Smart ankle pump exercise monitoring system and method for a shared platform

Abstract

The application discloses a shared platform intelligent ankle pump exercise monitoring system and method, and belongs to the technical field of measurement and diagnosis. The method specifically comprises the following steps: collecting joint angles, plantar pressure, local skin temperature and microcirculation blood flow parameters of ankle pump exercise, and performing fusion processing to obtain a joint feature vector; based on rehabilitation data of different users in the shared platform, a population stratified rehabilitation reference model is constructed; the population stratified rehabilitation reference model is divided in multiple dimensions according to disease types, rehabilitation stages and training scenes; the joint feature vector is matched with the population stratified rehabilitation reference model to identify the non-synchronization of cycle improvement and action amplitude, and differential auxiliary regulation is performed. The application can support more scientific and efficient rehabilitation training under the support of intelligent guidance and group comparison, accumulate large-scale rehabilitation data resources for medical institutions, and form a combination mode of clinical application and group rehabilitation.

Description

Technical Field

[0001] This invention belongs to the field of measurement and diagnostic technology, specifically a smart ankle pump motion monitoring system and method for a shared platform. Background Technology

[0002] Ankle pump exercises are a basic rehabilitation training method that promotes venous return in the lower limbs through repeated flexion and extension of the ankle joint. They are widely used to prevent deep vein thrombosis, improve blood circulation, and aid in postoperative lower limb function recovery. Current rehabilitation practices largely rely on patients performing these exercises themselves or under the guidance of therapists, and their effectiveness depends heavily on the patient's mastery of the rhythm, range of motion, and frequency of the movements. However, differences among patients in terms of disease type, rehabilitation stage, and physical condition often make it difficult to achieve ideal personalized rehabilitation results with traditional standardized movement guidance.

[0003] With the development of wearable devices and sensor technology, some intelligent ankle joint motion monitoring systems have gradually emerged. These systems typically collect users' motion data through accelerometers, angle sensors, or pressure sensors and provide feedback on the terminal device. However, the technical approach of these systems is mostly focused on monitoring a single data source, such as only detecting angle changes or the number of movements, lacking comprehensive analysis of physiological data such as blood circulation indicators, skin temperature, and microcirculation, making it difficult to dynamically adapt to individual differences and achieve refined control. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes an intelligent ankle pump motion monitoring system and method with a shared platform. Through multi-source data fusion, group reference model construction, differentiated regulation, and collaborative interaction mechanisms, it achieves intelligent, personalized, and dynamic evolution of the rehabilitation process.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] The shared platform's intelligent ankle pump motion monitoring method includes:

[0007] The joint angle, plantar pressure, local skin temperature, and microcirculation blood flow parameters of ankle pump movement were collected and fused to obtain a joint feature vector.

[0008] Based on rehabilitation data from different users on a shared platform, a stratified rehabilitation reference model is constructed, which is divided into multiple dimensions according to disease type, rehabilitation stage and training scenario.

[0009] The joint feature vector is matched with the population-stratified rehabilitation reference model to identify the asynchronous nature of cyclical improvement and movement amplitude, and differentiated auxiliary regulation is performed.

[0010] Specifically, the construction of a stratified rehabilitation reference model based on rehabilitation data from different users on the shared platform includes:

[0011] The rehabilitation data of different users in the shared platform are initially stratified according to disease type, rehabilitation stage and training scenario to obtain multiple stratified datasets;

[0012] The motion signals and physiological signals in the hierarchical dataset are aligned on the time axis to form a multi-dimensional time-series matching sequence.

[0013] Motion amplitude parameters, cyclic change parameters, and associated joint features are extracted from the temporal matching sequence and aggregated within the hierarchical dataset.

[0014] Based on the aggregation results, corresponding rehabilitation reference curves are generated according to the hierarchical dimensions;

[0015] When new users' rehabilitation data is integrated, the rehabilitation reference curve is automatically adjusted based on the differences from the existing stratification dimensions, forming a continuously evolving population-stratified rehabilitation reference model.

[0016] Specifically, motion amplitude parameters, cyclic change parameters, and associated joint features are extracted from the temporal matching sequence and aggregated within the hierarchical dataset, including:

[0017] The ankle joint angle change in the time-series matching sequence is segmented, and the extreme points of amplitude and their change trajectories are extracted to form a set of motion amplitude parameters.

[0018] The circulating blood flow signal and skin temperature signal in the time-series matching sequence are sampled and discretized to obtain a multi-dimensional parameter set of cyclic changes;

[0019] The motion amplitude parameter set and the multi-dimensional parameter set are cross-paired on the same time axis to generate cross-modal joint features;

[0020] The cross-modal joint features are segmented according to time windows, and a corresponding pattern segment sequence is generated within each time window;

[0021] The pattern fragment sequences are aggregated according to the hierarchical dimensions of the hierarchical dataset to form the corresponding joint feature aggregation results.

[0022] Specifically, generating corresponding rehabilitation reference curves based on the aggregation results and according to the hierarchical dimensions includes:

[0023] The combined feature aggregation results are divided into intervals according to rehabilitation stage, exercise amplitude level, and cyclic change rate to form a multi-dimensional feature distribution set;

[0024] Representative turning points and steady-state segments are identified in the set of feature distributions and used as candidate reference nodes.

[0025] The candidate reference nodes are connected according to time sequence and hierarchical dimension to form multiple node links;

[0026] The node links are fused within the hierarchical dimension to generate corresponding rehabilitation reference curves.

[0027] Specifically, the step of performing difference matching between the joint feature vector and the stratified rehabilitation reference model for the population, identifying the asynchrony between cyclical improvement and movement amplitude, and performing differentiated auxiliary regulation includes:

[0028] The joint feature vector is projected onto the corresponding dimension of the population-stratified rehabilitation reference model, and the offset of each feature point from the reference node is compared on the time axis, and continuous offset intervals are marked.

[0029] Within the continuous offset interval, the rhythm of motion amplitude variation and the fluctuation rhythm of the cyclic signal are identified respectively, and the time correspondence between the two is established to identify the pattern segments with phase misalignment, peak delay or rhythm drift.

[0030] The pattern segments are quantified in multiple dimensions according to offset magnitude, duration and number of repetitions, and then divided into different difference levels based on these dimensions.

[0031] Based on the difference level, the corresponding intervention action is retrieved from the preset regulation rule base, and combined with the user's rehabilitation stage and hierarchical dimension, an auxiliary regulation instruction set is generated.

[0032] The auxiliary control instruction set is embedded into the subsequent ankle pump exercise sequence in the form of timestamps, and combined with the user's real-time training data through segmented execution, so as to perform differentiated auxiliary control according to the evolution over time.

[0033] Specifically, the step of projecting the joint feature vector onto the corresponding dimension of the stratified rehabilitation reference model, comparing the offset of each feature point with the reference node on the time axis, and marking continuous offset intervals includes:

[0034] The joint feature vectors are projected into the stratified rehabilitation reference model of the population, and the corresponding reference dimensions are determined.

[0035] Each feature value is retrieved point by point in the projection result and compared with the node value of the reference dimension to obtain the offset data sequence;

[0036] Retrieve consecutive data segments exceeding a preset threshold in the offset data sequence and identify them as candidate difference segments;

[0037] The candidate difference segments are numbered in chronological order, and the corresponding offset intervals are marked in the projection results.

[0038] Specifically, the pattern fragments are quantified in multiple dimensions according to offset magnitude, duration, and number of repetitions, and then divided into different difference levels, including:

[0039] The difference intervals between the motion amplitude and the cyclic signal in the pattern segment are compared point by point, and the measurement results of the offset amplitude are generated.

[0040] The duration of the pattern segment on the time axis is retrieved to form a determination result of the duration;

[0041] Identify the number of repetitions of the pattern segment in the time-series matching sequence and generate the number of repetitions;

[0042] The offset amplitude measurement results, duration determination results, and number of repetitions are combined, and the pattern segments are divided into different difference levels according to preset multi-dimensional rules.

[0043] Specifically, based on the difference level, the corresponding intervention action is retrieved from the preset regulation rule base, and combined with the user's rehabilitation stage and hierarchical dimensions, an auxiliary regulation instruction set is generated, including:

[0044] Based on the level of difference, retrieve the corresponding set of intervention actions from the preset control rule base;

[0045] The set of intervention actions is matched with the user's rehabilitation stage and the corresponding hierarchical dimension to obtain the stage-appropriate action set;

[0046] The set of actions adapted to each stage is sorted and combined to form a temporary instruction sequence;

[0047] The temporary instruction sequence is converted into an executable set of auxiliary control instructions.

[0048] The intelligent ankle pump motion monitoring system of the shared platform is used to implement the intelligent ankle pump motion monitoring method of the shared platform, including: a data acquisition module, a model construction module, and an auxiliary control module;

[0049] The data acquisition module is used to collect joint angles, plantar pressure, local skin temperature, and microcirculation blood flow parameters of ankle pump movement, and perform fusion processing to obtain a joint feature vector.

[0050] The model building module is used to construct a stratified rehabilitation reference model for a population based on rehabilitation data of different users in the shared platform. The stratified rehabilitation reference model is divided into multiple dimensions according to disease type, rehabilitation stage and training scenario.

[0051] The auxiliary control module is used to perform differential matching between the joint feature vector and the population stratified rehabilitation reference model, identify the asynchronous nature of cyclical improvement and movement amplitude, and perform differentiated auxiliary control.

[0052] Specifically, the auxiliary control module includes: a pattern segment recognition unit, a difference level division unit, and an auxiliary control unit;

[0053] The pattern segment recognition unit is used to mark continuous offset intervals and identify pattern segments within the continuous offset intervals.

[0054] The difference level division unit is used to quantify the pattern segment in multiple dimensions according to the offset amplitude, duration and number of repetitions, and divide it into different difference levels based on these dimensions.

[0055] The auxiliary control unit is used to retrieve the corresponding intervention action from the preset control rule base according to the difference level, and generate an auxiliary control instruction set in combination with the user's rehabilitation stage and hierarchical dimension, and perform auxiliary control.

[0056] Compared with the prior art, the beneficial effects of the present invention are:

[0057] This invention proposes an intelligent ankle pump exercise monitoring system and method on a shared platform. It constructs a stratified rehabilitation reference model for the population and performs differential matching based on users' joint feature vectors to generate differentiated control instructions, which are then embedded into the training process. This achieves dynamic coupling monitoring of movement amplitude and circulatory improvement, adaptive adjustment of individualized training programs, and cross-user collaborative interaction. This approach avoids the limitations of traditional single-user systems, such as single monitoring indicators and coarse feedback, and overcomes the isolation of the rehabilitation process. It enables patients to undergo more scientific and efficient rehabilitation training with intelligent guidance and group control support. Simultaneously, it accumulates large-scale, evolvable rehabilitation data resources for medical institutions, forming a combined model of clinical application and group rehabilitation. Attached Figure Description

[0058] Figure 1 Flowchart of the intelligent ankle pump motion monitoring method for the shared platform provided by the present invention;

[0059] Figure 2 The rehabilitation reference curve provided by this invention;

[0060] Figure 3 A schematic diagram of auxiliary control provided by the present invention;

[0061] Figure 4 This is a diagram of the architecture of the intelligent ankle pump motion monitoring system provided by the present invention. Detailed Implementation

[0062] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0063] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0064] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. In addition, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0065] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0066] Example 1

[0067] Please see Figures 1-3 The present invention provides an embodiment of an intelligent ankle pump motion monitoring method for a shared platform, comprising the following specific steps:

[0068] Step S1: Collect joint angles, plantar pressure, local skin temperature, and microcirculation blood flow parameters of ankle pump movement, and perform fusion processing to obtain a joint feature vector.

[0069] In this embodiment, a multi-source sensor collaborative acquisition method is used to synchronously acquire and fuse kinematic and physiological signals generated during ankle pumping movements. Specifically, joint angles are extracted by an inertial measurement unit to reflect the flexion and extension amplitude and rhythm of the ankle joint; plantar pressure is acquired by a distributed pressure sensing element to characterize the plantar load distribution and force rhythm; local skin temperature is detected by a thermistor to capture the dynamic changes in peripheral temperature during training; and microcirculation blood flow parameters are measured by a photoplethysmography (PPG) sensor module to reflect the changing trend of peripheral blood flow in the lower limbs. It should be noted that the signals have differences in sampling frequency and dimensions in their original state. Therefore, after acquisition, timestamp calibration and amplitude normalization are performed to ensure the comparability of multi-source signals on the same time axis. Subsequently, a feature fusion algorithm is used to map and synthesize kinematic and physiological signals in the same feature space to obtain a joint feature vector that can simultaneously characterize movement performance and circulatory response.

[0070] Step S2: Based on the rehabilitation data of different users in the shared platform, construct a stratified rehabilitation reference model for the population. The stratified rehabilitation reference model for the population is divided into multiple dimensions according to disease type, rehabilitation stage and training scenario.

[0071] The specific steps of step S2 are as follows:

[0072] Step S201: Initially stratify the rehabilitation data of different users in the shared platform according to disease type, rehabilitation stage and training scenario to obtain multiple stratified datasets.

[0073] In this embodiment, multi-dimensional labeling and classification of multi-user rehabilitation data collected in the shared platform are used to construct an initial hierarchical data structure. Specifically, firstly, user data is classified according to disease type based on medical diagnosis results and medical records to distinguish rehabilitation differences under different pathological backgrounds. Secondly, according to the time process of individual rehabilitation, the data is divided into stages such as acute phase, recovery phase, and maintenance phase to reflect the temporal characteristics of training effects. Furthermore, combined with training scenario labels, data generated by users in bedside rehabilitation, rehabilitation center guidance, or home self-training environments are distinguished to reflect the impact of environmental differences on rehabilitation performance. It should be noted that, to avoid overlap or omission in data stratification, a multi-indexing method is used to label disease category, rehabilitation stage, and training scenario as composite dimensions, so that the data of the same user can be classified into the corresponding datasets under multiple dimensions. Through the above steps, multiple datasets with clear hierarchical attributes are finally obtained.

[0074] Step S202: Align the motion signals and physiological signals in the hierarchical dataset with the time axis to form a multi-dimensional time-series matching sequence.

[0075] In this embodiment, a unified temporal framework that can simultaneously reflect kinematic and physiological characteristics is established through the calibration and synchronization of multi-source signals in the time dimension. Specifically, motion signals such as joint angle changes and plantar pressure signals usually have high sampling frequencies, while physiological signals such as skin temperature and microcirculation blood flow parameters have relatively low sampling frequencies, resulting in inconsistent temporal resolution in the original data state. To resolve this difference, firstly, a unified timestamp benchmark is used to align the start times of various signals. Secondly, a combination of interpolation and downsampling is used to map high-frequency signals to the sampling points of low-frequency signals while maintaining their dynamic trend characteristics without distortion. Furthermore, all processed signals are reconstructed on the same time axis, forming multi-dimensional data points containing motion amplitude, pressure distribution, temperature changes, and blood flow fluctuations within each time slice. It should be noted that this alignment process not only ensures the comparability between different signals but also enables subsequent analysis to retrieve the correspondence between action performance and physiological response in a unified temporal sequence, thereby forming a multi-dimensional temporal matching sequence.

[0076] Step S203: Extract motion amplitude parameters, cyclic change parameters, and associated joint features from the temporal matching sequence, and aggregate them within the hierarchical dataset.

[0077] The specific steps of step S203 are as follows:

[0078] Step S2031: The ankle joint angle change in the time-series matching sequence is segmented, and the extreme value points of amplitude and their change trajectory are extracted to form a set of motion amplitude parameters.

[0079] In this embodiment, key points and trajectories characterizing the amplitude of motion are extracted by segmenting and analyzing the continuous temporal data of the ankle joint angle signal. Specifically, firstly, the joint angle change curve in the temporal matching sequence is divided into several segments according to the periodic fluctuation characteristics to ensure that each segment corresponds to a complete flexion-extension movement. Secondly, the peak and trough points of the angle change are detected in each segment to obtain the maximum extension amplitude and minimum retraction amplitude of the movement. Subsequently, these extreme points are connected in chronological order to obtain a trajectory curve reflecting the trend of amplitude change. Furthermore, the amplitude difference, trajectory fluctuation characteristics, and rhythm information of each segment are uniformly organized to construct a set of motion amplitude parameters. It should be noted that this set of motion amplitude parameters not only records the maximum and minimum angle positions of a single flexion-extension movement but also retains the continuity characteristics of multiple movements in the time dimension.

[0080] Step S2032: Sample and discretize the circulating blood flow signal and skin temperature signal in the time-series matching sequence to obtain a multi-dimensional parameter set of cyclic changes.

[0081] In this embodiment, by uniformly sampling and discretizing the circulatory blood flow signal and skin temperature signal in the time-series matching sequence, a multi-dimensional parameter set capable of describing the dynamic characteristics of circulation is formed. Specifically, firstly, sampling interval points are set on a unified time axis to ensure that the blood flow fluctuation signal and temperature change signal have corresponding data at the same time node. Secondly, peaks, troughs, and rising and falling slopes are extracted from the continuous blood flow curves according to the sampling points to construct a discrete feature set that reflects the rhythm of circulatory fluctuations. At the same time, the temperature value and the difference value between adjacent points are extracted from the skin temperature curve at the same sampling point to reveal the subtle change trend of peripheral temperature during the training process. Subsequently, the blood flow parameters and temperature parameters are combined in parallel at the sampling point level to obtain the composite circulatory features corresponding to the time segment. Finally, these composite features are arranged in sequence throughout the entire time interval to form a multi-dimensional parameter set of circulatory changes. It should be noted that this parameter set retains the temporal correspondence between circulatory fluctuations and temperature changes while discretizing.

[0082] Step S2033: The motion amplitude parameter set and the multi-dimensional parameter set are cross-paired on the same time axis to generate cross-modal joint features.

[0083] In this embodiment, cross-modal joint features that can simultaneously reflect both motor performance and physiological response are formed by cross-pairing the set of motion amplitude parameters and the multi-dimensional parameter set of cyclic changes under a unified time axis. Specifically, firstly, using time segments as the basic unit, parameters such as the extreme values ​​of motion amplitude and trajectory differences in each segment are paired with the blood flow fluctuation characteristics and temperature difference characteristics at the corresponding time points. Secondly, by establishing a matrix form with one-to-one correspondence of parameters within the time window, the coupling relationship between amplitude changes and cyclic fluctuations is solidified into a joint data structure. Furthermore, the pairing results in continuous time windows are sequentially superimposed to form a traceable correspondence between the motion mode and the cyclic mode in the time dimension. Finally, the joint data of each time window are collected to form a cross-modal joint feature sequence. It should be noted that this joint feature not only achieves the correspondence between motion and cycle at a single time point, but also establishes a dynamic coupled trajectory at the cross-time period level.

[0084] Step S2034: Segment the cross-modal joint features according to time windows, and generate corresponding mode fragment sequences within each time window.

[0085] In this embodiment, cross-modal joint features are segmented using time windowing, enabling the coupling relationship between motion and cyclic signals within different time periods to be structurally expressed as a sequence of pattern segments. Specifically, firstly, sliding or fixed window boundaries are set on a unified time axis to divide continuous cross-modal joint features into multiple adjacent or partially overlapping time intervals. Secondly, the corresponding combinations of amplitude parameters and cyclic parameters are extracted in each time window and arranged in chronological order to form a local feature chain under that window. Further, each local feature chain is merged and abstracted into an independent pattern segment to represent the motion-cyclic coupling pattern within that time window. Subsequently, the pattern segments generated by all time windows are concatenated in the order of the windows to form a complete sequence of pattern segments. It should be noted that through this time window segmentation and fragmentation process, cross-modal joint features are reorganized in the temporal structure.

[0086] Step S2035: Aggregate the pattern fragment sequence according to the hierarchical dimension of the hierarchical dataset to form the corresponding joint feature aggregation result.

[0087] In this embodiment, the pattern segment sequences obtained after the aforementioned time window segmentation are collected and integrated according to the hierarchical dimensions of the hierarchical dataset to obtain comparable joint feature aggregation results at the group level. Specifically, firstly, based on the disease type, rehabilitation stage, and training scenario in the hierarchical dimensions, the pattern segment sequences corresponding to each user are respectively assigned to the corresponding hierarchical dimensions. Secondly, the pattern segments of different users are merged within the same hierarchical dimension, and the motion amplitude features and cyclic features under the same time window are combined to form cross-user segment clusters. Further, representative key nodes and feature chains are retrieved in each segment cluster to eliminate noise interference caused by individual differences and highlight common feature patterns. Subsequently, the results of different segment clusters are accumulated and organized step by step at the hierarchical dimensions to generate an aggregation structure that can reflect the feature distribution at the group level. Finally, these aggregation structures are summarized into joint feature aggregation results. It should be noted that this aggregation process, through the merging of hierarchical dimensions and the alignment of cross-user segments, transforms individualized signals into comparable data units at the group level.

[0088] Step S204: Based on the aggregation results, generate the corresponding rehabilitation reference curve according to the hierarchical dimensions.

[0089] like Figure 2 As shown, the specific steps of step S204 are as follows:

[0090] Step S2041: Divide the combined feature aggregation results into intervals according to rehabilitation stage, movement amplitude level and cyclic change rate to form a multi-dimensional feature distribution set.

[0091] In this embodiment, by performing multi-dimensional interval processing on the aggregated results of joint features, the data features corresponding to different rehabilitation stages, movement amplitude levels, and cyclic change rates are hierarchically organized to construct a feature distribution set that reflects the rehabilitation patterns of the group. Specifically, firstly, based on the time progression of the rehabilitation stages, the aggregated results are divided into different stage intervals such as the acute recovery period, the functional reconstruction period, and the maintenance and consolidation period to reflect the temporal differences as the rehabilitation progresses. Secondly, based on the numerical range of the movement amplitude parameter, the user's movement amplitude is divided into low amplitude, medium amplitude, and high amplitude level intervals to reveal the performance characteristics under different training intensities. Furthermore, combined with the dynamic indicators of the cyclic signal, the blood flow change rate is graded and three intervals—slow, moderate, and fast—are set to distinguish different levels of cyclic response. Subsequently, the interval division results of the three dimensions are cross-combined in the data structure to generate composite interval units covering multiple rehabilitation scenarios. Finally, the corresponding joint features are aggregated in each composite interval unit to form a multi-dimensional feature distribution set. It should be noted that this interval division process, through multi-dimensional processing of stages, levels, and rates, makes the aggregated results comparable and hierarchical.

[0092] Step S2042: Identify representative turning points and steady-state segments in the feature distribution set and use them as candidate reference nodes.

[0093] In this embodiment, key locations that characterize the pattern of the rehabilitation process are extracted from the established multi-dimensional feature distribution set. Turning points and steady-state segments are used as candidate reference nodes to establish anchor points for the subsequent generation of reference curves. Specifically, firstly, trend analysis is performed on the feature sequences within each composite interval unit to identify nodes where the rate of numerical change changes significantly, which are then identified as potential turning points. Secondly, intervals where the values ​​remain stable or fluctuate below a threshold within a certain time range are retrieved and marked as steady-state segments. Furthermore, to avoid misjudgments caused by random fluctuations, common turning points and steady-state segments that repeatedly appear in multiple sample groups are screened through cross-interval comparison. Subsequently, these nodes are indexed in chronological order. Finally, the screened turning points and steady-state segments are stored together as candidate reference nodes in the data structure. It should be noted that through this process, the complex multi-dimensional feature distribution is simplified into a limited number of representative nodes.

[0094] Step S2043: Connect the candidate reference nodes according to time sequence and hierarchical dimension to form multiple node links.

[0095] In this embodiment, the identified candidate reference nodes are organized and linked according to the temporal evolution pattern and the hierarchical data attribution relationship to form a logically continuous node link. Specifically, firstly, within each hierarchical dimension, the candidate reference nodes are arranged according to their timestamp order to ensure that the sequential relationship between different nodes is preserved. Secondly, within the same hierarchical dimension, connecting edges are established between adjacent nodes to form local links for a single user group or a single stage. Furthermore, to enhance the hierarchical nature of the links, nodes that cross hierarchical dimensions but have overlapping time intervals are connected across dimensions to form a parallel link structure that can reflect group differences. Subsequently, each local link and cross-dimensional link is summarized to generate a composite node link set covering multiple groups and multiple stages. Finally, all node links are uniformly encoded and indexed. It should be noted that through this connection process, the scattered set of nodes is transformed into a link structure with order and hierarchy.

[0096] Step S2044: The node links are fused within the hierarchical dimension to generate the corresponding rehabilitation reference curve.

[0097] In this embodiment, by performing hierarchical fusion processing on the constructed node links, discrete node sequences are transformed into continuous curves to form a reference trajectory that can represent the rehabilitation pattern of the group. Specifically, firstly, within the same hierarchical dimension, node links belonging to the same category are aligned, and the connection relationship between adjacent nodes is smoothed according to their time index. Secondly, overlapping nodes appearing in the same time interval in multiple links are weighted and merged to reduce the interference of individual differences on the overall curve. Furthermore, the node distribution characteristics unique to each hierarchical dimension are preserved during the fusion process, and representative nodes across links are used as anchor points to ensure that the curve maintains temporal consistency and hierarchical specificity globally. Subsequently, by connecting anchor points and merged nodes segment by segment, a continuous curve skeleton covering the entire rehabilitation stage is generated. Finally, the curve skeleton is recorded in the data structure as a rehabilitation reference curve. It should be noted that this fusion process, through hierarchical integration of node links, transforms the originally discrete multiple links into a reference curve with overall trends and hierarchical characteristics.

[0098] Figure 2The diagram shows a rehabilitation reference curve, used to characterize the dynamic evolution trend of joint features in different rehabilitation stages. Specifically, the vertical axis represents the intensity of the joint features formed by the fusion of motion amplitude parameters and cyclic change parameters, and the horizontal axis represents the sequence of stages corresponding to the rehabilitation process. The curve gradually rises in the initial stage, indicating that the user's motion amplitude and cyclic signal gradually establish a matching relationship in the early rehabilitation process. Subsequently, the curve shows significant peaks and troughs, representing asynchronous fluctuations between motion and cyclic response in the middle stage of rehabilitation. The data in this stage is an important reference for identifying differences and generating intervention instructions. As it continues to evolve, the curve tends to have a relatively stable fluctuation pattern, indicating that the group of users shows convergence and stability in rhythm in the later stage of rehabilitation. It should be noted that this rehabilitation reference curve is generated through cross-user, multi-dimensional feature aggregation, which not only reflects the overall trend in the rehabilitation process, but also provides a standardized reference for comparing differences between individual users and group models.

[0099] Step S205: When new users' rehabilitation data is accessed, the rehabilitation reference curve is automatically adjusted according to the differences from the existing stratification dimensions to form a continuously evolving population stratification rehabilitation reference model.

[0100] In this embodiment, the original model is dynamically corrected and continuously evolved by performing difference analysis on the rehabilitation data of new users and the reference curves under the existing hierarchical dimensions. Specifically, firstly, when new user data is accessed, their rehabilitation stage, movement amplitude, and cyclical change rate are automatically identified, and their position in the existing hierarchical system is determined. Secondly, the joint feature vector of the user is projected onto the reference curve of the corresponding layer, and the segment closest to its features is retrieved to compare the degree of difference between the actual trajectory and the reference trajectory. Further, for intervals with significant differences, the offset and duration are recorded, and the frequency of similar differences is statistically analyzed at the cross-user data level to determine whether the hierarchical curve needs to be corrected. Subsequently, if the difference meets the preset correction rules, new nodes are introduced into the reference curve or the weights of existing nodes are adjusted so that the curve can cover the characteristic performance of new users. Finally, through iterative methods, the reference curve continuously evolves with the continuous supplementation of group data, gradually forming a population-stratified rehabilitation reference model that can reflect the characteristics of different rehabilitation groups. It should be noted that this process enables the reference model to maintain a stable structure while having the ability to be dynamically updated, thereby ensuring its long-term applicability and group representativeness.

[0101] Step S3: Perform difference matching between the joint feature vector and the population stratified rehabilitation reference model to identify the asynchronous nature of cyclic improvement and movement amplitude, and perform differentiated auxiliary regulation.

[0102] like Figure 3 As shown, the specific steps of step S3 are as follows:

[0103] Step S301: Project the joint feature vector onto the corresponding dimension of the population stratified rehabilitation reference model, compare the offset of each feature point with the reference node on the time axis, and mark the continuous offset intervals.

[0104] The specific steps of step S301 are as follows:

[0105] Step S3011: Project the joint feature vector into the population stratified rehabilitation reference model and determine the corresponding reference dimension.

[0106] In this embodiment, the location of individual data in a multidimensional reference system is achieved by establishing a mapping relationship between a joint feature vector and a stratified rehabilitation reference model. Specifically, firstly, the joint feature vector, composed of movement amplitude, plantar pressure, skin temperature, and circulatory blood flow parameters, is dimensionally standardized to eliminate differences in the dimensions of different signals. Secondly, the stratified dimensions corresponding to the vector, such as disease type, rehabilitation stage, and training scenario, are retrieved in the stratified rehabilitation reference model, and a mapping position is assigned to it in the multidimensional coordinate system. Further, each component of the joint feature vector is projected point by point to the corresponding dimension of the reference model according to the time index, so that the individual's data can establish a correspondence with the group reference node. Subsequently, by comparing the projected position with the node position in the existing reference curve, an initial location result reflecting the differences between the individual and the group is generated. Finally, the reference dimension attribution of the target user in the stratified model is obtained. It should be noted that this projection process ensures that individualized features can be accurately located in the group model, thereby achieving comparability across users and stages.

[0107] Step S3012: Retrieve each feature value point by point in the projection result and compare it with the node value of the reference dimension to obtain the offset data sequence.

[0108] In this embodiment, by comparing the projected feature points with the standard nodes in the reference dimension point by point, an offset data sequence that can characterize the difference between an individual and the reference model is obtained. Specifically, firstly, feature values ​​such as joint angle, plantar pressure, skin temperature, and circulating blood flow are retrieved one by one from the projection results according to the time index. Secondly, the reference node values ​​at the same time position are called in the corresponding dimension of the population stratified rehabilitation reference model as a comparison benchmark. Further, each feature value is compared with its corresponding reference node value by difference or interval to determine the offset of the feature at that moment. Subsequently, the offsets at all time points are arranged in sequence to form an offset data chain extending in time order. Finally, the offset data chain is organized into an offset data sequence. It should be noted that this point-by-point comparison method ensures the correspondence between individual data and group reference in the time dimension, so that the offset can reflect real-time and continuous dynamic differences.

[0109] Step S3013: Retrieve continuous data segments exceeding a preset threshold in the offset data sequence and identify them as candidate segments for difference.

[0110] In this embodiment, by filtering and identifying intervals that continuously exceed a threshold in the offset data sequence, candidate segments for potential abnormal or asynchronous performance are determined. Specifically, firstly, offset thresholds are set for different feature dimensions, which can be predefined based on population statistical distribution or clinical experience parameters. Secondly, the offset data sequence is searched point by point in chronological order. When the offset of a certain feature first exceeds the threshold, that time point is used as the starting mark of the difference segment. Further, if adjacent time points continue to exceed the threshold, they are continuously merged into the same segment until the offset falls back below the threshold, and the termination point is marked. Subsequently, all continuous data segments between the starting and ending points are collected and numbered to form a preliminary set of candidate segments. Finally, these candidate segments are stored as candidate difference segments. It should be noted that this continuous threshold-exceeding retrieval method can accurately capture intervals with significant differences in the temporal dimension, thereby avoiding misjudgments caused by single-point offsets. The preset threshold is set by those skilled in the art according to actual conditions.

[0111] Step S3014: Number the candidate difference segments according to time sequence and mark the corresponding offset intervals in the projection results.

[0112] In this embodiment, by uniformly numbering the candidate difference segments and marking them in the projection results, traceable positioning of difference intervals is achieved on the time axis and spatial mapping. Specifically, firstly, the candidate difference segments obtained in the previous processing step are arranged sequentially according to their starting points in the time series, and each segment is assigned a unique number for reference and comparison in subsequent analysis. Secondly, in the results of projecting the joint feature vector onto the population stratified rehabilitation reference model, the time segments corresponding to each candidate segment are retrieved and offset markings are made on the corresponding coordinate dimensions. Further, to ensure the consistency of difference markings across feature dimensions, the numbering information and the offset of each feature dimension are recorded synchronously to ensure that the differences between kinematic and physiological parameters within the same segment are accurately marked. Subsequently, all marked offset intervals are stored as a set of difference segments with numbered indexes. Finally, a difference marking result that retains both time sequence and spatial positioning is formed. It should be noted that this marking and numbering process allows the offset intervals to be structurally extracted from the original data chain, facilitating subsequent analysis and dynamic control.

[0113] Step S302: Within the continuous offset interval, identify the variation rhythm of motion amplitude and the fluctuation rhythm of the cyclic signal respectively, establish the time correspondence between the two, and identify the pattern segments with phase misalignment, peak delay or rhythm drift.

[0114] In this embodiment, within a defined continuous offset interval, rhythmic analysis is performed on the temporal characteristics of the motion amplitude signal and the circulatory signal, respectively, and potential asynchronous modes are identified by establishing a time correspondence. Specifically, firstly, the ankle joint angle change curve within the offset interval is periodically divided, and rhythmic characteristic parameters of flexion and extension movements are extracted, including the duration of a single cycle, the peak-to-valley interval, and the repetition frequency. Secondly, the circulatory blood flow signal and skin temperature signal within the same interval are subjected to fluctuation analysis, and parameters such as peak spacing, rise and fall periods, and fluctuation amplitude are extracted to describe the dynamic rhythm of circulation. Further, the motion rhythm parameters and circulatory rhythm parameters are matched one-to-one according to timestamps to establish a matching relationship between the two types of signals in the time dimension. Subsequently, the synchronicity of the two types of rhythms is compared, and segments exhibiting phase misalignment, peak delay, or rhythm drift are retrieved and labeled. Finally, these identified segments are output as mode segments. It should be noted that this method, through dual rhythmic analysis of motion and circulatory signals, enables cross-modal asynchronous performance to be clearly captured and recorded in the temporal structure.

[0115] Step S303: Quantify the pattern fragments in multiple dimensions according to offset amplitude, duration and number of repetitions, and classify them into different difference levels based on these dimensions.

[0116] The specific steps of step S303 are as follows:

[0117] Step S3031: Compare the difference intervals between the motion amplitude and the cyclic signal in the mode segment point by point, and generate the measurement result of the offset amplitude.

[0118] In this embodiment, the difference range between motion amplitude and cyclic signal in a pattern segment is accurately measured by point-by-point comparison, thereby obtaining the offset amplitude result that can quantify the asynchrony. Specifically, firstly, corresponding points of motion amplitude curve and cyclic signal curve on the same time axis are selected in the pattern segment, and the numerical parameters at each time point are extracted. Secondly, the values ​​of the two types of signals at the same time point are differentially calculated or the interval distance is measured to obtain the instantaneous offset. Further, the offset is recorded point-by-point throughout the difference range, and an offset curve that changes with time is generated to retain the dynamic characteristics of the difference in the time dimension. Subsequently, the numerical set in the offset curve is organized into the offset amplitude measurement result, including the average offset value, the maximum offset value, and the offset distribution. Finally, the measurement result is used as input data for subsequent difference grading to characterize the coupling degree between motion and cyclic signals in the difference range. It should be noted that this point-by-point comparison method can not only identify local instantaneous differences, but also characterize the continuity of differences in the time dimension.

[0119] Step S3032: Search for the duration of the pattern segment on the time axis to form a determination result of the duration.

[0120] In this embodiment, by retrieving the start and end positions of a pattern segment on the time axis, its duration in the temporal dimension is calculated to obtain a determination result for characterizing the persistence of differences. Specifically, firstly, the starting point where the offset first exceeds the threshold is identified in the time index of the pattern segment, and the ending point where the offset falls back below the threshold is determined. Secondly, the time span between the starting and ending points is taken as the duration of the pattern segment and marked on the time axis. Further, to avoid false segments caused by local fluctuations, a minimum interval threshold is set between adjacent offset intervals to merge excessively short discontinuous segments into continuous segments. Subsequently, the duration of all pattern segments is statistically analyzed and organized to generate the corresponding duration determination result. Finally, this determination result serves as an important indicator for difference classification. It should be noted that this method can effectively capture the stability and extensibility of differences in the temporal dimension, enabling difference identification to not only remain at the amplitude level but also have a basis for determining dynamic persistence.

[0121] Step S3033: Identify the number of times the pattern segment recurs in the time-series matching sequence and generate the number of times it recurs.

[0122] In this embodiment, by cyclically searching the time-series matching sequence, it is determined whether a specific pattern segment recurs in multiple time intervals, and the frequency of its occurrence is counted accordingly, thus forming a frequency index of the difference. Specifically, firstly, a feature template for the pattern segment is set in the time-series matching sequence. This template is composed of multi-dimensional parameters such as motion amplitude offset, cyclic signal fluctuation characteristics, and time span. Secondly, the entire sequence is scanned segment by segment along the time axis, and the parameters of each candidate interval are compared with the template for similarity. When the similarity exceeds a preset threshold, the interval is marked as a match. Furthermore, to avoid false matches caused by local signal noise, invalid results are eliminated by setting a minimum segment length and a segment overlap limit. Subsequently, all valid matching intervals are summarized in chronological order, and the cumulative number of occurrences is counted. Finally, this number is output as the recurrence count of the pattern segment, providing a frequency dimension for the determination of difference classification. It should be noted that this retrieval and statistical process can reveal the repetitive characteristics of asynchronous differences, so that the difference analysis not only has a single description of amplitude and time, but also reflects the occurrence pattern of differences in the group rehabilitation process.

[0123] Step S3034: Combine the offset amplitude measurement result, duration determination result, and number of repetitions, and divide the pattern segment into different difference levels according to the preset multi-dimensional rules.

[0124] In this embodiment, the quantitative results of differences across different dimensions are comprehensively processed, and pattern segments are divided into distinct difference levels through a preset rule system to achieve multi-dimensional classification. Specifically, firstly, the offset amplitude measurement results, duration determination results, and repetition frequency are matched according to time index to ensure that the three dimensions of data can form a correlation within the same pattern segment. Secondly, based on the preset multi-dimensional judgment rules, weight coefficients are assigned to the magnitude of the offset amplitude, the duration, and the number of repetitions to reflect the importance of each parameter in the overall evaluation. Further, the weighted multi-dimensional parameter values ​​are aggregated and compared under multi-level boundary conditions to determine whether the pattern segment should be classified into a mild, moderate, or severe difference level. Subsequently, the classified pattern segments are uniformly stored as a difference level sequence for subsequent call and generation of control instructions. Finally, a differentiated classification system covering multiple parameters and multiple levels is formed. It should be noted that this combination and classification process, through the weighted integration of multi-dimensional indicators, ensures that the difference classification no longer relies on a single factor but is based on a composite judgment of amplitude, time, and frequency, thereby guaranteeing the systematic and hierarchical nature of the difference classification.

[0125] Step S304: Based on the difference level, retrieve the corresponding intervention action from the preset regulation rule base, and generate an auxiliary regulation instruction set by combining the user's rehabilitation stage and hierarchical dimension.

[0126] The specific steps of step S304 are as follows:

[0127] Step S3041: Retrieve the corresponding set of intervention actions from the preset control rule base according to the difference level.

[0128] In this embodiment, the difference level is used as the retrieval index. The set of intervention actions corresponding to that level is called from a preset regulation rule base, thereby achieving differentiated action plan selection. Specifically, firstly, the level identifier corresponding to each pattern segment is read from the difference level sequence; this level identifier serves as a key parameter for rule base retrieval. Secondly, the index table in the rule base, divided according to level hierarchy, is queried. Entries corresponding to different levels store different categories of action plans, such as rhythm adjustment, amplitude correction, pause insertion, breathing guidance, and sequence rearrangement. Further, when an entry for the target level is retrieved, all its subordinate intervention action entries are extracted and output as a set. Subsequently, this action set is temporarily cached for later matching with the user's rehabilitation stage and hierarchical dimensions. Finally, an intervention action set corresponding one-to-one with the difference level is formed. It should be noted that this retrieval process, through the direct mapping between difference levels and rule base entries, enables regulation measures to be automatically called according to the level gradient, thereby ensuring a correspondence between the action plan and the degree of difference.

[0129] Step S3042: Match the set of intervention actions with the user's rehabilitation stage and the corresponding hierarchical dimension to obtain the stage-appropriate action set.

[0130] In this embodiment, a targeted set of stage-appropriate actions is selected by cross-matching the set of intervention actions retrieved under the difference level with the user's rehabilitation stage and stratification dimension. Specifically, firstly, the user's stage information, such as acute recovery period, functional reconstruction period, or maintenance and consolidation period, is extracted from the user's rehabilitation data, and the user's category in the stratification dimension, such as disease type, training scenario, and population group, is confirmed. Secondly, each action item in the intervention action set is retrieved and compared with the preset applicable stage and stratification conditions of the action item. When both are consistent with or highly correlated with the user's characteristics, the item is marked as a candidate action. Further, for candidate actions that meet multiple stratification conditions, they are sorted by priority weight to ensure the accuracy of the screening results. Subsequently, all candidate actions that meet the conditions and are sorted are merged into a stage-appropriate action set. Finally, this action set is output as the input for subsequent temporal combinations. It should be noted that this matching and screening process ensures that the action regulation not only corresponds to the difference level but also to the user's rehabilitation process and group characteristics, thereby achieving stratification and personalization of intervention measures.

[0131] Step S3043: Sort and combine the set of stage-adapted actions to form a temporary instruction sequence.

[0132] In this embodiment, the discrete action items are transformed into an executable temporary instruction sequence by temporalizing the stage-adapted action set, thus ensuring the reasonable connection of control measures in the time dimension. Specifically, firstly, each action item in the stage-adapted action set is initially sorted according to its preset execution priority, which is usually determined by the severity of the difference level, the degree of adaptation to the rehabilitation stage, and the weight of the hierarchical feature. Secondly, based on the sorting result, the temporal relationship between adjacent actions is logically sorted out. If there is a mutual exclusion or dependency relationship, it is corrected by adjusting the position or inserting a buffer interval. Further, after the action sorting is completed, each item is arranged sequentially in the form of a timestamp to generate a temporal chain with a start time, execution duration, and end position. Subsequently, this temporal chain is used as a combination framework to fill in the action items one by one while maintaining their original parameter attributes. Finally, a temporary instruction sequence with a clear sequence and execution nodes is formed. It should be noted that this sorting and combination process realizes the transition from an action set to a time sequence.

[0133] Step S3044: Convert the temporary instruction sequence into an executable auxiliary control instruction set.

[0134] In this embodiment, the temporary instruction sequence obtained through sorting and combination is transformed into an auxiliary control instruction set with clear execution logic and operation identifiers, thereby achieving standardized output that can directly drive the training process. Specifically, firstly, each action item in the temporary instruction sequence is formatted, a unique identifier is assigned to each action, and its execution parameters, including start time, duration, and rhythm parameters, are added. Secondly, actions with dependencies within the same sequence are logically bound to maintain sequential constraints or synchronous triggering during execution. Furthermore, different action items are transformed into instruction units according to a unified syntax rule, and error checking and redundant fields are added to the data structure to ensure the stability and fault tolerance of the instructions during execution. Subsequently, all instruction units are integrated to form an instruction set with a hierarchical structure, which includes both global control instructions and local refined action instructions. Finally, this instruction set is output as a directly executable auxiliary control scheme. It should be noted that this transformation process makes the abstract action sorting result operable and consistent in execution, thereby completing the mapping from logical sequence to execution instructions.

[0135] Step S305: The auxiliary control instruction set is embedded into the subsequent ankle pump exercise sequence in the form of timestamps, and combined with the user's real-time training data through segmented execution, and differentiated auxiliary control is performed according to the time evolution.

[0136] In this embodiment, differentiated and individualized auxiliary control is achieved by embedding the set of auxiliary control instructions into the temporal sequence of subsequent ankle pump exercises and dynamically executing them in conjunction with real-time training data. Specifically, firstly, based on the timestamp information in the instruction set, each control instruction is associated with a specific time period in the exercise sequence, ensuring that the instructions are triggered at the precise moment. Secondly, a segmented control strategy is adopted during execution, that is, instructions are loaded segment by segment according to preset time period boundaries at different exercise stages to avoid over-coupling caused by global intervention. Furthermore, the instruction execution results are compared with the real-time collected exercise and physiological data. Point comparison is performed, and if offset or asynchronous phenomena are found, the corresponding adjustment action is triggered first to achieve closed-loop regulation. Subsequently, the feedback data after segmented execution is updated and matched with subsequent unexecuted instructions to achieve dynamic adjustment and rolling optimization. Finally, through continuous iterative embedding and correction during the evolution of time, a differentiated auxiliary regulation mechanism that can flexibly adapt to changes in user status is formed. It should be noted that this method, through the combination of timestamp embedding and segmented execution, enables auxiliary regulation to have both global temporal continuity and local response flexibility, providing dynamic support for sustainable evolution in the rehabilitation process.

[0137] Example 2

[0138] Please see Figure 2 Another embodiment of the present invention provides: a smart ankle pump motion monitoring system for a shared platform, comprising: a data acquisition module, a model building module, and an auxiliary control module;

[0139] The data acquisition module is used to collect joint angles, plantar pressure, local skin temperature, and microcirculation blood flow parameters of ankle pump movement, and perform fusion processing to obtain a joint feature vector.

[0140] The model building module is used to construct a stratified rehabilitation reference model for a population based on rehabilitation data of different users in the shared platform. The stratified rehabilitation reference model is divided into multiple dimensions according to disease type, rehabilitation stage and training scenario.

[0141] The auxiliary control module is used to perform differential matching between the joint feature vector and the population stratified rehabilitation reference model, identify the asynchronous nature of cyclical improvement and movement amplitude, and perform differentiated auxiliary control.

[0142] The auxiliary control module includes: a pattern segment recognition unit, a difference level classification unit, and an auxiliary control unit;

[0143] The pattern segment recognition unit is used to mark continuous offset intervals and identify pattern segments within the continuous offset intervals.

[0144] The difference level division unit is used to quantify the pattern segment in multiple dimensions according to the offset amplitude, duration and number of repetitions, and divide it into different difference levels based on these dimensions.

[0145] The auxiliary control unit is used to retrieve the corresponding intervention action from the preset control rule base according to the difference level, and generate an auxiliary control instruction set in combination with the user's rehabilitation stage and hierarchical dimension, and perform auxiliary control.

[0146] In addition, the parts of the technical solutions provided in the embodiments of this application that are consistent with the implementation principles of the corresponding technical solutions in the prior art have not been described in detail, so as to avoid excessive elaboration.

[0147] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the invention. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for monitoring ankle pump movement on a shared platform, characterized in that, include: The joint angle, plantar pressure, local skin temperature, and microcirculation blood flow parameters of ankle pump movement were collected and fused to obtain a joint feature vector. Based on rehabilitation data from different users on a shared platform, a stratified rehabilitation reference model is constructed, which is divided into multiple dimensions according to disease type, rehabilitation stage and training scenario. The joint feature vector is matched with the population-stratified rehabilitation reference model to identify the asynchrony between cyclical improvement and movement amplitude, and differentiated auxiliary regulation is performed. The step of performing difference matching between the joint feature vector and the stratified rehabilitation reference model to identify the asynchrony between cyclical improvement and movement amplitude, and performing differentiated auxiliary regulation, includes: The joint feature vector is projected onto the corresponding dimension of the population-stratified rehabilitation reference model, and the offset of each feature point from the reference node is compared on the time axis, and continuous offset intervals are marked. Within the continuous offset interval, the rhythm of motion amplitude variation and the fluctuation rhythm of the cyclic signal are identified respectively, and the time correspondence between the two is established to identify the pattern segments with phase misalignment, peak delay or rhythm drift. The pattern segments are quantified in multiple dimensions according to offset magnitude, duration and number of repetitions, and then divided into different difference levels based on these dimensions. Based on the difference level, the corresponding intervention action is retrieved from the preset regulation rule base, and combined with the user's rehabilitation stage and hierarchical dimension, an auxiliary regulation instruction set is generated. The auxiliary control instruction set is embedded into the subsequent ankle pump exercise sequence in the form of timestamps, and combined with the user's real-time training data through segmented execution, so as to perform differentiated auxiliary control according to the evolution over time. The process of projecting the joint feature vector onto the corresponding dimension of the stratified rehabilitation reference model, comparing the offset of each feature point with the reference node on the time axis, and marking continuous offset intervals includes: The joint feature vectors are projected into the stratified rehabilitation reference model of the population, and the corresponding reference dimensions are determined. Each feature value is retrieved point by point in the projection result and compared with the node value of the reference dimension to obtain the offset data sequence; Retrieve consecutive data segments exceeding a preset threshold in the offset data sequence and identify them as candidate difference segments; The candidate difference segments are numbered in chronological order, and the corresponding offset intervals are marked in the projection results; The process involves quantifying the pattern fragments across multiple dimensions, including offset magnitude, duration, and frequency of repetition, and then classifying them into different levels of difference, including: The difference intervals between the motion amplitude and the cyclic signal in the pattern segment are compared point by point, and the measurement results of the offset amplitude are generated. The duration of the pattern segment on the time axis is retrieved to form a determination result of the duration; Identify the number of repetitions of the pattern segment in the time-series matching sequence and generate the number of repetitions; The measurement results of the offset amplitude, the determination results of the duration, and the number of repetitions are combined, and the pattern segments are divided into different difference levels according to the preset multi-dimensional rules. Based on the level of difference, the corresponding intervention action is retrieved from the preset regulation rule base, and combined with the user's rehabilitation stage and hierarchical dimensions, an auxiliary regulation instruction set is generated, including: Based on the level of difference, retrieve the corresponding set of intervention actions from the preset control rule base; The set of intervention actions is matched with the user's rehabilitation stage and the corresponding hierarchical dimension to obtain the stage-appropriate action set; The set of actions adapted to each stage is sorted and combined to form a temporary instruction sequence; The temporary instruction sequence is converted into an executable set of auxiliary control instructions.

2. The intelligent ankle pump motion monitoring method of the shared platform as described in claim 1, characterized in that, The aforementioned stratified rehabilitation reference model, constructed based on rehabilitation data from different users on the shared platform, includes: The rehabilitation data of different users in the shared platform are initially stratified according to disease type, rehabilitation stage and training scenario to obtain multiple stratified datasets; The motion signals and physiological signals in the hierarchical dataset are aligned on the time axis to form a multi-dimensional time-series matching sequence. Motion amplitude parameters, cyclic change parameters, and associated joint features are extracted from the temporal matching sequence and aggregated within the hierarchical dataset. Based on the aggregation results, corresponding rehabilitation reference curves are generated according to the hierarchical dimensions; When new users' rehabilitation data is integrated, the rehabilitation reference curve is automatically adjusted based on the differences from the existing stratification dimensions, forming a continuously evolving population-stratified rehabilitation reference model.

3. The intelligent ankle pump motion monitoring method of the shared platform as described in claim 2, characterized in that, Motion amplitude parameters, cyclic change parameters, and associated joint features are extracted from the temporal matching sequence and aggregated within the hierarchical dataset, including: The ankle joint angle change in the time-series matching sequence is segmented, and the extreme points of amplitude and their change trajectories are extracted to form a set of motion amplitude parameters. The circulating blood flow signal and skin temperature signal in the time-series matching sequence are sampled and discretized to obtain a multi-dimensional parameter set of cyclic changes; The motion amplitude parameter set and the multi-dimensional parameter set are cross-paired on the same time axis to generate cross-modal joint features; The cross-modal joint features are segmented according to time windows, and a corresponding pattern segment sequence is generated within each time window; The pattern fragment sequences are aggregated according to the hierarchical dimensions of the hierarchical dataset to form the corresponding joint feature aggregation results.

4. The intelligent ankle pump motion monitoring method of the shared platform as described in claim 3, characterized in that, The process of generating corresponding rehabilitation reference curves based on the aggregation results and according to the hierarchical dimensions includes: The combined feature aggregation results are divided into intervals according to rehabilitation stage, exercise amplitude level, and cyclic change rate to form a multi-dimensional feature distribution set; Representative turning points and steady-state segments are identified in the set of feature distributions and used as candidate reference nodes. The candidate reference nodes are connected according to time sequence and hierarchical dimension to form multiple node links; The node links are fused within the hierarchical dimension to generate corresponding rehabilitation reference curves.

5. A shared platform intelligent ankle pump motion monitoring system, used to implement the shared platform intelligent ankle pump motion monitoring method according to any one of claims 1-4, characterized in that, include: Data acquisition module, model building module, and auxiliary control module; The data acquisition module is used to collect joint angles, plantar pressure, local skin temperature, and microcirculation blood flow parameters of ankle pump movement, and perform fusion processing to obtain a joint feature vector. The model building module is used to construct a stratified rehabilitation reference model for a population based on rehabilitation data of different users in the shared platform. The stratified rehabilitation reference model is divided into multiple dimensions according to disease type, rehabilitation stage and training scenario. The auxiliary control module is used to perform differential matching between the joint feature vector and the population stratified rehabilitation reference model, identify the asynchronous nature of cyclical improvement and movement amplitude, and perform differentiated auxiliary control.

6. The intelligent ankle pump motion monitoring system of the shared platform as described in claim 5, characterized in that, The auxiliary control module includes: a pattern segment recognition unit, a difference level division unit, and an auxiliary control unit; The pattern segment recognition unit is used to mark continuous offset intervals and identify pattern segments within the continuous offset intervals. The difference level division unit is used to quantify the pattern segment in multiple dimensions according to the offset amplitude, duration and number of repetitions, and divide it into different difference levels based on these dimensions. The auxiliary control unit is used to retrieve the corresponding intervention action from the preset control rule base according to the difference level, and generate an auxiliary control instruction set in combination with the user's rehabilitation stage and hierarchical dimension, and perform auxiliary control.

Images