Rehabilitation nursing remote monitoring transmission system

Abstract

The invention discloses a rehabilitation nursing remote monitoring and transmission system, which comprises the following steps: collecting a plurality of physiological parameters of a user, judging the fluctuation condition of the physiological parameters, automatically sampling data when the parameter change exceeds a set threshold value, identifying state mutation by using an analysis method, and dynamically adjusting a sampling strategy by the system according to an identification result, the system can synchronize processed data and evaluation results between terminal equipment and a remote platform through a two-way communication mechanism, continuous monitoring and stage judgment of the rehabilitation process are achieved, and nursing intervention is assisted. The system has automatic analysis and feedback regulation and control capabilities, and is suitable for remote rehabilitation management scenes.

Description

Technical Field

[0001] The present invention relates to the technical field of remote rehabilitation management, and in particular to a rehabilitation nursing remote monitoring transmission system. Background Art

[0002] With the growing demand for telemedicine and intelligent rehabilitation, higher requirements are placed on continuous monitoring and individualized management of physiological status in rehabilitation care scenarios. Especially in applications such as home rehabilitation, postoperative recovery, and chronic disease intervention, the patient's physiological status often changes dynamically. How to achieve an efficient, accurate, and remotely controllable monitoring mechanism has become a key technical challenge in remote rehabilitation management.

[0003] In the existing technology, remote rehabilitation monitoring systems mostly use periodic sampling methods and combine static threshold judgment to perform health assessment. This type of method has obvious shortcomings in the following aspects: 1. Lack of event-driven mechanism: Most existing methods are based on fixed-frequency sampling and cannot dynamically adjust the sampling frequency according to the actual fluctuations of physiological state, resulting in a lot of redundant data and low energy efficiency.

[0004] 2. Delayed abnormality identification: Most methods judge mutation events based only on statistical analysis or fixed rules, making it difficult to capture and identify changes in recovery status in real time and at an early stage.

[0005] 3. Lack of feedback mechanism: The current system lacks the ability to adjust the closed loop between monitoring results and sampling strategies, and cannot reversely optimize the sampling trigger conditions based on the monitoring and analysis results.

[0006] 4. Rough expression of rehabilitation status: Traditional methods mostly judge health status through single-point indicator thresholds, lack modeling and hierarchical expression of the stage characteristics of the rehabilitation process, and are difficult to meet individualized management needs.

[0007] 5. Limited remote collaboration capabilities: The existing system has a relatively simple data synchronization mechanism between the edge and the remote management platform, and lacks two-way communication support for model results, warning information, and control parameters.

[0008] Therefore, how to provide a remote monitoring and transmission system for rehabilitation care is an urgent problem that those skilled in the art need to solve. Summary of the invention

[0009] One purpose of the present invention is to propose a remote monitoring and transmission system for rehabilitation care. The present invention adopts an event-driven dynamic sampling mechanism, a Bayesian online change point detection method, an adaptive sampling control strategy and a rehabilitation state segmentation recognition model to perform real-time analysis and multi-stage state evaluation of user physiological parameters. It describes in detail the entire process from data acquisition, change point recognition, sampling control to stage level output, and realizes rapid identification and response of key health events. It has the advantages of high monitoring sensitivity, low data redundancy, strong feedback adjustment capability and good remote collaborative management capability.

[0010] The rehabilitation nursing remote monitoring transmission system according to an embodiment of the present invention comprises the following steps: S1, data acquisition module, collects multi-source physiological parameters of the user, constructs a sliding time window and calculates the dynamic change amplitude of each multi-source physiological parameter, and performs data sampling when the dynamic change amplitude of any multi-source physiological parameter exceeds the set sampling trigger threshold to generate a sampling data sequence; S2, data preprocessing module, preprocesses the sampled data sequence, including filtering, noise reduction and time alignment, to generate standardized time series data; S3, change point detection module, based on standardized time series data, uses Bayesian online change point detection method to perform recursive analysis to obtain the posterior probability of the change point at each time point, and when the posterior probability of the change point is greater than the set probability judgment threshold, the corresponding time point is identified as the state mutation time point; S4, sampling control module, adjusts the sampling trigger condition according to the comparison result between the posterior probability of the change point at any time point and the set probability judgment threshold, and realizes closed-loop adaptive adjustment of sampling control; S5, a rehabilitation status recognition module, which segments the standardized time series data based on the identified state mutation time points, generates rehabilitation stage intervals, and outputs the rehabilitation stage level corresponding to each state mutation time point; S6, edge collaboration and transmission module, synchronously transmits standardized time series data, posterior probability of change points and rehabilitation stage level between the edge and the remote server through a two-way communication mechanism, where the edge is a local terminal device that performs data collection and analysis tasks, and the remote server is a rehabilitation monitoring platform for centralized storage and control. The communication mechanism includes uploading monitoring data and issuing remote control instructions to achieve remote monitoring and dynamic management of the rehabilitation process.

[0011] Optionally, the S1 specifically includes: S11, a physiological parameter acquisition unit, which acquires multi-source physiological parameters of the user to form multi-channel original physiological parameters for synchronous recording; S12, sliding window processing unit, constructs the original physiological parameters corresponding to each channel with a length of The sliding time window is updated according to a fixed step size to form multiple sliding time window segments that are continuous in time; S13, a change amplitude extraction unit, taking the original physiological parameters in each sliding time window as input, and calculating the corresponding dynamic change amplitude, the dynamic change amplitude Defined as: ; in, Indicates The original physiological parameter set within a sliding time window, Indicated in The last original physiological parameter in a sliding time window; S14, the sampling trigger judgment unit calculates the dynamic change amplitude The sampling trigger threshold is set Compare, if satisfied , a data sampling event is triggered; S15, a sampling data construction unit extracts and encapsulates all original physiological parameters in the sliding window corresponding to the trigger event into a group of sampling segments, and combines them in chronological order to form a sampling data sequence.

[0012] Optionally, the S14 specifically includes: S141, threshold configuration unit, setting sampling trigger threshold , the sampling trigger threshold is a positive real number, which is used to compare with the dynamic change amplitude to determine whether to trigger sampling; S142, single channel comparison unit, receiving dynamic change amplitude , and the corresponding and Compare and if the conditions are met , then mark the current sliding time window of the channel as a candidate trigger window; S143, a multi-channel fusion determination unit summarizes the candidate trigger windows of all channels, and generates a sampling trigger signal if at least one channel satisfies the comparison condition; S144, the trigger control unit extracts the data index corresponding to the current sliding time window based on the sampling trigger signal, and generates a sampling event instruction for driving the sampling data construction unit to perform a sampling data sequence generation operation.

[0013] Optionally, the S3 specifically includes: S31, a change point probability calculation unit receives standardized time series data, performs recursive analysis based on the Bayesian online change point detection method, and calculates the change point posterior probability value of each time point as a state mutation point. The change point posterior probability value is defined as: ; in, Indicates time point A standardized time series data subsequence of Indicates time point The length of the interval between the previous state mutation point, Indicates time point is the new state mutation point, Indicates the normalized time series data subsequence Under given conditions, the time point is the posterior distribution of the state mutation point, For time point The corresponding change point posterior probability value; S32, a change point probability sequence construction unit receives the change point posterior probability value at each time point , arranged in chronological order, generating a change point probability sequence ,in Indicates the end time point index of the time series data; S33, change point discrimination unit, setting probability judgment threshold , the posterior probability value of each change point in the change point probability sequence and Compare and judge the condition as follows ,in, To set the probability judgment threshold, the range is , used to define the confidence level of the state mutation point determination. When the conditions are met, the time point is determined. is a candidate state mutation point; S34, state mutation output unit, receives candidate state mutation points, performs tag confirmation and outputs a state mutation time point set ,in Indicates the time point identified as a state mutation.

[0014] Optionally, the S31 specifically includes: S311, observation sequence construction unit, receiving standardized time series data, at each time point Construct the corresponding time series data subsequence ; S312, prior modeling unit, define state variables Indicates time point The length of the interval between the previous state mutation point and the prior distribution of the for: ; in, is the prior probability of the state mutation point occurring, satisfying , , is the value of the interval length; S313, observation model evaluation unit, building conditional observation model , assuming that the standardized time series data subsequence within the stateless mutation point interval It obeys Gaussian distribution and is expressed as: ; in, Represents the mean of the current segment, which is updated by autoregression based on the data from the segment start point to the moment before the current time point. is the preset observation noise variance; S314, posterior probability calculation unit, based on prior distribution , conditional observation model , and the posterior distribution of the previous moment ,in, To standardize the time series data subsequence Under given conditions, the time point The posterior distribution generated by the change point recursion is calculated by recursion at the time point is the posterior probability value of the change point.

[0015] Optionally, the S4 specifically includes: S41, change point probability extraction unit, receiving change point probability sequence , and extract the time point The corresponding change point posterior probability value ; S42, threshold comparison unit, setting probability judgment threshold , the time point The corresponding change point posterior probability value Probability judgment threshold Compare and determine whether ; S43, sampling adjustment signal generating unit, when judging that the When the sampling adjustment command signal is generated ,in Indicates that the sampling trigger threshold needs to be adjusted dynamically. Indicates that there is no need to dynamically adjust the sampling trigger threshold; S44, triggering the threshold adjustment unit to receive the sampling adjustment instruction signal , dynamically update the sampling trigger threshold , the adjustment strategy is: ; in, is a positive real fixed adjustment coefficient, Indicates time point The sampling trigger threshold is the updated sampling trigger threshold; S45, feedback closed-loop control unit, updates the sampling trigger threshold Applied to the setting of trigger conditions in the data acquisition module, it is used to control the dynamic adjustment of subsequent sampling trigger thresholds to achieve closed-loop adaptive regulation of sampling control.

[0016] Optionally, the S44 specifically includes: S441, threshold receiving unit, receiving sampling trigger threshold And sampling adjustment command signal ,in, Indicates whether adjustment is needed; S442, the threshold value calculation unit, when sampling and adjusting the instruction signal When , the threshold adjustment operation is triggered, and the sampling trigger threshold is updated according to the linear decreasing strategy. The update formula is: ; in, is a positive real fixed adjustment coefficient; S443, the threshold holding unit, when sampling and adjusting the instruction signal When , the current sampling trigger threshold remains unchanged, that is: ; S444, threshold output unit, outputs the adjusted sampling trigger threshold , used to set the trigger conditions in the subsequent data acquisition module and as the starting threshold for the next cycle of dynamic sampling control.

[0017] Optionally, the S5 specifically includes: S51, mutation point receiving unit, receiving state mutation time point set ,in Indicates the time point identified as a state mutation; S52, stage segmentation unit, according to the state mutation time point set Segment the standardized time series data in chronological order to construct multiple non-overlapping time intervals , where each stage interval is defined as: ; in, , , Indicates The time interval corresponding to each rehabilitation stage is Represents the total number of time intervals in the rehabilitation phase; S53, stage feature extraction unit, for each rehabilitation stage interval The standardized time series data in the calculation of statistical characteristic indicators, including mean, variance, trend slope, constitute the stage characteristic vector ; S54, stage level determination unit, based on the stage feature vector Introducing stage-level mapping functions , mapping each stage interval to a rehabilitation stage level : ; in, For the Stages of rehabilitation level.

[0018] The beneficial effects of the present invention are: (1) The present invention combines an event-driven dynamic sampling mechanism, a Bayesian online change point detection method and an adaptive sampling control strategy to achieve efficient monitoring of multi-source physiological parameters of users during remote rehabilitation care. It can timely identify key physiological state mutations, dynamically adjust the sampling strategy, significantly reduce the amount of redundant data and system energy consumption, and at the same time improve the perception ability and response speed of abnormal states in the rehabilitation process.

[0019] (2) The present invention constructs a closed-loop control mechanism based on the posterior probability of the change point as feedback, which can automatically update the sampling trigger conditions according to the real-time judgment results of the state mutation, realize adaptive adjustment of the sampling frequency, and effectively improve the resource utilization efficiency and system monitoring sensitivity. It is particularly suitable for resource-constrained edge deployment environments.

[0020] (3) Based on the results of mutation point identification, the present invention performs phased segmentation processing on the rehabilitation data and extracts segmentation features. The multi-stage identification and evaluation of the rehabilitation process is realized through the rehabilitation stage level mapping model, so that the system has the ability to automatically distinguish and grade the rehabilitation stages, and supports individualized rehabilitation management and optimization of remote intervention timing.

[0021] (4) The present invention realizes the synchronous transmission of standardized time series data, change point posterior probability and rehabilitation stage grade results through a two-way communication mechanism between the edge device and the remote management platform. It has the capabilities of remote monitoring, model control and data interaction, and improves the collaborative intelligence level and remote operability of the rehabilitation nursing system. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings: Figure 1This is the overall architecture diagram of the rehabilitation nursing remote monitoring transmission system proposed in the present invention. DETAILED DESCRIPTION

[0023] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0024] refer to Figure 1 , a rehabilitation nursing remote monitoring transmission system, comprises the following steps: S1, data acquisition module, collects multi-source physiological parameters of the user, constructs a sliding time window and calculates the dynamic change amplitude of each multi-source physiological parameter, and performs data sampling when the dynamic change amplitude of any multi-source physiological parameter exceeds the set sampling trigger threshold to generate a sampling data sequence; S2, data preprocessing module, preprocesses the sampled data sequence, including filtering, noise reduction and time alignment, to generate standardized time series data; In this implementation, the sampling data sequence is first filtered using a bandpass filtering algorithm to effectively remove high-frequency noise and low-frequency drift in the physiological parameter signal, and then the time alignment of the multi-channel data is completed through the interpolation method and the unified timestamp strategy to ensure the synchronization of various physiological parameters on the same time axis. Finally, the processed data is standardized to make parameters of different dimensions and numerical ranges comparable, and to generate standardized time series data in a unified format. This processing flow provides a stable, synchronous, and computable input data source for subsequent change point detection and rehabilitation stage identification, which significantly improves the accuracy of model analysis and the overall robustness of the system.

[0025] S3, change point detection module, based on standardized time series data, uses Bayesian online change point detection method to perform recursive analysis to obtain the posterior probability of the change point at each time point, and when the posterior probability of the change point is greater than the set probability judgment threshold, the corresponding time point is identified as the state mutation time point; S4, sampling control module, adjusts the sampling trigger condition according to the comparison result between the posterior probability of the change point at any time point and the set probability judgment threshold, and realizes closed-loop adaptive adjustment of sampling control; S5, a rehabilitation status recognition module, which segments the standardized time series data based on the identified state mutation time points, generates rehabilitation stage intervals, and outputs the rehabilitation stage level corresponding to each state mutation time point; S6, edge collaboration and transmission module, synchronously transmits standardized time series data, posterior probability of change points and rehabilitation stage level between the edge and the remote server by building a two-way communication mechanism, where the edge is a local terminal device that performs data collection and analysis tasks, and the remote server is a rehabilitation monitoring platform for centralized storage and control. The communication mechanism includes uploading monitoring data and issuing remote control instructions to achieve remote monitoring and dynamic management of the rehabilitation process.

[0026] This implementation method deploys a data collection and analysis module at the edge end. After completing the standardized processing of the user's physiological parameters, state mutation identification, and rehabilitation stage level judgment in real time at the local terminal, the standardized time series data, the posterior probability of the change point, and the rehabilitation stage level are synchronously uploaded to the remote rehabilitation monitoring platform using the constructed two-way communication mechanism, and the control parameters and intervention instructions issued by the platform are received at the same time, realizing two-way synchronization of data and control instructions. This mechanism ensures the real-time consistency of the data state and control logic between the edge device and the remote platform, which not only improves the continuity and responsiveness of remote monitoring, but also realizes the implementation of efficient rehabilitation management and dynamic adjustment strategies based on edge computing.

[0027] In this implementation, S1 specifically includes: S11, a physiological parameter acquisition unit, which acquires multi-source physiological parameters of the user to form multi-channel original physiological parameters for synchronous recording; S12, sliding window processing unit, constructs the original physiological parameters corresponding to each channel with a length of The sliding time window is updated according to a fixed step size to form multiple sliding time window segments that are continuous in time; The sliding time window constructs a continuous time period of a fixed length for the original physiological parameter sequence, and gradually slides forward with a set step length to achieve window update. During implementation, the system first sets the time length covered by each window, and divides the original multi-channel data in chronological order so that each window contains multiple physiological parameter samples in adjacent time periods. The system moves the window starting position according to the preset step length to generate multiple overlapping or non-overlapping window segments, thereby realizing continuous slicing processing of the data stream, which is convenient for subsequent trend analysis and trigger judgment.

[0028] S13, a change amplitude extraction unit, taking the original physiological parameters in each sliding time window as input, and calculating the corresponding dynamic change amplitude, the dynamic change amplitude Defined as: ; in, Indicates The original physiological parameter set within a sliding time window, Indicated in The last original physiological parameter in a sliding time window; S14, the sampling trigger judgment unit calculates the dynamic change amplitude The sampling trigger threshold is set Compare, if satisfied , a data sampling event is triggered; S15, a sampling data construction unit extracts and encapsulates all original physiological parameters in the sliding window corresponding to the trigger event into a group of sampling segments, and combines them in chronological order to form a sampling data sequence.

[0029] This embodiment obtains the multi-source physiological signals of the user through the physiological parameter acquisition unit, and dynamically processes each signal channel based on the sliding time window mechanism, extracting the change amplitude of the physiological parameters in each window as a characteristic indicator of state fluctuation. The system compares the change amplitude with the preset threshold, triggers data sampling when a mutation trend is detected, and encapsulates the original data in this period into a complete sequence of sampling fragments to construct a standard input for subsequent analysis. This method not only achieves the accurate capture of key physiological changes, but also avoids redundant sampling through the window mechanism, improves data processing efficiency and system response sensitivity, and provides basic data guarantee for subsequent change point detection and rehabilitation evaluation.

[0030] In this implementation manner, the S14 specifically includes: S141, threshold configuration unit, setting sampling trigger threshold , the sampling trigger threshold is a positive real number, which is used to compare with the dynamic change amplitude to determine whether to trigger sampling; S142, single channel comparison unit, receiving dynamic change amplitude , and the corresponding and Compare and if the conditions are met , then mark the current sliding time window of the channel as a candidate trigger window; S143, a multi-channel fusion determination unit summarizes the candidate trigger windows of all channels, and generates a sampling trigger signal if at least one channel satisfies the comparison condition; S144, the trigger control unit extracts the data index corresponding to the current sliding time window based on the sampling trigger signal, and generates a sampling event instruction for driving the sampling data construction unit to perform a sampling data sequence generation operation.

[0031] This implementation achieves intelligent identification of key physiological state fluctuations by setting a sampling trigger threshold and combining the dynamic change amplitude of multi-channel physiological parameters for fusion judgment. Specifically, the system first compares the dynamic change amplitude of the current sliding time window with the preset threshold on each channel, and marks the candidate trigger window that meets the conditions; then, cross-channel aggregation is performed in the multi-channel fusion judgment unit. As long as any channel shows significant fluctuations, the sampling control unit can be triggered to generate a sampling event instruction, thereby driving the subsequent data extraction and processing process. This mechanism effectively improves the response sensitivity to abnormal conditions, avoids the waste of resources caused by fixed-frequency sampling, and enhances the real-time and adaptability of the system in actual rehabilitation monitoring scenarios.

[0032] In this implementation, S3 specifically includes: S31, a change point probability calculation unit receives standardized time series data, performs recursive analysis based on the Bayesian online change point detection method, and calculates the change point posterior probability value of each time point as a state mutation point. The change point posterior probability value is defined as: ; in, Indicates time point A standardized time series data subsequence of Indicates time point The length of the interval between the previous state mutation point, Indicates time point is the new state mutation point, Indicates the normalized time series data subsequence Under given conditions, the time point is the posterior distribution of the state mutation point, For time point The corresponding change point posterior probability value; This formula normalizes a subsequence of time series data. , the posterior probability , the posterior probability value of the change point They are all dimensionless, taking into account the relationship between the current time series data and the historical time series data, and recursively calculating the probability value of a new change point at each time point, thereby realizing real-time judgment of the mutation moment. The core principle of this method is to use the Bayesian inference framework to integrate prior information with the current time series data, dynamically update the basis for judging the change point, and have online processing capabilities. It can continuously evaluate the possibility of state changes during the process of continuous data updating, thereby providing a reliable basis for mutation judgment for subsequent stage identification and sampling regulation.

[0033] S32, a change point probability sequence construction unit receives the change point posterior probability value at each time point , arranged in chronological order, generating a change point probability sequence ,in Indicates the end time point index of the time series data; S33, change point discrimination unit, setting probability judgment threshold , the posterior probability value of each change point in the change point probability sequence and Compare and judge the condition as follows ,in, To set the probability judgment threshold, dimensionless, range is , used to define the confidence level of the state mutation point determination. When the conditions are met, the time point is determined. is a candidate state mutation point; The change point identification method based on threshold judgment is implemented as follows: after receiving the change point posterior probability value corresponding to each time point, the system compares the probability value with the preset judgment threshold one by one. When the probability value of a certain time point exceeds the threshold, it is considered that there is a possibility of state mutation at this time point, and it is marked as a candidate change point. This method controls the sensitivity of change point identification by setting a threshold, so that the system can effectively filter out misjudgments caused by short-term fluctuations while maintaining real-time performance, thereby improving the accuracy of state mutation identification.

[0034] S34, state mutation output unit, receives candidate state mutation points, performs tag confirmation and outputs a state mutation time point set ,in Indicates the time point identified as a state mutation.

[0035] This implementation method recursively analyzes the standardized time series data based on the Bayesian online change point detection method, uses the state of the previous moment and the current observation value to jointly calculate the posterior probability of each time point being a state mutation point, and constructs a change point probability sequence arranged in chronological order. By setting the probability judgment threshold, the change point probability of each time point is judged, the potential state mutation point is identified, and finally the state mutation time point set is output. This method can detect the mutation moment of the user's physiological state in a timely manner when the amount of data is limited, improves the accuracy and real-time performance of abnormal recognition in the rehabilitation process, and provides highly reliable mutation information support for the subsequent rehabilitation state segmentation and sampling strategy adaptive adjustment.

[0036] In this implementation manner, the S31 specifically includes: S311, observation sequence construction unit, receiving standardized time series data, at each time point Construct the corresponding time series data subsequence ; S312, prior modeling unit, define state variables Indicates time point The length of the interval between the previous state mutation point and the prior distribution of the for: ; in, is the prior probability of the state mutation point occurring, satisfying , , is the value of the interval length; This formula adopts the form of geometric distribution. Its principle is based on the assumption that mutation events occur with a fixed probability on the time axis. It can effectively express the temporal sparsity and uncertainty of the occurrence of change points, and provides a mathematical basis for state transition modeling in subsequent Bayesian recursion. At the same time, it enhances the system's ability to model sudden state changes.

[0037] S313, observation model evaluation unit, build conditional observation model , assuming that the standardized time series data subsequence within the stateless mutation point interval It obeys Gaussian distribution and is expressed as: ; in, Represents the mean of the current segment, which is updated by autoregression based on the data from the segment start point to the moment before the current time point. is the preset observation noise variance; Based on the assumption of normal distribution, the formula regards the fluctuation of physiological parameters in each non-mutation stage as random fluctuations around a stable mean. The model can be used to calculate the degree of deviation of the current observation value from the historical data, thereby providing a stable probability basis for change point detection. The formula establishes a reference baseline for the stability of the physiological state by fitting the mean and variance of the data in the stage, so that when the state has a significant deviation, it can be reflected in the posterior probability in a timely manner, improving the sensitivity and accuracy of change point identification.

[0038] S314, posterior probability calculation unit, based on prior distribution , conditional observation model , and the posterior distribution at the previous moment ,in, To standardize the time series data subsequence Under given conditions, the time point The posterior distribution generated by the change point recursion is calculated by recursion at the time point is the posterior probability value of the change point.

[0039] This implementation is based on a standardized time series data sequence. First, a corresponding observation subsequence is constructed at each time point for subsequent Bayesian analysis. On this basis, the state variable is defined to represent the interval between the current moment and the previous change point, and its prior distribution is set to a geometric distribution to express the temporal probability structure of the mutation; the observation model is further constructed, assuming that the data in the continuous segment without change points obeys a Gaussian distribution, and the mean is estimated by autoregression based on the historical data of the segment, and a complete likelihood function is constructed in combination with the preset noise variance. Subsequently, by inputting the prior distribution, the observation model and the posterior probability of the previous moment into the Bayesian recursive formula, the posterior probability of each time point being a change point is calculated in real time. This method realizes the online identification of state mutations during the rehabilitation process. It not only has high timeliness and sensitivity, but also can maintain stable judgment capabilities in marginal environments with limited data, providing reliable support for the optimization of rehabilitation early warning and monitoring strategies.

[0040] In this implementation, S4 specifically includes: S41, change point probability extraction unit, receiving change point probability sequence , and extract the time point The corresponding change point posterior probability value ; S42, threshold comparison unit, setting dimensionless probability judgment threshold , the time point The corresponding change point posterior probability value Probability judgment threshold Compare and determine whether ; S43, sampling adjustment signal generating unit, when judging that the When the sampling adjustment command signal is generated ,in Indicates that the sampling trigger threshold needs to be adjusted dynamically. Indicates that there is no need to dynamically adjust the sampling trigger threshold; The method determines whether the posterior probability of the change point at the current time point exceeds the preset judgment threshold. If the condition is met, a control signal for adjusting the sampling strategy is generated. In actual operation, the system extracts the current change point probability value in each monitoring cycle and compares it with the threshold in real time. If a significant state mutation trend is identified, a control instruction is generated to start the update process of the subsequent sampling trigger threshold. The control signal is expressed in Boolean form, with clear judgment basis and operating conditions, ensuring that the sampling adjustment is effectively triggered only when the monitoring state mutates, forming a regulation mechanism driven by state probability.

[0041] S44, triggering the threshold adjustment unit to receive the sampling adjustment instruction signal , dynamically update the sampling trigger threshold , the adjustment strategy is: ; in, is a positive real fixed adjustment coefficient, Indicates time point The sampling trigger threshold is the updated sampling trigger threshold; This formula implements dynamic sampling strategy control based on the probability of state mutation. Its principle is to determine whether there is a mutation trend in the current system state, and then make a decreasing adjustment to the sampling trigger condition, so that the system automatically lowers the trigger threshold when possible physiological abnormalities are detected, thereby increasing the sampling frequency and enhancing the system's ability to respond to emergencies. This formula embodies a linear feedback mechanism, which updates the threshold with a limited amplitude according to the preset step factor, ensuring that the system maintains a dynamic balance between sensitivity and resource consumption, and achieving the adaptive sampling effect of closed-loop control.

[0042] S45, feedback closed-loop control unit, updates the sampling trigger threshold Applied to the setting of trigger conditions in the data acquisition module, it is used to control the dynamic adjustment of subsequent sampling trigger thresholds to achieve closed-loop adaptive regulation of sampling control.

[0043] This implementation method dynamically generates a sampling adjustment signal based on the comparison result between the posterior probability value of the change point and the set threshold, and adjusts the triggering condition of data sampling accordingly. Specifically, the posterior probability of the change point at the current time point is extracted and compared with the preset probability judgment threshold. If it is judged to be a state mutation trend, a sampling adjustment instruction signal is generated to drive the system to reduce the sampling trigger threshold in a fixed step size, and then the updated threshold is reapplied to the data acquisition module through the feedback mechanism to achieve closed-loop adaptive adjustment of sampling control. This mechanism significantly improves the dynamic response capability of the sampling strategy, enabling the system to automatically increase the sampling frequency when the state changes, reduce resource consumption in a stable state, and achieve the beneficial effect of effectively controlling energy consumption and data redundancy while ensuring monitoring sensitivity.

[0044] In this implementation manner, the S44 specifically includes: S441, threshold receiving unit, receiving sampling trigger threshold And sampling adjustment command signal ,in, Indicates whether adjustment is needed; S442, the threshold value calculation unit, when sampling and adjusting the instruction signal When , the threshold adjustment operation is triggered, and the sampling trigger threshold is updated according to the linear decreasing strategy. The update formula is: ; in, is a positive real fixed adjustment coefficient; S443, the threshold holding unit, when sampling and adjusting the instruction signal When , the current sampling trigger threshold remains unchanged, that is: ; S444, threshold output unit, outputs the adjusted sampling trigger threshold , used to set the trigger conditions in the subsequent data acquisition module and as the starting threshold for the next cycle of dynamic sampling control.

[0045] This implementation is based on the above content. By setting the adjustment signal control mechanism, when the sampling adjustment instruction signal is received as 1, the current sampling trigger threshold is dynamically updated using a linear decreasing strategy, and the update amplitude is determined by a fixed step coefficient; if the adjustment signal is 0, the threshold is kept unchanged. The updated sampling trigger threshold is applied as feedback input to the data sampling judgment of the next cycle, forming a closed-loop control link with self-adjustment of parameters. This method realizes the adaptive response of the sampling strategy to changes in monitoring sensitivity, can automatically enhance the data sampling capability when the key state changes suddenly, and suppress invalid sampling in the stable stage, which significantly improves the resource utilization and dynamic adaptability of the system.

[0046] In this implementation manner, S5 specifically includes: S51, mutation point receiving unit, receiving state mutation time point set ,in Indicates the time point identified as a state mutation; S52, stage segmentation unit, according to the state mutation time point set Segment the standardized time series data in chronological order to construct multiple non-overlapping time intervals , where each stage interval is defined as: ; in, , , Indicates The time interval corresponding to each rehabilitation stage is Represents the total number of time intervals in the rehabilitation phase; This formula divides the complete rehabilitation process into several non-overlapping time intervals by segmenting the continuous time series data according to the adjacent state mutation time points. This method uses the mutation point as the stage boundary to ensure that the data in each stage has relatively consistent dynamic characteristics. Its principle is to use state mutation as the natural dividing point of rehabilitation state change, thereby constructing a stage division that conforms to the logic of physiological changes, making the subsequent feature extraction and level determination more accurate and reliable, and avoiding the stage division error that may be caused by artificially setting a fixed time window.

[0047] The stage division method adopted uses the identified state mutation time points as the boundaries, and slices the standardized time series data in chronological order. In the specific implementation, all state mutation time points are first arranged in ascending time order, and then the time range between the current mutation point and the previous mutation point is used as a rehabilitation stage interval. The first stage starts from the monitoring start time, and the last stage continues to the monitoring end time, ensuring that each stage interval has clear start and end boundaries and completely covers the original time series data without overlap or omission, thereby realizing structured stage organization of the data.

[0048] S53, stage feature extraction unit, for each rehabilitation stage interval The standardized time series data in the calculation of statistical characteristic indicators, including mean, variance, trend slope, constitute the stage characteristic vector ; The statistical characteristic index traverses the standardized time series data in each rehabilitation stage interval, calculates the mean of all data points in the interval, adds up all the standardized time series data values ​​in the stage interval and divides them by the number of data points to characterize the overall level of the stage, calculates the variance, and calculates the square of the difference between each data point and the mean on the basis of the mean and takes the average to reflect the degree of data fluctuation, calculates the trend slope, uses the least squares method to perform a first-order linear regression fit on the data points, and extracts the slope value as an indicator of the data change trend within the stage. The above operations are performed independently in each segmented interval, and the three eigenvalues ​​output together constitute the stage eigenvector, which serves as the input for the subsequent rehabilitation stage level determination.

[0049] S54, stage level determination unit, based on the stage feature vector Introducing stage-level mapping functions , mapping each stage interval to a rehabilitation stage level : ; in, For the Stages of rehabilitation level.

[0050] The stage level determination is implemented by constructing a rule mapping function, specifically: the feature vector of each stage interval is input into a predefined grading rule set, the rule set is clearly divided based on the value range of the mean, variance and trend slope in the feature vector, and each set of rules corresponds to a unique rehabilitation stage level label; The system matches each rule condition one by one, determines the level of the current stage, and outputs the corresponding rehabilitation stage level result, which is used to indicate the rehabilitation status of the stage and realize the standardized hierarchical expression of the rehabilitation process.

[0051] This embodiment receives a set of state mutation time points identified by a change point detection module, divides the standardized time series data into multiple rehabilitation stage intervals in chronological order, extracts statistical features such as mean, variance and trend slope in each stage interval, constructs a stage feature vector, and then analyzes each feature vector through a preset stage level mapping function to output the corresponding rehabilitation stage level label, thereby realizing automatic segmentation and level classification of the rehabilitation process, helping to accurately judge the individual's rehabilitation progress, providing a phased decision-making basis for remote nursing intervention, and improving the intelligence and individualization of rehabilitation management.

[0052] Embodiment 1: In order to verify the feasibility of the present invention in implementation, the present invention was applied to the rehabilitation medicine center of a tertiary hospital to carry out a three-month actual clinical pilot. The center mainly focuses on neurological rehabilitation and orthopedic rehabilitation, and receives a large number of patients recovering from postoperative injuries, stroke rehabilitation and long-term chronic diseases every day. The rehabilitation assessment process of the center mainly relies on regular manual interviews and staged detection equipment, such as gait analyzers and surface electromyography, but due to the long detection cycle and the process relying on professional personnel to operate, there are problems such as the inability to continuously track changes in physiological status and the difficulty in timely capturing fluctuations in the patient's rehabilitation status, making it difficult to achieve accurate and dynamic rehabilitation intervention.

[0053] In this pilot, the hospital equipped 150 rehabilitation inpatients with edge acquisition terminal devices of the system of the present invention, which collected multi-source physiological parameters including heart rate, galvanic skin response, lower limb electromyography, respiratory rate, etc. The system automatically starts real-time data acquisition after each patient wears the device, dynamically analyzes the change range of physiological parameters through a sliding time window, and automatically triggers data sampling when any indicator changes beyond the set threshold, avoiding the waste of resources caused by redundant sampling.

[0054] After filtering, denoising and time alignment, the sampled data is passed to the embedded processing module, which calculates the probability of state mutation in real time based on the Bayesian online change point detection algorithm. The system adaptively adjusts the sampling trigger conditions based on the calculated change point posterior probability value. If the system detects that the mutation probability is higher than the judgment threshold, it will dynamically reduce the sampling trigger threshold, thereby increasing the sampling frequency during the critical period and achieving an automatic balance between resource allocation and monitoring sensitivity.

[0055] During the rehabilitation process, the system automatically segments the standardized data based on the change point determination results, extracts features such as the mean, variance, and change trend within each segment, and outputs the patient's rehabilitation stage level through the established rehabilitation stage mapping model, which is automatically synchronized to the remote rehabilitation management platform. Rehabilitation physicians can remotely evaluate the patient's current progress based on the level change trend and stage length, and adjust the rehabilitation plan accordingly, such as increasing the frequency of rehabilitation training, changing intervention methods, etc. At the same time, the remote platform can also return the control parameters to the edge terminal to achieve a true remote monitoring and two-way closed-loop adjustment.

[0056] During the three months of operation, the hospital collected rehabilitation data from 150 patients and conducted comparative analysis between the system and traditional rehabilitation methods. The following is some of the core data: Table 1 Comparison of the effects of the system of the present invention and traditional rehabilitation assessment methods

[0057] The data table shows that while the sampling frequency of the system of the present invention is reduced by 42%, the effective data capture rate is greatly improved to 91.8%, which significantly improves the energy efficiency and analysis effectiveness of the system. In terms of rehabilitation stage identification and grade output, the accuracy rate is improved from 73.2% of the traditional method to 94.5%, and the system can complete automatic grade output within 5 minutes, which greatly shortens the time required for manual evaluation. Since the staged evaluation mechanism is clearer, doctors can adjust the intervention plan in time according to the system feedback results. The response time is shortened from the traditional 24 hours to less than 6 hours, and the matching degree and satisfaction of the rehabilitation plan are significantly improved.

[0058] For example, during the pilot process, a postoperative stroke patient experienced abnormal fluctuations in heart rate and skin electricity during the second stage of rehabilitation. The system identified the sudden change in state and increased the sampling frequency through a dynamic adjustment mechanism. It confirmed the existence of rehabilitation fluctuations in this stage within 15 minutes and judged it as a degeneration from "moderate rehabilitation" to "initial recovery" level through the hierarchical model. After receiving this feedback, the doctor immediately adjusted the rehabilitation training plan from traditional endurance training to a combination of mild repetitive exercise and neural stimulation, avoiding regression caused by rehabilitation overload and shortening the subsequent recovery cycle.

[0059] In summary, the rehabilitation nursing remote monitoring and transmission system proposed in the present invention constructs a closed-loop intelligent rehabilitation management process from perception to evaluation to intervention feedback through real-time perception of physiological state, change point analysis, adaptive adjustment and stage level identification, which solves the problems of long cycle, slow response and rough classification of traditional rehabilitation monitoring, and significantly improves the visualization of the rehabilitation process, intelligent evaluation capability and intervention response efficiency, providing effective support for promoting the application of intelligent rehabilitation in clinical nursing.

[0060] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. Rehabilitation nursing remote monitoring transmission system, characterized in that: include: S1, data acquisition module, collects multi-source physiological parameters of the user, constructs a sliding time window and calculates the dynamic change amplitude of each multi-source physiological parameter, and performs data sampling when the dynamic change amplitude of any multi-source physiological parameter exceeds the set sampling trigger threshold to generate a sampling data sequence; S2, data preprocessing module, preprocesses the sampled data sequence, including filtering, noise reduction and time alignment, to generate standardized time series data; S3, change point detection module, based on standardized time series data, uses Bayesian online change point detection method to perform recursive analysis to obtain the posterior probability of the change point at each time point, and when the posterior probability of the change point is greater than the set probability judgment threshold, the corresponding time point is identified as the state mutation time point; S4, sampling control module, adjusts the sampling trigger condition according to the comparison result between the posterior probability of the change point at any time point and the set probability judgment threshold, and realizes closed-loop adaptive adjustment of sampling control; S5, a rehabilitation status recognition module, which segments the standardized time series data based on the identified state mutation time points, generates rehabilitation stage intervals, and outputs the rehabilitation stage level corresponding to each state mutation time point; S6, edge collaboration and transmission module, synchronously transmits standardized time series data, change point posterior probability and rehabilitation stage level between the edge and the remote server through a two-way communication mechanism, where the edge is a local terminal device that performs data collection and analysis tasks, and the remote server is a rehabilitation monitoring platform for centralized storage and control.

2. The rehabilitation nursing remote monitoring transmission system according to claim 1 is characterized in that: The S1 specifically includes: S11, a physiological parameter acquisition unit, which acquires multi-source physiological parameters of the user to form multi-channel original physiological parameters for synchronous recording; S12, sliding window processing unit, constructs the original physiological parameters corresponding to each channel with a length of The sliding time window is updated according to a fixed step size to form multiple sliding time window segments that are continuous in time; S13, a change amplitude extraction unit, taking the original physiological parameters in each sliding time window as input, and calculating the corresponding dynamic change amplitude, the dynamic change amplitude Defined as: ； in, Indicates The original physiological parameter set within a sliding time window, Indicated in The last original physiological parameter in a sliding time window; S14, the sampling trigger judgment unit calculates the dynamic change amplitude The sampling trigger threshold is set Compare, if satisfied , a data sampling event is triggered; S15, a sampling data construction unit extracts and encapsulates all original physiological parameters in the sliding window corresponding to the trigger event into a group of sampling segments, and combines them in chronological order to form a sampling data sequence.

3. The rehabilitation nursing remote monitoring transmission system according to claim 2 is characterized in that: The S14 specifically includes: S141, threshold configuration unit, setting sampling trigger threshold , the sampling trigger threshold is a positive real number, which is used to compare with the dynamic change amplitude to determine whether to trigger sampling; S142, single channel comparison unit, receiving dynamic change amplitude , and the corresponding and Compare and if the conditions are met , then mark the current sliding time window of the channel as a candidate trigger window; S143, a multi-channel fusion determination unit summarizes the candidate trigger windows of all channels, and generates a sampling trigger signal if at least one channel satisfies the comparison condition; S144, the trigger control unit extracts the data index corresponding to the current sliding time window based on the sampling trigger signal, and generates a sampling event instruction for driving the sampling data construction unit to perform a sampling data sequence generation operation.

4. The rehabilitation nursing remote monitoring transmission system according to claim 1 is characterized in that: The S3 specifically includes: S31, a change point probability calculation unit receives standardized time series data, performs recursive analysis based on the Bayesian online change point detection method, and calculates the change point posterior probability value of each time point as a state mutation point. The change point posterior probability value is defined as: ； in, Indicates time point A standardized time series data subsequence of Indicates time point The length of the interval between the previous state mutation point, Indicates time point is the new state mutation point, Indicates the normalized time series data subsequence Under given conditions, the time point is the posterior distribution of the state mutation point, For time point The corresponding change point posterior probability value; S32, a change point probability sequence construction unit receives the change point posterior probability value at each time point , arranged in chronological order, generating a change point probability sequence ,in Indicates the end time point index of the time series data; S33, change point discrimination unit, setting probability judgment threshold , the posterior probability value of each change point in the change point probability sequence and Compare and judge the condition as follows ,in, To set the probability judgment threshold, the range is , used to define the confidence level of the state mutation point determination. When the conditions are met, the time point is determined. is a candidate state mutation point; S34, state mutation output unit, receives candidate state mutation points, performs tag confirmation and outputs a state mutation time point set ,in Indicates the time point identified as a state mutation.

5. The rehabilitation nursing remote monitoring transmission system according to claim 4 is characterized in that: The S31 specifically includes: S311, observation sequence construction unit, receiving standardized time series data, at each time point Construct the corresponding time series data subsequence ; S312, prior modeling unit, define state variables Indicates time point The length of the interval between the previous state mutation point and the prior distribution of the for: ； in, is the prior probability of the state mutation point occurring, satisfying , , is the value of the interval length; S313, observation model evaluation unit, building conditional observation model , assuming that the standardized time series data subsequence within the stateless mutation point interval It obeys Gaussian distribution and is expressed as: ； in, Represents the mean of the current segment, which is updated by autoregression based on the data from the segment start point to the moment before the current time point. is the preset observation noise variance; S314, posterior probability calculation unit, based on prior distribution , conditional observation model , and the posterior distribution at the previous moment ,in, To standardize the time series data subsequence Under given conditions, the time point The posterior distribution generated by the change point recursion is calculated by recursion at the time point is the posterior probability value of the change point.

6. The rehabilitation nursing remote monitoring transmission system according to claim 4 is characterized in that: The S4 specifically includes: S41, change point probability extraction unit, receiving change point probability sequence , and extract the time point The corresponding change point posterior probability value ; S42, threshold comparison unit, setting probability judgment threshold , the time point The corresponding change point posterior probability value Probability judgment threshold Compare and determine whether ; S43, sampling adjustment signal generating unit, when judging that the When the sampling adjustment command signal is generated ,in Indicates that the sampling trigger threshold needs to be adjusted dynamically. Indicates that there is no need to dynamically adjust the sampling trigger threshold; S44, triggering the threshold adjustment unit to receive the sampling adjustment instruction signal , dynamically update the sampling trigger threshold , the adjustment strategy is: ； in, is a positive real fixed adjustment coefficient, Indicates time point The sampling trigger threshold is the updated sampling trigger threshold; S45, feedback closed-loop control unit, updates the sampling trigger threshold Applied to the setting of trigger conditions in the data acquisition module, it is used to control the dynamic adjustment of subsequent sampling trigger thresholds to achieve closed-loop adaptive regulation of sampling control.

7. The rehabilitation nursing remote monitoring transmission system according to claim 6 is characterized in that: The S44 specifically includes: S441, threshold receiving unit, receiving sampling trigger threshold And sampling adjustment command signal ,in, Indicates whether adjustment is needed; S442, the threshold value calculation unit, when sampling and adjusting the instruction signal When , the threshold adjustment operation is triggered, and the sampling trigger threshold is updated according to the linear decreasing strategy. The update formula is: ； in, is a positive real fixed adjustment coefficient; S443, the threshold holding unit, when sampling and adjusting the instruction signal When , the current sampling trigger threshold remains unchanged, that is: ； S444, threshold output unit, outputs the adjusted sampling trigger threshold , used to set the trigger conditions in the subsequent data acquisition module and as the starting threshold for the next cycle of dynamic sampling control.

8. The rehabilitation nursing remote monitoring transmission system according to claim 4 is characterized in that: The S5 specifically includes: S51, mutation point receiving unit, receiving state mutation time point set ,in Indicates the time point identified as a state mutation; S52, stage segmentation unit, according to the state mutation time point set Segment the standardized time series data in chronological order to construct multiple non-overlapping time intervals , where each stage interval is defined as: ； in, , , Indicates The time interval corresponding to each rehabilitation stage is Represents the total number of time intervals in the rehabilitation phase; S53, a stage feature extraction unit, for each time interval corresponding to the rehabilitation stage The standardized time series data in the calculation of statistical characteristic indicators, including mean, variance, trend slope, constitute the stage characteristic vector ; S54, stage level determination unit, based on the stage feature vector Introducing stage-level mapping functions , mapping each stage interval to a rehabilitation stage level : ； in, For the Stages of rehabilitation levels.

Images