical examination operation parameter set includes spatial path coordinate sequence, movement speed curve, and pressure control parameters.

[0033] The timestamp alignment operation of the multi-source data synchronization submodule adopts a combination of hardware synchronization and software compensation. A highly stable clock chip is deployed on each sensor node, and the main controller periodically sends time synchronization signals. For nodes that cannot be synchronized by hardware, a software timestamp correction algorithm is used to compensate for time errors by calculating network transmission delay and processing delay. The outlier detection of the anomaly filtering submodule adopts a dynamic threshold mechanism. The threshold range is adaptively adjusted according to the recent data fluctuations of the user. When a suspected outlier is detected, the system will start a multi-sensor verification mechanism to determine whether it is indeed an anomaly by checking the consistency of the associated parameters. The resampling operation of the standardization processing submodule adopts an anti-aliasing digital filter. High-frequency sampled data is downsampled first, and then low-frequency data is interpolated and upsampled to ensure that all data streams reach a uniform sampling frequency without introducing frequency aliasing errors.

[0034] The motion parameter extraction submodule's data parsing process includes kinematic feature extraction, calculating the curvature change of the end effector's motion trajectory from the robotic arm joint angle sequence, and extracting the standard time distribution of different inspection actions from the execution duration data. The path optimization submodule uses a multi-objective genetic algorithm for optimization, setting three fitness functions: path length, region coverage, and time efficiency, and finding the Pareto optimal solution set through iterative evolution. The pressure gradient allocation of the force adaptation submodule adopts a fuzzy control strategy, using body surface characteristic parameters and load levels as input variables, and inferring the applicable pressure parameter range through a fuzzy rule base. The time-aligned raw dataset is immediately processed by the anomaly filtering submodule to avoid processing delays caused by data backlog. The update of the physical examination motion parameter library and the path optimization operation are performed in parallel. When new load level data is generated, the system asynchronously updates the optimized path scheme in the background. The force adaptation submodule communicates with the robot's force control system in real time to ensure that pressure parameters can be accurately converted into the actuator's motor torque output.

[0035] Taking Mr. Zhang, a 56-year-old male user, as an example, his heart rate fluctuated abnormally during his daily health monitoring. The multi-source data synchronization submodule of the health data acquisition module started working, with data deployed on the smart bracelet, chest patch sensor, and infrared thermometer being transmitted at different frequencies: the bracelet heart rate was measured once per second, the patch ECG at 100 times per second, and the thermometer every 30 seconds. After detecting timestamp differences, the system aligned all data sources with nanosecond-level precision. Finding a 500-millisecond delay in the temperature data, it used cubic spline interpolation to generate a time-aligned dataset, ensuring all parameters were analyzed under the same time reference. The anomaly filtering submodule detected three consecutive abnormal peaks exceeding 180 bpm in the heart rate data. This module called the preset age-related heart rate range (set by the user to 50-160 bpm) and identified these points as outliers. Simultaneously, it found two missing points in the temperature data, which were filled using linear interpolation of the preceding and following valid data. An outlier was found in the blood oxygen saturation data, momentarily dropping to 85%. This was determined to be caused by motion artifacts and replaced with adjacent normal values. The final output was a 5-minute cleaned sequence of health data, containing 360 heart rate points, 300 blood oxygen saturation points, and 10 body temperature points. The standardization submodule normalized the cleaned data: heart rate data was converted from raw bpm values ​​to Z-scores, blood oxygen saturation was converted to percentage deviation, and body temperature was converted to an offset relative to basal body temperature. All data was resampled to a uniform frequency of 4 times per second, generating a health dataset containing 1440 standardized data points for use by subsequent modules.

[0036] The physical examination parameter optimization module initiates processing. The motion parameter extraction submodule retrieves similar cases from the historical database and finds a similar heart rate fluctuation case from 3 months ago. At that time, the robotic arm used a 30-degree joint angle and a 5-degree angular velocity per second to scan the precordial area, with a dwell time of 2 seconds per data set. These parameters are categorized into high-load levels and stored in the physical examination motion parameter library, marked as "tachycardia mode". The path optimization submodule analyzes the current abnormal fluctuation area and determines that it mainly occurs in the area from the 4th intercostal space on the left sternal border to the apex of the heart. The system selects 5 physical examination paths covering this area from the motion parameter library and calculates the efficiency score of each path: Path A covers 90% of the area and takes 12 seconds, scoring 8.7; Path B covers 95% of the area and takes 15 seconds, scoring 8.2. Finally, the path with the highest score, Path A, is selected as the candidate path. The force adaptation submodule obtains the user's chest wall features through a 3D vision sensor and detects that the area has a large surface curvature and thin tissue thickness. Combined with the high-load level, the initial contact pressure is set to 0.8N, and the pressure gradient is 0.2N / time. The system generates the final set of physical examination operation parameters, including the coordinate sequence of path A (x1, y1, z1) to (x8, y8, z8), the movement speed curve v=5deg / s, and the pressure parameter set {0.8N, 0.2N / step}.

[0037] Throughout the process, the data synchronization submodule continuously monitors the clock offset of each sensor, performing fine-tuning calibration every 10 seconds. The anomaly filtering submodule dynamically adjusts the threshold range based on real-time data quality, temporarily relaxing the upper and lower limits of heart rate by 10 bpm when the user is slightly active. The standardization processing submodule automatically activates a high-order interpolation algorithm to ensure data smoothness when it detects a low body temperature sampling rate. The motion parameter extraction submodule simultaneously searches for other similar user cases, finding that male users of the same age typically require 30% more time for examination in this area. The path optimization submodule considers path length, area coverage, and time cost when calculating efficiency scores, with weights set to 0.4, 0.4, and 0.2, respectively. The force adaptation submodule, based on real-time electromyography signal feedback, lowers the initial contact pressure from 0.8N to 0.6N to avoid user discomfort. All parameter adjustments are recorded in the operation log, including the reason for each adjustment, numerical changes, and effect evaluation. This data is updated in real-time to the user's personal health record, providing optimization references for subsequent examination operations. After generating the final set of physical examination operation parameters, the system immediately starts the robotic arm positioning program to prepare for the actual health examination operation.

[0038] Example 3: The environmental adaptability adjustment module continuously acquires environmental data through the environmental factor acquisition submodule. This submodule deploys high-precision temperature and humidity sensors within the robot's operating space and uses a depth vision sensor to capture the three-dimensional offset of the user's body position relative to the standard detection position. The raw environmental data undergoes noise suppression processing based on Kalman filtering, and then a moving average algorithm with a window size of 5 is used to smooth data fluctuations, ultimately generating an environmental feature vector containing temperature, humidity, and body position offset. The parameter influence analysis submodule receives the physical examination operation parameter set from the physical examination parameter optimization module, calculates the correlation coefficient between each element in the environmental feature vector and each parameter in the physical examination parameter set, uses Pearson correlation analysis to quantify the degree of influence of environmental factors on the physical examination operation, and selects parameters with an absolute correlation coefficient greater than 0.7 as significantly influential parameters, generating a parameter sensitivity list sorted by the degree of influence. The dynamic matching submodule adjusts the physical examination path and intensity parameters based on this sensitivity list, iteratively optimizes the physical examination plan by constructing an environment-parameter response model, evaluates the quality indicators of the physical examination data after each adjustment, until the environmental adaptability score reaches the predetermined standard, and finally outputs the environmental response physical examination plan.

[0039] The rate of change extraction submodule of the dynamic health threshold setting module monitors the physiological parameter flow in real time, calculates the first derivative using the central difference method to obtain the instantaneous rate of change, tracks the extreme points of the rate of change within a configurable sliding time window, and records the occurrence time and magnitude of the maximum and minimum values ​​to form a dynamic rate of change sequence. The threshold stratification submodule accesses the user's historical health database, extracts the statistical distribution characteristics of recent health data, uses the percentile method to delineate the boundaries of abnormal levels, and incorporates the extreme value information of the dynamic rate of change sequence into the threshold adjustment process. The adjusted threshold boundaries are calculated in the following way:

[0040] in: This represents the basic threshold calculated based on historical data percentiles. and These represent the maximum and minimum values ​​of the rate of change within the sliding window, respectively. It is the time interval between the maximum and the minimum. To adjust the coefficients, based on parameter types and empirical values, the final generated multi-level health warning thresholds include three levels: normal range, attention interval, and abnormal interval.

[0041] The environmental factor acquisition submodule employs a multi-sensor fusion strategy for data processing. Temperature and humidity sensors collect environmental data once per second, while the depth vision system captures changes in user posture at 30Hz. All data streams are timestamped before being fed into the processing pipeline. A Kalman filter is configured with different process and observation noise parameters for each type of sensor. The moving average window size is dynamically adjusted based on data stability, automatically shrinking to improve response speed when data fluctuations are significant. The parameter influence analysis submodule uses batch processing for correlation calculation. Each time a new environmental feature vector is received, it is compared with the most recent 100 sets of examination parameters to form a sample set for correlation analysis. The threshold for determining significantly influential parameters can be adjusted according to the actual application scenario, and the list of sensitive parameters is updated every 5 minutes. The iterative optimization process of the dynamic matching submodule includes a parameter sensitivity weighting mechanism, prioritizing highly sensitive parameters. After each parameter adjustment, the adaptability is evaluated by analyzing the stability of the subsequent three examination data. The environmental adaptability score comprehensively considers three dimensions: data acquisition completeness, measurement error rate, and operation time.

[0042] The sliding window mechanism of the rate of change extraction submodule supports dynamic configuration. Heart rate and blood oxygen parameters use a 10-second window, while body temperature parameters use a 30-second window. The extreme value detection algorithm avoids false peaks caused by noise, requiring extreme points to satisfy the condition of monotonic change between preceding and following data points. The historical data range of the threshold stratification submodule is set to the most recent 30 days by default. Percentile division uses P5 and P95 as normal range boundaries, and P2.5 and P97.5 as abnormal interval boundaries. The dynamic adjustment coefficient α is set according to the parameter type: 0.2 for heart rate, 0.1 for blood oxygen, and 0.05 for body temperature. The update frequency of the graded health warning threshold is consistent with the physiological parameter sampling rate. The threshold boundary is recalculated each time new data is collected, but significant adjustments require three consecutive verifications before taking effect. Environmental data acquisition and processing use a dedicated coprocessor to ensure timeliness. Threshold calculation tasks are assigned to background threads to avoid blocking the main data stream. All parameter adjustments are subject to rate-of-change limits to prevent oscillations. The system maintains an audit trail of all historical adjustment records for troubleshooting.

[0043] Example 4: The health trend prediction module processes real-time physiological data streams through a multi-parameter fusion submodule. This submodule receives time-series data of heart rate, blood oxygen, and body temperature from a standardized health dataset. First, it extracts trend features for each parameter sequence, calculating its slope, fluctuation amplitude, and stability index within a certain time window. Then, it uses a dynamic weight allocation mechanism to calculate the weight coefficient based on the measurement quality, historical coefficient of variation, and clinical importance of each parameter at the current moment. The weight calculation process considers sensor confidence scores and data integrity indicators. The final multimodal health feature set includes a weighted comprehensive health index and independent trend indicators for each parameter.

[0044] The pattern inference submodule performs feature matching on the historical abnormal pattern library. This pattern library stores multi-parameter change patterns under various typical abnormal conditions, such as tachycardia with decreased blood oxygenation and hypothermia with increased heart rate variability. Each pattern template contains a set of standardized parameter feature vectors and their corresponding risk labels. The matching process uses a similarity algorithm based on dynamic time warping to calculate the morphological similarity between the current health feature set and each pattern template. The top K patterns with the highest similarity are selected as candidate reference patterns, as shown in Table 1.

[0045] Table 1: Fragments of the Historical Anomaly Pattern Library

[0046] In the deviation calculation stage, for each candidate reference pattern, the Mahalanobis distance between the current feature set and the pattern template in each dimension is calculated. This distance metric takes into account the correlation between parameters and can effectively eliminate the influence of the difference in the scale of each parameter. The deviation of each dimension is converted into a probability value through a Gaussian weighting function. The final risk probability output is the fusion value of the weighted probabilities of each candidate pattern. The weighting coefficient depends on the similarity ranking of each pattern with the current feature.

[0047] The risk prediction value generation process includes a multi-stage verification mechanism. The initial trend prediction result is based on the first and second derivative features of the weighted health index, and is corrected by combining the risk probability value. When the risk probability exceeds a set threshold, the system initiates multi-timescale verification to check the consistency between recent and long-term trends, and compares whether the direction of change of related parameters conforms to physiological laws. The corrected health risk prediction value includes a risk level identifier, confidence score, and description of expected evolution trend. The historical abnormal pattern library is updated and maintained using an incremental learning mechanism. Whenever a new abnormal event is confirmed, the system adds the event's feature vector to the pattern library after anonymization. At the same time, a clustering algorithm is used to periodically clean up the pattern library, merging similar patterns and eliminating outdated patterns. The feature vector dimension of the pattern template is completely consistent with the multimodal health feature set, ensuring the consistency of the matching process.

[0048] The dynamic weight adjustment of the multi-parameter fusion submodule follows an adaptive principle. When the sensor report quality of a certain parameter decreases, the system automatically reduces its weight coefficient while increasing the weight allocation of other parameters. The weight calculation cycle is synchronized with the data acquisition frequency, but the rate of change of the weight coefficient is limited to avoid drastic fluctuations in the comprehensive health index caused by frequent weight adjustments. The K-value selection in the pattern matching process is adaptively determined based on the noise level of the current data. A smaller K-value is used to improve sensitivity when the data quality is high, and a larger K-value is used to enhance stability when the data quality is low. The similarity calculation adopts an improved dynamic time warping algorithm, which is robust to the scaling and offset of the time axis and can effectively handle pattern matching of different durations.

[0049] The risk probability fusion calculation employs a confidence-based weighted average method, assigning the highest-ranked candidate pattern the largest weight coefficient. However, considering the differences between patterns, the system lowers the overall confidence score and initiates additional data collection procedures to acquire more physiological parameters when multiple candidate patterns provide inconsistent risk directions. Multi-parameter fusion and pattern inference are executed in parallel. The pattern matching process runs in a background thread without affecting the real-time data stream. Risk prediction values ​​are generated using an asynchronous update mechanism, ensuring continuous risk assessment output even when some parameters are temporarily missing.

[0050] Example 5: The physical examination feedback control module continuously monitors the real-time changes of physiological parameters through the error calculation submodule. This submodule receives the health risk prediction value from the health trend prediction module and the actual collected physiological parameter sequence. It compares the absolute deviation between the two using a sliding time window. The window size is dynamically adjusted according to the parameter characteristics. The heart rate parameter uses a 15-second window, the blood oxygen parameter uses a 30-second window, and the body temperature parameter uses a 60-second window. The deviation is calculated by accumulating the absolute values. The sum of the differences between the predicted value and the measured value is calculated within each window period. Then, according to the body region mapping rules, the error value is assigned to the corresponding spatial region to generate an error distribution map containing the error magnitude and spatial distribution. The parameter iteration submodule initiates the adjustment program based on the error distribution map. For areas where the error value exceeds the set threshold, the density of nodes in the examination path is automatically increased. The number of nodes added is directly proportional to the magnitude of the error. At the same time, the movement speed of the examination robot arm in that area is adjusted. Areas with larger errors correspond to lower scanning speeds and higher data acquisition frequencies. After each parameter adjustment, the error change rate of the new round is recorded immediately. The adjustment effect is evaluated by comparing the error change trends before and after the adjustment. If the error reduction rate does not reach the expected target, a new round of parameter optimization is initiated. This process is repeated until the error change rate converges to the predetermined threshold range.

[0051] The error calculation submodule employs a multi-dimensional evaluation strategy for deviation analysis. Besides calculating absolute numerical differences, it also considers the consistency of parameter change trends. Even small numerical deviations with opposite trends are still marked as significant errors. Spatial region mapping is based on a 3D human body model, mapping each physiological parameter measurement point to a corresponding area on the body surface. The error distribution map is visualized using a heatmap, with different color depths representing the severity of the error. The path adjustment algorithm in the parameter iteration submodule includes kinematic constraint detection. Newly added path nodes must meet the requirements of reachability and motion smoothness for the robotic arm. Node density increases follow a gradual principle, with each iteration increasing the number of nodes by a maximum of 50%. Motion speed adjustment is coordinated with node density changes; lower motion speeds are used in high-density node areas to ensure measurement accuracy. The speed adjustment range is controlled between 30% and 100% of the rated speed. The error change rate monitoring adopts a differential calculation method to record the error change over three consecutive sampling periods after each parameter adjustment. Convergence determination is based on the second derivative of the error change rate. When the change in the rate of change is less than a set threshold, it is considered to have reached the convergence state. The convergence threshold is set according to the parameter type: 0.5% per second for heart rate parameter, 0.2% per second for blood oxygen parameter, and 0.1% per second for body temperature parameter.

[0052] When the error in a certain area remains consistently high, cross-module collaborative adjustment is triggered. The health trend prediction module is notified to reassess the risk prediction model for that area, while the health data acquisition module is instructed to improve signal acquisition quality in that area, forming a closed-loop optimization system. The robotic arm motion control employs an impedance control strategy, adjusting the contact stiffness and damping coefficient of the end effector according to the error level. Lower stiffness and higher damping are used in high-error areas to ensure contact safety. Motion trajectory generation uses B-spline curve interpolation to ensure smooth path transitions and avoid mechanical vibrations caused by abrupt turns. Data acquisition frequency adjustment is synchronized with changes in motion speed; the sampling rate is increased accordingly during low-speed motion. Time-interval sampling is used, and the dwell time at each sampling point is set according to parameter characteristics. Heart rate measurement requires at least three complete heartbeat cycles, blood oxygen measurement requires 5 seconds of stable reading time, and body temperature measurement requires 10 seconds of thermal equilibrium time.

[0053] During the iterative optimization process, operation logs are maintained, detailing the error state before each parameter adjustment, the adjusted parameters, the error changes after adjustment, and the convergence status. This data is used to optimize and adjust algorithm parameters and provides decision-making references for similar situations in the future. The system has a maximum iteration limit; if convergence is not achieved after reaching the iteration limit, it automatically switches to conservative mode, using default parameter settings and issuing a manual intervention request to avoid infinite looping and consuming system resources. Conservative mode parameters are set based on the average value of historical successful cases. The generated adaptive body control strategy includes the optimized path node coordinate sequence, motion speed parameters for each node, data acquisition frequency settings, and robotic arm force control parameters. These parameters are encapsulated into an executable instruction set and directly sent to the robot control system for execution.

[0054] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0055] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A robotic non-invasive health checkup and monitoring platform, characterized in that, The platform comprises: The health data acquisition module obtains user physiological parameters based on multi-modal sensors, integrates body temperature, heart rate, and blood oxygen data, eliminates abnormal values, and performs time series alignment to generate a standardized health data set; The physical sign load analysis module calculates the fluctuation difference of different physiological parameters based on the standardized health data set, marks abnormal fluctuation areas, and generates a physical sign load distribution state value; The physical examination parameter optimization module extracts the angle and duration combination of robot physical examination actions based on the physical sign load distribution state value, filters the optimal physical examination path and contact force parameter, and generates a physical examination operation parameter set; The environmental adaptability adjustment module collects environmental temperature and humidity and user body position data based on the physical examination operation parameter set, analyzes the influence of environmental factors on physical examination parameters, and generates an environmental response physical examination scheme; The health threshold dynamic setting module extracts the real-time physiological parameter change rate based on the environmental response physical examination scheme, dynamically adjusts the abnormal judgment threshold in combination with historical health data, and generates a graded health warning threshold; The health trend prediction module uses fuzzy reasoning to jointly analyze the change trend of multiple parameters based on the graded health warning threshold, and generates a health risk prediction value; The physical examination feedback control module compares the error between real-time data and the prediction value based on the health risk prediction value, adjusts the physical examination operation parameters and optimizes the scanning frequency, and generates an adaptive physical examination control strategy.

2. The robot non-invasive health physical examination and monitoring platform according to claim 1, wherein: The physical sign load distribution state value includes a physiological parameter fluctuation set, a regional load level set, and an abnormal marker set; The physical examination operation parameter set includes path planning parameters and force control parameters; The environmental response physical examination scheme includes temperature and humidity adaptation parameters and body position compensation parameters; The graded health warning threshold includes real-time change rate parameters and historical baseline parameters; The health risk prediction value includes trend joint analysis parameters and mode classification parameters; The adaptive physical examination control strategy includes error compensation parameters and frequency optimization parameters.

3. The robotic non-invasive health check-up and monitoring platform as claimed in claim 1, wherein, The health data acquisition module comprises: The multi-source data synchronization submodule aligns the physiological parameter acquisition time stamp based on the time sequence of multi-modal sensors, eliminates time offset data, and generates a time-aligned original data set; The abnormality filtering submodule identifies and replaces outliers based on the preset physiological parameter reasonable range, uses adjacent data to interpolate missing values, and generates a cleaned health data sequence; The standardized processing submodule normalizes the cleaned health data sequence, unifies the dimension and sampling frequency, and generates a standardized health data set.

4. The robotic non-invasive health check-up and monitoring platform as claimed in claim 1, wherein, The physical examination parameter optimization module comprises: The action parameter extraction submodule analyzes robot joint angle and execution duration historical data based on the physical sign load distribution state value, stores them according to load level classification, and generates a physical examination action parameter library; The path optimization submodule filters physical examination paths that cover abnormal fluctuation areas in the physical examination action parameter library, calculates path efficiency scores, and generates a candidate physical examination path set; The force adaptation submodule adjusts the contact force parameter according to the user's body surface characteristics, assigns a pressure gradient based on the load level, and generates a physical examination operation parameter set.

5. The robotic non-invasive health check-up and monitoring platform as claimed in claim 1, wherein, The environmental adaptability adjustment module comprises: The environment factor acquisition submodule acquires environment temperature and humidity and user body position offset data in real time, performs noise filtering and sliding average processing, and generates an environment feature vector; The parameter influence analysis submodule calculates a correlation coefficient of the environment feature vector and the physical examination operation parameter set, marks significant influence parameter items, and generates a parameter sensitivity list; The dynamic matching submodule adjusts the physical examination path and intensity parameters based on the parameter sensitivity list, iteratively verifies environment adaptability, and generates an environment response physical examination scheme.

6. The robotic non-invasive health check-up and monitoring platform as claimed in claim 1, wherein, The health threshold dynamic setting module includes: The change rate extraction submodule captures the first derivative of real-time physiological parameters, calculates the change extreme value in the sliding window, and generates a dynamic change rate sequence; The threshold layering submodule divides abnormal grades according to historical health data distribution percentiles, adjusts threshold boundaries in combination with the dynamic change rate sequence, and generates graded health warning thresholds.

7. The robotic non-invasive health check-up and monitoring platform as claimed in claim 1, wherein, The health trend prediction module includes: The multi-parameter fusion submodule assigns dynamic weights to heart rate, blood oxygen, and body temperature data, calculates a joint change trend index, and generates a multi-modal health feature set; The mode inference submodule matches the current health feature set based on a historical abnormal mode library, corrects trend prediction bias, and generates a health risk prediction value.

8. The robotic non-invasive health check-up and monitoring platform as claimed in claim 7, wherein, The mode inference submodule performs the following operations: Extract the top K modes with the highest similarity to the current health feature set from the historical abnormal mode library; Calculate the deviation of the current feature from each mode, and output the risk probability after weighted fusion; Correct the initial trend prediction result according to the risk probability, and generate a health risk prediction value.

9. The robotic non-invasive health check-up and monitoring platform as claimed in claim 1, wherein, The physical examination feedback control module includes: The error calculation submodule compares the absolute deviation of real-time physiological parameters and health risk prediction values, divides error levels by region, and generates an error distribution map; The parameter iteration submodule adjusts the physical examination path coverage range according to the error distribution map, optimizes the single physical examination duration and repetition frequency, and generates an adaptive physical examination control strategy.

10. The robotic non-invasive health check-up and monitoring platform as claimed in claim 9, wherein, The parameter iteration submodule performs the following operations: Increase the node density of the physical examination path in the high error area, dynamically adjust the movement speed of the physical examination robot arm according to the error level, record the error change rate after parameter adjustment, and cyclically optimize until the convergence threshold is reached.

Images