Sarcopenia prevention exercise method and system for the elderly

Abstract

The present application relates to the technical field of sports science, and particularly relates to a sarcopenia prevention exercise method and system for the elderly. Time series correlation analysis is performed on multi-dimensional body state data of the target elderly person to extract a muscle function degradation process characteristic vector, a muscle group function dependency topological structure is constructed based on the characteristic vector, a cascading influence node muscle group and a function dependency path are identified to determine a target intervention muscle group set, a hierarchical progressive exercise scheme is generated, a bidirectional mapping relationship between action execution quality and physiological load response is established, a deviation direction and a deviation amplitude of the current action execution quality are obtained by using forward prediction and backward calculation, and real-time feedback of instant correction instructions containing an action adjustment vector is generated, a functional bottleneck muscle group is identified, and an alternative muscle group path is retrieved from the topological structure to insert an auxiliary exercise action. The present application realizes personalized dynamic adjustment, improves exercise pertinence and effectiveness, and effectively prevents sarcopenia.

Description

Technical Field

[0001] This invention relates to the field of sports science and technology, and in particular to a method and system for preventing sarcopenia in the elderly through exercise. Background Technology

[0002] For preventative exercise targeting sarcopenia in the elderly, current technologies widely employ periodic assessment methods based on single-dimensional physiological indicators (such as grip strength, gait speed, or muscle mass measurements). These conventional approaches typically develop exercise plans based on fixed thresholds or universally recommended standards, such as resistance training or aerobic exercise programs assigned to specific groups after basic fitness tests. Some systems use wearable devices to record simple parameters like exercise duration, steps, or heart rate, and then make rough adjustments to the exercise intensity accordingly. This approach lacks the ability to comprehensively analyze the multidimensional biomechanical and neuromuscular control factors involved in the process of muscle function decline.

[0003] The flaw in conventional approaches lies in neglecting the dynamic evolution of functional dependencies between muscle groups. Existing programs typically treat each muscle group as an independent training unit, failing to recognize the cascading effects caused by compensatory imbalances during muscle function decline—that is, the functional decline of a weak muscle group can be transmitted to other related muscle groups through the biomechanical chain, leading to abnormal overall movement patterns. This structural lack of functional dependency makes the selection of intervention targets untargeted, often resulting in the over-strengthening of secondary muscle groups while core problem muscles are not effectively stimulated, or even exacerbating movement pattern disorders.

[0004] Another drawback of conventional approaches lies in the lack of an adaptive closed-loop mechanism based on real-time individual physiological-motor responses during the exercise execution phase. Existing systems mostly rely on fixed template correction prompts or post-exercise assessments, failing to establish a two-way mapping relationship between movement execution quality and physiological load. When older adults experience movement compensation or fatigue accumulation, the system cannot immediately determine the direction and degree of deviation and generate precise movement adjustment instructions. This results in a lag in corrective feedback and makes it difficult to optimize movement strategies in reverse based on changes in current physiological load. Furthermore, due to the lack of automatic identification of functional bottleneck muscle groups during exercise, existing programs cannot retrieve alternative pathways from muscle group compensation relationships and dynamically insert auxiliary exercise content, thus limiting the adaptability and sustainability of the training program. Summary of the Invention

[0005] The present invention provides a method and system for preventing sarcopenia in the elderly, which can solve the problems in the prior art.

[0006] A first aspect of the present invention provides a method for preventing sarcopenia in the elderly through exercise, comprising:

[0007] Temporal correlation analysis was performed on multidimensional physical condition data of the target elderly population to extract feature vectors of muscle function degeneration process;

[0008] Based on the feature vector of the muscle function degeneration process, a muscle group function dependency topology is constructed. The cascaded influence nodes and functional dependency paths between nodes in the muscle group function dependency topology are identified, and the set of target intervention muscle groups with cascaded influence effects is determined.

[0009] A tiered, progressive exercise program is generated for the target muscle group set. During the execution of the tiered, progressive exercise program by the target elderly person, real-time execution data is collected to establish a two-way mapping relationship between the quality of movement execution and physiological load response.

[0010] The real-time execution data is input into the bidirectional mapping relationship for forward prediction and backward calculation to obtain the deviation direction and deviation magnitude of the current action execution quality. Based on the deviation direction and deviation magnitude, an instant correction instruction containing an action adjustment vector is generated and fed back to the target elderly person in real time through the interactive interface.

[0011] Based on the real-time execution data, identify the muscle groups that are experiencing functional bottlenecks during the current training phase.

[0012] The alternative muscle group paths that have a functional compensatory relationship with the functional bottleneck muscle group are retrieved from the muscle group function-dependent topology, and the auxiliary exercise movements corresponding to the alternative muscle group paths are inserted into the layered progressive exercise program.

[0013] A time-series correlation analysis was performed on the multidimensional physical condition data of the target elderly population to extract feature vectors of muscle function degeneration process, including:

[0014] A multi-scale time window sequence is constructed for the multi-dimensional physical state data of the target elderly. Instantaneous change features, stage evolution features and cumulative degeneration features of the physical state data are extracted in each time window. The stage inflection points of muscle function degeneration at different time scales are identified by the feature transfer relationship across windows.

[0015] Based on the aforementioned stage turning point, the time series is divided into multiple degeneration stage intervals. A dynamic correlation matrix between body parts is constructed in each degeneration stage interval. The dynamic correlation matrix quantifies the mutual influence of functional decline of body parts in the current degeneration stage.

[0016] By comparing the differences in the dynamic correlation matrix across different degradation stage intervals, the evolution trajectory features of functional dependencies between body parts as degradation progresses are extracted.

[0017] The instantaneous change features, stage evolution features, and cumulative degradation features extracted from the multi-scale time window sequence are aggregated with the dynamic correlation matrix and evolution trajectory features of different degradation stage intervals at multiple levels. By quantifying the functional decline acceleration of each body part in different degradation stages and the transformation rate of functional dependence between body parts, a feature vector of the muscle function degradation process containing temporal evolution dimension and cross-part correlation dimension is generated.

[0018] Based on the feature vector of the muscle function degeneration process, a muscle group function dependency topology is constructed. The cascading influence nodes and functional dependency paths between nodes in the muscle group function dependency topology are identified. The set of target intervention muscle groups with cascading influence effects is determined, including:

[0019] From the feature vector of the muscle function degeneration process, feature components containing temporal evolution dimension and cross-site association dimension are extracted. The muscle groups corresponding to each body part are used as topological nodes, and the functional dependencies between body parts represented in the feature components are used as topological edges to construct an initial muscle group functional dependency topological structure.

[0020] The initial muscle group function-dependent topology is directionally labeled and intensity quantized. The directional labeling is determined based on the dominance order of the driving muscle group and the responding muscle group, and the intensity quantization is determined based on the influence coefficient of the functional state change of the driving muscle group on the functional state change of the responding muscle group. The initial muscle group function-dependent topology is transformed into a weighted directed graph structure.

[0021] In the weighted directed graph structure, based on the multi-hop path propagation attenuation coefficient of each topological node from the current node to other nodes along the topological edge, the cascaded influence node muscle group that can influence multiple downstream nodes through a finite number of path hops is identified.

[0022] Extract the functionally dependent paths from the currently described cascaded influence node muscle groups to each downstream node, and calculate the cumulative conduction gain of each functionally dependent path;

[0023] The cumulative conduction gain of the functionally dependent path from the current cascaded affected muscle group to all downstream nodes is summed to obtain the total conduction gain contribution value of each cascaded affected muscle group. The target intervention muscle group set is selected based on the total conduction gain contribution value.

[0024] A tiered, progressive exercise program is generated for the target muscle group set. During the execution of the tiered, progressive exercise program by the target elderly individual, real-time execution data is collected, and a bidirectional mapping relationship between movement execution quality and physiological load response is established, including:

[0025] Extract the topological hierarchy of each muscle group in the target intervention muscle group set;

[0026] The layered and progressive training program divides the training movements into a basic stability layer, a functional linkage layer, and a comprehensive coordination layer.

[0027] Based on the aforementioned topological hierarchy, muscle groups located at the starting position are assigned to the basic stabilization layer, and single-joint isolated activation exercises are generated based on the characteristics of the driving muscle groups; muscle groups located in the middle position that receive transmission from multiple pathways are assigned to the functional linkage layer, and cross-joint joint recruitment exercises are generated based on the characteristics of multi-path convergence; muscle groups located at the end position that integrate multi-muscle group synergy are assigned to the comprehensive coordination layer, and whole-body integrated exercises are generated based on the characteristics of multi-muscle group synergy.

[0028] The motion trajectory data and physiological monitoring data during the execution of the layered progressive training program are collected as the real-time execution data, and the motion execution quality characteristics and physiological load response characteristics are extracted respectively.

[0029] A bidirectional mapping relationship is established between the action execution quality characteristics and the physiological load response characteristics, including a forward prediction mapping and a reverse correction mapping. The forward prediction mapping predicts the expected physiological load response characteristics based on the action execution quality characteristics, and the reverse correction mapping adjusts the target range of the action execution quality characteristics in reverse based on the deviation between the actual physiological load response characteristics and the expected physiological load response characteristics.

[0030] The real-time execution data is input into the bidirectional mapping relationship for forward prediction and backward calculation to obtain the deviation direction and magnitude of the current action execution quality. Based on the deviation direction and magnitude, an immediate correction instruction containing an action adjustment vector is generated and fed back to the target elderly person in real time through an interactive interface, including:

[0031] Based on the real-time execution data, the physiological load response characteristics are compared with the expected physiological load response characteristics calculated by the positive prediction mapping in the bidirectional mapping relationship to obtain the physiological load response deviation.

[0032] The physiological load response deviation is input into the reverse correction mapping in the bidirectional mapping relationship. The contribution of each dimension of the action execution quality feature in the reverse correction mapping to the physiological load response deviation is calculated. The action execution quality feature dimension with the largest contribution is selected as the deviation direction. The magnitude of the physiological load response deviation is extracted as the deviation amplitude.

[0033] Based on the deviation direction, a preset mapping table of action execution quality feature dimensions and body motion space parameters is queried. The deviation direction is mapped to the specific dimension of the body motion space parameters as the spatial dimension of action adjustment. The action adjustment trend is determined according to the positive or negative value of the deviation amplitude. The degree of quantification of action adjustment is determined according to the magnitude of the deviation amplitude.

[0034] The spatial dimension, the motion adjustment trend, and the quantification level are combined to generate a motion adjustment vector, and the motion adjustment vector is converted into an instant correction instruction containing motion correction indications;

[0035] The instant correction instructions are fed back to the target elderly person in real time through an interactive interface that includes visual and voice feedback channels.

[0036] Based on the real-time execution data, the muscle groups that experience functional bottlenecks during the current training phase include:

[0037] Extract activation intensity data and fatigue accumulation data of each muscle group during the current training phase from the real-time execution data;

[0038] For each muscle group in the target intervention muscle group set, the deviation of the activation intensity data from the preset benchmark activation intensity is calculated, the degree of exceeding the fatigue accumulation data from the preset fatigue threshold is calculated, and the deviation and the degree of exceeding the threshold are weighted and fused to obtain the functional limitation score of each muscle group.

[0039] Construct a collaborative dependency graph among the muscle groups in the target intervention muscle group set, count the number of downstream dependent muscle groups of each muscle group node in the collaborative dependency graph, select muscle group nodes whose number of downstream dependent muscle groups exceeds a preset downstream dependency number threshold as high-impact node muscle groups, and multiply the functional limitation score of the high-impact node muscle group with the number of downstream muscle groups constrained by the high-impact node muscle group to obtain the bottleneck impact index.

[0040] All muscle groups in the target intervention muscle group set are sorted in descending order according to the bottleneck impact index, and the muscle groups whose bottleneck impact index ranks in the top preset position are selected as the functional bottleneck muscle groups.

[0041] Retrieving alternative muscle group pathways from the muscle group function-dependent topology that have a functional compensatory relationship with the functional bottleneck muscle group, and inserting the auxiliary exercise movements corresponding to the alternative muscle group pathways into the layered progressive exercise program includes:

[0042] Based on the node position of the functional bottleneck muscle group in the muscle group function-dependent topology, the set of precursor muscle groups that have a direct functional input relationship with the functional bottleneck muscle group are traced upstream, and the set of successor muscle groups that are directly constrained by the functional bottleneck muscle group are traced downstream.

[0043] In the muscle group function-dependent topology, search for bypass muscle group nodes that bypass functional bottlenecks and connect the precursor muscle group set and the successor muscle group set. Quantitatively evaluate the functional compensation ability of each muscle group node in the bypass muscle group chain, and select the bypass muscle group chain whose functional compensation ability meets the preset compensation ability threshold as the alternative muscle group path.

[0044] The set of exercise movements corresponding to each muscle group node in the alternative muscle group path is retrieved from the preset muscle group and exercise movement mapping library. The execution order of the exercise movement set is determined according to the order of each muscle group node in the alternative muscle group path in the bypass muscle group chain, and an auxiliary exercise movement sequence for the alternative muscle group path is generated.

[0045] Analyze the target muscle group coverage of each training stage in the layered progressive training program, and identify the target training stage with the highest functional correlation to the functional bottleneck muscle group.

[0046] The auxiliary exercise sequence is inserted into the target exercise phase, replacing the main exercise movements corresponding to the functional bottleneck muscle groups.

[0047] A second aspect of the present invention provides an exercise system for preventing sarcopenia in the elderly, comprising:

[0048] The feature extraction unit is used to perform time-series correlation analysis on the multi-dimensional physical state data of the target elderly and extract the feature vector of muscle function degeneration process.

[0049] The cascade identification unit is used to construct a muscle group function dependency topology based on the feature vector of the muscle function degeneration process, identify the cascade influence nodes of the muscle group function dependency topology and the functional dependency path between the nodes, and determine the set of target intervention muscle groups with cascade influence effect.

[0050] The program generation unit is used to generate a hierarchical progressive exercise program for the target intervention muscle group set. During the execution of the hierarchical progressive exercise program by the target elderly, real-time execution data is collected to establish a two-way mapping relationship between the quality of movement execution and physiological load response.

[0051] The instant correction unit is used to input the real-time execution data into the bidirectional mapping relationship to perform forward prediction and backward calculation respectively, to obtain the deviation direction and deviation magnitude of the current action execution quality, and to generate an instant correction instruction containing an action adjustment vector based on the deviation direction and the deviation magnitude and to provide real-time feedback to the target elderly person through the interactive interface.

[0052] The bottleneck identification unit is used to identify the functional bottleneck muscle groups that appear in the current training phase based on the real-time execution data.

[0053] The program adjustment unit is used to retrieve alternative muscle group paths that have a functional compensatory relationship with the functional bottleneck muscle group from the muscle group function-dependent topology, and insert the auxiliary exercise movements corresponding to the alternative muscle group paths into the layered progressive exercise program.

[0054] A third aspect of the present invention provides an electronic device, comprising:

[0055] processor;

[0056] Memory used to store processor-executable instructions;

[0057] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0058] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0059] The beneficial effects of this application are as follows:

[0060] Preliminary temporal correlation analysis of multi-dimensional physical condition data enables early detection of subtle trends in muscle function decline and extraction of feature vectors representing the degenerative process. The constructed muscle group function-dependent topology accurately identifies key muscle groups with cascading effects and their functional dependency pathways, pinpointing the target muscle groups requiring priority intervention. This avoids the blindness and lag of traditional approaches, allowing preventative interventions to directly target core weaknesses, significantly improving the targeting and timeliness of interventions. It provides a scientific basis for tiered, progressive training programs, delaying the progression of sarcopenia from its source.

[0061] The exercise program establishes a two-way mapping relationship between movement execution quality and physiological load response. It performs forward prediction and backward calculation on real-time collected execution data, accurately assessing the direction and magnitude of current movement deviations. It then instantly generates corrective instructions containing movement adjustment vectors and provides real-time feedback through an interactive interface. This instant error correction mechanism ensures the accuracy and appropriateness of each exercise movement, effectively preventing muscle damage caused by compensatory movements or excessive load. This significantly improves the safety and effectiveness of exercise, enhancing older adults' trust in the exercise system and their adherence to continued participation.

[0062] The system can dynamically identify functional bottleneck muscle groups during the current training phase. It quickly retrieves alternative muscle pathways with compensatory relationships to the bottleneck muscles from the muscle group function dependency topology, and flexibly inserts corresponding auxiliary exercises into a tiered, progressive training program. This adaptive adjustment capability maintains the continuity and integrity of the training plan, preventing training interruptions or stagnation due to bottleneck muscle group limitations. Simultaneously, it promotes synergistic compensation and functional balance among muscle groups, optimizes training load distribution, and ensures that the training program always aligns with the real-time changes in the elderly's physical function. This enables long-term, sustainable, personalized, and precise training, delaying muscle function decline, reducing the risk of falls and fractures, and improving quality of life. Attached Figure Description

[0063] Figure 1 A flowchart illustrating exercise methods for preventing sarcopenia in the elderly;

[0064] Figure 2 A flowchart illustrating the process of identifying the target muscle group set for intervention. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0067] Figure 1 This is a flowchart illustrating an exercise method for preventing sarcopenia in the elderly according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0068] Temporal correlation analysis was performed on multidimensional physical condition data of the target elderly population to extract feature vectors of muscle function degeneration process;

[0069] Based on the feature vector of the muscle function degeneration process, a muscle group function dependency topology is constructed. The cascaded influence nodes and functional dependency paths between nodes in the muscle group function dependency topology are identified, and the set of target intervention muscle groups with cascaded influence effects is determined.

[0070] A tiered, progressive exercise program is generated for the target muscle group set. During the execution of the tiered, progressive exercise program by the target elderly person, real-time execution data is collected to establish a two-way mapping relationship between the quality of movement execution and physiological load response.

[0071] The real-time execution data is input into the bidirectional mapping relationship for forward prediction and backward calculation to obtain the deviation direction and deviation magnitude of the current action execution quality. Based on the deviation direction and deviation magnitude, an instant correction instruction containing an action adjustment vector is generated and fed back to the target elderly person in real time through the interactive interface.

[0072] Based on the real-time execution data, identify the muscle groups that are experiencing functional bottlenecks during the current training phase.

[0073] The alternative muscle group paths that have a functional compensatory relationship with the functional bottleneck muscle group are retrieved from the muscle group function-dependent topology, and the auxiliary exercise movements corresponding to the alternative muscle group paths are inserted into the layered progressive exercise program.

[0074] In one optional implementation, a time-series correlation analysis is performed on the multi-dimensional physical state data of the target elderly person to extract a feature vector of muscle function degeneration process, including:

[0075] A multi-scale time window sequence is constructed for the multi-dimensional physical state data of the target elderly. Instantaneous change features, stage evolution features and cumulative degeneration features of the physical state data are extracted in each time window. The stage inflection points of muscle function degeneration at different time scales are identified by the feature transfer relationship across windows.

[0076] Based on the aforementioned stage turning point, the time series is divided into multiple degeneration stage intervals. A dynamic correlation matrix between body parts is constructed in each degeneration stage interval. The dynamic correlation matrix quantifies the mutual influence of functional decline of body parts in the current degeneration stage.

[0077] By comparing the differences in the dynamic correlation matrix across different degradation stage intervals, the evolution trajectory features of functional dependencies between body parts as degradation progresses are extracted.

[0078] The instantaneous change features, stage evolution features, and cumulative degradation features extracted from the multi-scale time window sequence are aggregated with the dynamic correlation matrix and evolution trajectory features of different degradation stage intervals at multiple levels. By quantifying the functional decline acceleration of each body part in different degradation stages and the transformation rate of functional dependence between body parts, a feature vector of the muscle function degradation process containing temporal evolution dimension and cross-part correlation dimension is generated.

[0079] To capture the evolution of muscle function decline at different time granularities, a multi-dimensional physical condition data set was constructed for the target elderly population. Data sources included surface electromyography (EMG), joint angular velocity, gait pressure distribution, grip strength measurement, balance swing amplitude, and biochemical indicators such as serum creatine kinase. These data were streamed in parallel across multiple channels. Specifically, a multi-scale time window sequence was constructed for the aforementioned multi-dimensional data, comprising three levels: short window, medium window, and long window. The short window covers a single action execution cycle (approximately a few seconds), the medium window covers a single exercise session (approximately tens of minutes), and the long window covers continuous exercise cycles (several weeks to several months). The three levels of windows were sampled by sliding along the time axis, with a certain proportion of overlap between adjacent windows to avoid truncation distortion of features at window boundaries.

[0080] Within a short window, instantaneous changes in body state data are extracted, including peak amplitude of electromyographic signals, instantaneous fluctuations in discharge frequency, abrupt slopes of joint angular velocity changes, and instantaneous shifts in the gait stress center. These instantaneous features reflect the immediate response capability of neuromuscular activation during a single movement execution and are highly sensitive to the identification of early fatigue accumulation and movement compensation patterns. Within a medium window, phased evolution features are extracted, including the slope of the decay trend of activation amplitude of each muscle group within a single training session, the phased narrowing of joint range of motion, and the rate of change of the standard deviation of balance swing within the session. These features describe the phased changes in muscle function over time during a single training session. Within a long window, cumulative degenerative features are extracted, including the decrease in maximum isometric contractile force across cycles, the long-term slowing trend of gait speed, and the monthly decline slope of grip strength. These features characterize the irreversible cumulative degenerative process of muscle function over a longer time span.

[0081] The cross-window feature transfer relationship is achieved by establishing feature mappings between adjacent scale windows. Specifically, the instantaneous feature sequences extracted from the short window are statistically aggregated at the medium window granularity, and their mean, variance, and extreme value distributions are calculated to form a supplementary description of the medium window features. Similarly, the phased features of the medium window are trend-fitted at the long window granularity to obtain the evolution slope across the course. By observing whether structural abrupt changes occur in this cross-scale feature transfer (e.g., a sudden and significant increase in the variance of instantaneous fatigue features within a certain medium window, or a breakpoint in the monthly decline slope within a long window), the phased turning points of muscle function degeneration at different time scales are identified. The determination of turning points uses a change point detection algorithm, which performs segmented fitting on the cross-scale aggregated feature sequences. When the difference in fitting parameters between adjacent segments exceeds a preset threshold, the corresponding time position is marked as a degeneration stage turning point.

[0082] Based on the identified turning points, the complete time series is divided into multiple degradation stage intervals, each representing a relatively stable state of functional degradation. Within each degradation stage interval, a dynamic correlation matrix between body parts is constructed. Let the number of body parts involved in each interval be . Then the dynamic correlation matrix for The real-valued matrix, where the first... Line 1 Column elements Indicates the first Within each stage of degeneration, body parts Functional decline in body parts The influence intensity of functional state is obtained by calculating the mutual information or time-delay cross-correlation coefficient of the corresponding feature sequences of two sites within that interval. The construction method of the dynamic correlation matrix ensures the quantitative expression of the functional dependencies between sites within different degradation stages, rather than static and fixed anatomical correlations, and can reflect the dynamic coupling law that changes with the degradation process.

[0083] By comparing the differences between the dynamic correlation matrices of different degradation stage intervals, the evolution trajectory features of functional dependencies between body parts as degradation progresses are extracted. The difference matrix is ​​obtained by subtracting the dynamic correlation matrices of two adjacent degradation stage intervals. The parts corresponding to elements with larger absolute values ​​in the difference matrix indicate that their functional dependencies have changed significantly between the two stages, and these parts are marked as active nodes in the relationship evolution. Furthermore, the difference matrices of all adjacent stages are sequentially arranged, and the rate and direction of change of each part pair on the evolution trajectory are calculated to form a trajectory feature vector describing the evolution of functional dependencies as the degradation process progresses. For parts pairs with continuously strengthening functional dependencies, it means that the compensatory coupling between the two deepens as degradation progresses; for parts pairs with gradually weakening functional dependencies, it means that degradation has led to a decrease in their co-activation ability.

[0084] After completing the feature extraction at each level, the instantaneous change features, stage evolution features, and cumulative degradation features extracted from the multi-scale time window sequence are aggregated with the dynamic correlation matrix and evolution trajectory features of different degradation stage intervals at multiple levels. The aggregation process is divided into two dimensions: the temporal evolution dimension and the cross-part correlation dimension. In the temporal evolution dimension, the features of each scale window are concatenated in chronological order, and the functional decline acceleration of each body part in different degradation stages is calculated, that is, the difference in the rate of decline of the same functional index within adjacent degradation stage intervals. A positive acceleration indicates that the degradation rate is accelerating, and a negative acceleration indicates that the degradation rate is leveling off. In the cross-part correlation dimension, the dynamic correlation matrix of each degradation stage is flattened and concatenated with the evolution trajectory features, and the transformation rate of functional dependence between body parts is calculated, that is, the slope of the change of the absolute value of each element in the difference matrix as the stage progresses.

[0085] The aforementioned temporal evolution dimension features and cross-site correlation dimension features are integrated into a unified feature vector through feature concatenation and weighted fusion. This vector is a muscle function degeneration process feature vector that includes both temporal evolution and cross-site correlation dimensions. This feature vector not only encodes the absolute degree of degeneration of each muscle group at different time scales, but also encodes the dynamic evolution of functional dependencies between muscle groups as the degeneration process progresses. It provides a multi-dimensional, high-information input representation for subsequent construction of muscle group function dependency topology and identification of cascading influence nodes, enabling intervention programs to accurately correspond to individualized degeneration process patterns, rather than relying on static group average parameters.

[0086] In one optional implementation, a muscle group function dependency topology is constructed based on the feature vector of the muscle function degeneration process. The cascading influence nodes and functional dependency paths between nodes in the muscle group function dependency topology are identified. The set of target intervention muscle groups with cascading influence effects is determined, including:

[0087] From the feature vector of the muscle function degeneration process, feature components containing temporal evolution dimension and cross-site association dimension are extracted. The muscle groups corresponding to each body part are used as topological nodes, and the functional dependencies between body parts represented in the feature components are used as topological edges to construct an initial muscle group functional dependency topological structure.

[0088] The initial muscle group function-dependent topology is directionally labeled and intensity quantized. The directional labeling is determined based on the dominance order of the driving muscle group and the responding muscle group, and the intensity quantization is determined based on the influence coefficient of the functional state change of the driving muscle group on the functional state change of the responding muscle group. The initial muscle group function-dependent topology is transformed into a weighted directed graph structure.

[0089] In the weighted directed graph structure, based on the multi-hop path propagation attenuation coefficient of each topological node from the current node to other nodes along the topological edge, the cascaded influence node muscle group that can influence multiple downstream nodes through a finite number of path hops is identified.

[0090] Extract the functionally dependent paths from the currently described cascaded influence node muscle groups to each downstream node, and calculate the cumulative conduction gain of each functionally dependent path;

[0091] The cumulative conduction gain of the functionally dependent path from the current cascaded affected muscle group to all downstream nodes is summed to obtain the total conduction gain contribution value of each cascaded affected muscle group. The target intervention muscle group set is selected based on the total conduction gain contribution value.

[0092] like Figure 2 As shown, the method includes:

[0093] Feature components are extracted from the feature vector of muscle function degeneration process, distinguishing two dimensions: temporal evolution dimension and cross-site correlation dimension. The feature components of the temporal evolution dimension describe the changing trend of muscle function indicators of a certain body part as the degeneration stage progresses, such as the decline slope of peak muscle strength and the increase in muscle activation delay. The feature components of the cross-site correlation dimension describe the synergistic or inhibitory relationships in functional states between different body parts, such as the statistical correlation strength between decreased hip flexor function and compensatory activation of knee extensors. By abstracting the muscle groups corresponding to each body part as topological nodes and the functional dependencies represented in the cross-site correlation dimension as topological edges, an initial muscle group functional dependency topological structure can be constructed. This initial structure only reflects whether functional correlations exist between muscle groups and does not yet contain directional and strength information; it is an unweighted and undirected initial graph.

[0094] Based on the initial topology, each topological edge is directionally labeled and its intensity is quantized, transforming it into a weighted directed graph structure. Directional labeling is determined based on the order of control between driving and responding muscle groups: driving muscle groups are those that initiate functional changes in the kinetic chain and exert a traction effect on downstream muscle groups, while responding muscle groups receive upstream driving signals and produce functional responses. Specifically, the causal direction between driving and responding can be determined by analyzing the temporal sequence of electromyographic signals, the direction of joint torque transmission, and the neural control pathways in multi-dimensional body state data. This assigns a directional attribute to the topological edge, meaning the edge's direction points from the driving muscle group node to the responding muscle group node. Intensity quantization is determined based on the influence coefficient of the driving muscle group's functional state change on the responding muscle group's functional state change: let the driving muscle group node be... Response muscle group nodes are The edge weight between the two Indicates when node When the functional state of a node changes by a unit, The corresponding change in the functional state can be obtained by performing regression analysis or conditional mutual information estimation on historical time-series data. After the above processing, the initial topology is transformed into a weighted directed graph, where each directed edge carries direction information and quantization weight.

[0095] In weighted directed graph structures, the core basis for identifying cascading influence nodes and muscle groups is the multi-hop path conduction attenuation coefficient. From a certain node... Start, and pass along the directed edge. Jump path to reach node Conductivity attenuation coefficient This reflects the degree of attenuation of the functionally influential signal after passing through multiple intermediate nodes. Let the weights of each hop edge on the path be as follows: , … ,in If a node is an intermediate node on the path, then the conduction attenuation coefficient of the path is the product of the weights of each jump edge, i.e. In practical applications, a maximum number of jumps should be set as an upper limit. (For example, taking 3 or 4 hops) to ensure that path search is computationally feasible while conforming to the physiologically meaningful propagation range of functional impact. For each candidate node, its range within a certain limit is counted. The number of downstream nodes that can be affected within the jump range, and whether the conduction attenuation coefficient of each path exceeds the preset effective conduction threshold. When the number of downstream nodes that a node can influence through a finite number of hops exceeds a set threshold, and the conduction attenuation coefficient of each path is not lower than a certain threshold. At that time, the node is identified as a cascading influence node muscle group. This identification mechanism can effectively screen out muscle groups that play a pivotal role in the entire muscle group functional network. The functional state of these muscle groups has an amplifying effect on overall motor function.

[0096] After extracting the functionally dependent paths from each cascade-affected node muscle group to all its downstream nodes, the cumulative conduction gain is calculated for each path. Let a certain path originate from a cascade-affected node... to downstream nodes The functional dependency path is The cumulative conduction gain of this path Defined as the product of the weights of all directed edges on the path, i.e. ,in Representing a path The directed edge on, This is the weight of the edge. When from node To the node When multiple paths exist, the maximum value of the cumulative conduction gain of all paths is taken as the node. For nodes Effective conduction gain This approach reflects the contribution of the optimal conduction pathway in the propagation of functional effects. This method aligns with physiological reality, as the neuromuscular system tends to select the most efficient conduction pathway during functional compensation.

[0097] After calculating the cumulative conduction gain for each path, the muscle groups at each cascaded influence node are analyzed. Total conduction gain contribution Perform the calculation. Define as a node The sum of the effective conduction gains for all its reachable downstream nodes, i.e. ,in Indicates from node Departure The set of all downstream nodes reachable within the jump range. The larger the value, the wider the influence of the node muscle group on the overall muscle group functional network and the higher the transmission intensity. The improvement of its functional state can drive the functional improvement of more downstream muscle groups, and therefore it has a higher intervention priority.

[0098] Based on the total conduction gain contribution value All cascading muscle groups are ranked in descending order, and the top-ranked muscle groups are selected to form the target intervention muscle group set. The specific number of selected groups can be dynamically adjusted based on the individual elderly person's overall functional status, exercise tolerance, and intervention period. For example, for elderly people with poor functional status or low exercise tolerance, priority can be given to selecting… The top three muscle groups with the highest activity levels are selected as the target intervention set to concentrate training resources and maximize the cascading improvement effect. For older adults with relatively good functional status, the range of target intervention muscle groups can be appropriately expanded to include more muscle group nodes with moderate conduction gain contributions, in order to achieve a more comprehensive improvement in muscle function. Through this systematic approach based on topology and quantified gain, the most valuable target muscle groups can be accurately located from the complex muscle group function dependency network, providing a scientific basis for the generation of subsequent hierarchical and progressive training programs.

[0099] In one optional implementation, a tiered, progressive exercise program is generated for the target muscle group set. During the execution of the tiered, progressive exercise program by the target elderly individual, real-time execution data is collected, and a bidirectional mapping relationship is established between the quality of movement execution and the physiological load response, including:

[0100] Extract the topological hierarchy of each muscle group in the target intervention muscle group set;

[0101] The layered and progressive training program divides the training movements into a basic stability layer, a functional linkage layer, and a comprehensive coordination layer.

[0102] Based on the aforementioned topological hierarchy, muscle groups located at the starting position are assigned to the basic stability layer, and single-joint isolated activation exercise movements are generated based on the characteristics of the driving muscle groups.

[0103] The muscle groups located in the middle position and receiving transmission from multiple paths are assigned to the functional linkage layer, and cross-joint joint recruitment exercise movements are generated based on the multi-path convergence feature.

[0104] The muscle groups located at the end position and integrating multi-muscle group synergy are assigned to the comprehensive coordination layer, and whole-body integrated exercise movements are generated based on the multi-muscle group synergy characteristics.

[0105] The motion trajectory data and physiological monitoring data during the execution of the layered progressive training program are collected as the real-time execution data, and the motion execution quality characteristics and physiological load response characteristics are extracted respectively.

[0106] A bidirectional mapping relationship is established between the action execution quality characteristics and the physiological load response characteristics, including a forward prediction mapping and a reverse correction mapping. The forward prediction mapping predicts the expected physiological load response characteristics based on the action execution quality characteristics, and the reverse correction mapping adjusts the target range of the action execution quality characteristics in reverse based on the deviation between the actual physiological load response characteristics and the expected physiological load response characteristics.

[0107] In the constructed muscle group function-dependent topology, each muscle group node is assigned a different hierarchical label based on its in-degree, out-degree, and path convergence degree in the directed graph. Muscle groups with zero in-degree or only a single input path are typically located at the beginning of the topology, and their functional state changes drive multiple downstream muscle groups. Muscle groups in the middle levels simultaneously receive functional transmissions from multiple upstream paths and continue to transmit them downstream. Muscle groups at the terminal level integrate collaborative inputs from multiple upstream muscle groups and are responsible for completing the final motor function output. By performing hierarchical analysis of the topology, the positional attributes of each muscle group in the function-dependent chain can be clarified, providing a structured basis for the subsequent allocation of movement types.

[0108] The tiered, progressive training program divides all exercises into three levels: a foundational stability level, a functional linkage level, and a comprehensive coordination level. These three levels increase sequentially in activation range, movement complexity, and physiological load intensity, forming a progressive training path from isolated local activation to whole-body integration and coordination. This tiered design fully considers the uneven nature of muscle function decline in older adults, avoiding excessive synergistic loads on weak muscle groups in the early stages, while also providing a sufficient foundation for functional integration in later stages.

[0109] Muscle groups located at the initiation of the topological structure are assigned to the basal stabilization layer. These muscle groups typically have high excursion, and their functional state directly affects the activation efficiency of multiple downstream muscle groups. For these muscle groups, single-joint isolation activation exercises are generated based on the characteristics of the driving muscle group. This means that the contraction range of the target muscle group is isolated as much as possible in the exercise design to reduce the compensatory involvement of adjacent muscle groups. For example, when the quadriceps is the driving muscle group, a seated knee extension exercise can be designed. By fixing the hip joint angle and limiting trunk sway, the knee extension movement is completed solely by the quadriceps. The load intensity of the exercise is determined based on the baseline muscle strength assessment results of the target elderly person. In the initial stage, the core objectives are low impedance and high control precision, with a focus on establishing the neural activation pathways of the target muscle group.

[0110] Muscle groups located in the intermediate topological level and receiving transmission from multiple pathways are assigned to the functional linkage layer. These muscle groups are at the convergence node position of the function-dependent network, and their activation quality is jointly influenced by multiple upstream driving muscle groups. Based on the multi-path convergence characteristics, cross-joint joint recruitment exercises are generated for these muscle groups, that is, simultaneously activating the target muscle group and its main upstream driving muscle groups in a single movement, simulating the collaborative activation mode of function-dependent pathways in real movement. Taking the gluteus medius as an example, if its topological in-degree comes from dual-path transmission from the quadratus lumborum and transverse abdominis, the corresponding linkage layer movement can be designed as a compound movement of lateral hip abduction combined with abdominal stability control, requiring the target elderly person to complete hip abduction while maintaining core stability, thereby simultaneously activating three related muscle groups. The movement design of the functional linkage layer needs to meet the requirements of multi-muscle group coordination in both the spatial and temporal dimensions of the movement trajectory, thus correspondingly increasing the requirements for the monitoring accuracy of movement execution quality.

[0111] Muscle groups located at the terminal topological position and integrating multi-muscle synergy are assigned to the comprehensive coordination layer. The function of these muscle groups depends on the coordinated operation of the entire functional dependence network, and their activation quality reflects the overall level of motor function integration in older adults. Based on the characteristics of multi-muscle synergy, whole-body integrative exercises are generated for these muscle groups, with exercise design encompassing elements such as multi-joint linkage, center of gravity transfer control, and dynamic balance maintenance. Typical exercise forms include functional stand-up-sit transitions, single-leg dynamic balance support, and gait simulation training. These exercises activate the terminal integrative muscle groups while also comprehensively mobilizing the muscle groups in the basic stability layer and the functional linkage layer, achieving functional continuity between the three levels.

[0112] During the implementation of a tiered and progressive exercise program for the target elderly population, two types of real-time execution data were collected: movement trajectory data and physiological monitoring data. Movement trajectory data was acquired using joint angle sensors, inertial measurement units, or visual posture estimation models, recording parameters such as angular velocity, angular acceleration, movement trajectory deviation, and joint range of motion for each joint during movement. Physiological monitoring data was collected using surface electromyography (EMG) sensors, heart rate monitoring devices, and blood oxygen sensors, recording indicators such as EMG activation amplitude, activation sequence, heart rate variability, and subjective fatigue perception scores for the target muscle groups. Movement execution quality characteristics, including joint angle arrival accuracy, movement rhythm stability, and movement trajectory smoothness, were extracted from the movement trajectory data; physiological load response characteristics, including root mean square EMG values, heart rate reserve utilization, and muscle activation synergy index, were extracted from the physiological monitoring data.

[0113] A bidirectional mapping relationship is established between action execution quality features and physiological load response features, comprising two directions: forward prediction mapping and backward correction mapping. Forward prediction mapping uses the action execution quality feature vector as input to predict the expected physiological load response features under the current action execution state. This mapping is established through a regression model, with training data derived from paired records of action quality and physiological response in historical execution samples. Taking joint angle attainment accuracy as an example, if the target elderly person stably reaches the target range for knee extension angle during a seated knee extension movement, the forward prediction mapping will output the expected range of the corresponding quadriceps root mean square electromyography (EMG) value. Backward correction mapping uses the actually collected physiological load response features as input, compares them with the expected physiological load response features obtained from forward prediction, calculates the deviation between the two, and maps the deviation information backward into the action execution quality feature space to infer which quality dimensions have deviated during the current action execution, as well as the direction and magnitude of the deviation. For example, if the actual EMG activation amplitude is significantly lower than the expected value, the backward correction mapping will infer that the target elderly person has quality problems such as insufficient joint angle or excessively fast movement rhythm, and adjust the target range of the action execution quality features accordingly, providing a basis for the generation of subsequent real-time correction instructions. The establishment of a two-way mapping relationship enables a closed-loop monitoring mechanism between the quality of movement execution and the physiological load response. This mechanism can both predict the risk of physiological load based on the quality of movement and locate the problem of movement quality based on abnormal physiological response, thereby achieving refined and personalized management of the training process.

[0114] In one optional implementation, the real-time execution data is input into the bidirectional mapping relationship for forward prediction and backward calculation to obtain the deviation direction and magnitude of the current action execution quality. Based on the deviation direction and the deviation magnitude, an immediate correction instruction containing an action adjustment vector is generated and fed back to the target elderly person in real time through an interactive interface, including:

[0115] Based on the real-time execution data, the physiological load response characteristics are compared with the expected physiological load response characteristics calculated by the positive prediction mapping in the bidirectional mapping relationship to obtain the physiological load response deviation.

[0116] The physiological load response deviation is input into the reverse correction mapping in the bidirectional mapping relationship. The contribution of each dimension of the action execution quality feature in the reverse correction mapping to the physiological load response deviation is calculated. The action execution quality feature dimension with the largest contribution is selected as the deviation direction. The magnitude of the physiological load response deviation is extracted as the deviation amplitude.

[0117] Based on the deviation direction, a preset mapping table of action execution quality feature dimensions and body motion space parameters is queried. The deviation direction is mapped to the specific dimension of the body motion space parameters as the spatial dimension of action adjustment. The action adjustment trend is determined according to the positive or negative value of the deviation amplitude. The degree of quantification of action adjustment is determined according to the magnitude of the deviation amplitude.

[0118] The spatial dimension, the motion adjustment trend, and the quantification level are combined to generate a motion adjustment vector, and the motion adjustment vector is converted into an instant correction instruction containing motion correction indications;

[0119] The instant correction instructions are fed back to the target elderly person in real time through an interactive interface that includes visual and voice feedback channels.

[0120] During the implementation of a tiered and progressive exercise program for the target elderly population, the continuously collected real-time execution data contains two core types of information: one is the quality characteristics of movement execution, namely kinematic parameters such as limb movement trajectory, joint range of motion, and movement rhythm collected by devices such as inertial sensors and joint angle sensors; the other is the physiological load response characteristics, namely physiological signals such as muscle activation intensity, heart rate variability, and respiratory rate obtained by physiological sensors such as heart rate monitoring and surface electromyography (EMG) signal acquisition. The bidirectional mapping relationship consists of a forward prediction mapping and a backward correction mapping. The forward prediction mapping describes the mapping law from movement execution quality characteristics to physiological load response characteristics, while the backward correction mapping describes how deviations in physiological load response are traced back to various dimensions of movement execution quality characteristics.

[0121] In the forward prediction phase, the currently collected action execution quality features are input into the forward prediction mapping to obtain the expected physiological load response features that should theoretically be generated under this action execution quality. This expected physiological load response feature is then compared dimension-by-dimensionally with the actual physiological load response features collected at the same time, and the difference between the two is calculated to obtain the physiological load response deviation. The physiological load response deviation is a multi-dimensional vector, with each component corresponding to the degree of deviation of a physiological signal. For example, if the expected electromyographic activation intensity is a certain value but the measured value is lower, the corresponding component is negative; if the measured heart rate is higher than expected, the corresponding component is positive. This method of constructing the multi-dimensional deviation can completely preserve the mismatch information of the physiological load response, providing sufficient input for subsequent reverse tracing.

[0122] In the reverse correction phase, the aforementioned physiological load response deviation is input into the reverse correction map. The core function of the reverse correction map is to calculate the contribution of each dimension of the action execution quality characteristics to the current physiological load response deviation. Specifically, for each dimension in the action execution quality characteristic space, the explanatory power of changes in that dimension on the physiological load response deviation is calculated, i.e., to what extent adjustments to that dimension can eliminate the currently observed deviation. The contribution can be expressed as... ,in Dimensional index for action execution quality characteristics. The larger the value, the stronger the explanatory power of that dimension for the current bias. Select the dimension with the largest contribution. The direction of this correction is directed towards the dimension of motion quality that most urgently needs adjustment. Simultaneously, the overall magnitude of the physiological load response deviation is extracted as the deviation amplitude, reflecting the overall degree of deviation between the current execution state and the ideal state.

[0123] After determining the direction of deviation, a pre-defined mapping table between the dimensions of motion execution quality characteristics and the parameters of body movement space is consulted. This mapping table is a pre-established correspondence during the treatment design phase, mapping abstract dimensions of motion execution quality characteristics to specific parameters of body movement space. For example, the dimension of "hip flexion angular velocity" is mapped to the angular velocity parameter of the hip joint in the sagittal plane, and the dimension of "lateral displacement of the center of gravity" is mapped to the lateral displacement parameter of the center of mass in the coronal plane. By consulting the mapping table, the dimension of motion execution quality characteristics indicated by the direction of deviation is converted into a specific dimension of body movement space parameters, serving as the spatial dimension for motion adjustment. This conversion process allows the subsequently generated corrective instructions to be expressed in a spatial movement language that older adults can directly understand and execute, rather than remaining at the level of abstract signal characteristics.

[0124] The determination of the trend and quantification level of movement adjustment depends on the numerical characteristics of the deviation amplitude. The sign of the deviation amplitude determines the direction of the movement adjustment: a positive deviation amplitude indicates that the current execution state exceeds expectations in that spatial dimension and needs to be adjusted towards reduction; a negative deviation amplitude indicates that the current execution state is insufficient in that spatial dimension and needs to be adjusted towards increase. The absolute value of the deviation amplitude determines the quantification level of the movement adjustment: a smaller absolute value corresponds to a slight adjustment instruction; a larger absolute value corresponds to a significant adjustment instruction. The quantification level can be divided into three levels—mild, moderate, and severe—according to preset grading thresholds. Each level corresponds to a different degree of movement correction amplitude description, ensuring that the intensity of the correction instruction matches the actual deviation level and avoiding unnecessary fluctuations in the exercise load for the elderly.

[0125] By combining three elements—spatial dimension, movement adjustment trend, and quantification level—a movement adjustment vector is constructed. This vector is a structured descriptive unit that clearly indicates "in which spatial dimension, in which direction, and by how much," possessing complete operational semantics. Based on this, the movement adjustment vector is converted into an immediate corrective instruction containing movement correction guidelines. The conversion process relies on a pre-defined instruction generation template library, mapping the vectorized adjustment information into natural language descriptions or graphical movement diagrams, making the instructions intuitive, readable, and easy to execute for older adults. For example, when the spatial dimension is knee flexion angle, the adjustment trend is increasing, and the quantification level is moderate, the corresponding generated instruction text would be "Please appropriately increase the knee flexion range," along with a corresponding joint angle diagram or dynamic arrow animation.

[0126] Real-time corrective instructions are provided to the target elderly person through an interactive interface that includes both visual and voice feedback channels. The visual feedback channel uses a screen display to present the adjustment instructions as text descriptions, body posture comparison diagrams, or dynamic trajectory arrows, helping the elderly person intuitively perceive the difference between their current posture and the target posture. The voice feedback channel converts the instruction text into spoken audio, guiding the elderly person to correct their movements with clear and concise conversational language. This is particularly suitable for scenarios where the elderly person's gaze is focused on body movements and they cannot continuously watch the screen. The parallel output of visual and voice channels ensures that the corrective information is effectively received and understood by the elderly person without interrupting the exercise rhythm, thus achieving closed-loop real-time intervention on the quality of movement execution.

[0127] In one optional implementation, identifying the functional bottleneck muscle groups in the current training phase based on the real-time execution data includes:

[0128] Extract activation intensity data and fatigue accumulation data of each muscle group during the current training phase from the real-time execution data;

[0129] For each muscle group in the target intervention muscle group set, the deviation of the activation intensity data from the preset benchmark activation intensity is calculated, the degree of exceeding the fatigue accumulation data from the preset fatigue threshold is calculated, and the deviation and the degree of exceeding the threshold are weighted and fused to obtain the functional limitation score of each muscle group.

[0130] Construct a collaborative dependency graph among the muscle groups in the target intervention muscle group set, count the number of downstream dependent muscle groups of each muscle group node in the collaborative dependency graph, select muscle group nodes whose number of downstream dependent muscle groups exceeds a preset downstream dependency number threshold as high-impact node muscle groups, and multiply the functional limitation score of the high-impact node muscle group with the number of downstream muscle groups constrained by the high-impact node muscle group to obtain the bottleneck impact index.

[0131] All muscle groups in the target intervention muscle group set are sorted in descending order according to the bottleneck impact index, and the muscle groups whose bottleneck impact index ranks in the top preset position are selected as the functional bottleneck muscle groups.

[0132] During the implementation of a tiered, progressive exercise program for the target elderly population, continuously collected real-time execution data includes temporal signals of activation intensity and fatigue accumulation indicators for each muscle group in each exercise phase. Activation intensity data typically originates from surface electromyography (EMG) signal acquisition devices. By filtering, rectifying, and extracting the envelope of the raw EMG signals, the root mean square amplitude value of each muscle group within a single movement cycle is obtained, thereby quantifying the actual activation level of the muscle group. Fatigue accumulation data comprehensively considers the downward shift trend of the median EMG frequency, the attenuation slope of activation intensity during continuous movement, and subjective fatigue scores to form a comprehensive indicator reflecting the current fatigue state of the muscle group. These two types of data together constitute the original input basis for identifying functional bottleneck muscle groups.

[0133] For each muscle group in the target intervention muscle group set, the deviation between its activation intensity data and the preset baseline activation intensity was calculated. The preset baseline activation intensity was determined based on the standard movement specifications for sarcopenia prevention exercises in the elderly, corresponding to the minimum effective activation level that the target muscle group should achieve in this exercise movement. The deviation was calculated by dividing the absolute value of the difference between the actual collected mean activation intensity and the baseline activation intensity by the baseline activation intensity to obtain the normalized relative deviation ratio. The larger this value, the more significant the gap between the current activation level of the muscle group and the target requirement, indicating that the muscle group has functional limitations during exercise. At the same time, the degree of exceeding the preset fatigue threshold was calculated based on the cumulative fatigue data. The degree of exceeding the threshold was defined as the ratio of the portion of the cumulative fatigue index that exceeds the fatigue threshold to the fatigue threshold itself; if the cumulative fatigue index does not exceed the fatigue threshold, the degree of exceeding the threshold is recorded as zero. The preset fatigue threshold was dynamically configured based on the target elderly person's age group, basic physical fitness assessment results, and current exercise stage to ensure that the assessment standards match the individual's actual endurance.

[0134] The deviation from the target and the degree of exceeding the limit are weighted and fused to obtain the functional limitation score for each muscle group. Let the deviation from the target for a certain muscle group be denoted as . The degree of exceeding the limit is The corresponding fusion weights are respectively and The functional limitation score of this muscle group is... satisfy: ,in The muscle group index in the target intervention muscle group set, and All are positive and satisfy The specific values ​​of the two weights were configured based on clinical intervention experience for sarcopenia in the elderly: in the initial stage of the exercise program, insufficient activation has a more prominent early warning significance for functional decline, therefore... Take the larger value; in the later stages of the training program, the risk of compensation due to excessive fatigue accumulation needs more attention, so it is appropriate to increase the value. The percentage of functional limitations. (Rating) The higher the value, the more limited the current functional state of that muscle group, and the more likely it is to become a bottleneck restricting the overall training effect.

[0135] After obtaining the functional limitation scores of each muscle group, a synergistic dependency graph is constructed among the muscle groups in the target intervention muscle group set. This graph uses each muscle group as a node and the synergistic driving or functional dependency relationships between muscle groups during exercise as directed edges, with the edges pointing from the functional output muscle group to the downstream dependent muscle group it drives. The synergistic dependency relationships are established based on the path information identified in the muscle group functional dependency topology, combined with the movement patterns involved in the current exercise stage for subgraph extraction, thus obtaining a local synergistic dependency graph highly relevant to the current exercise stage. In this graph, the number of downstream dependent muscle groups for each muscle group node is counted, i.e., the total number of all downstream nodes reachable from that node along directed edges. A larger number of downstream dependent muscle groups indicates a wider influence of the functional state of that muscle group on other muscle groups in the current exercise stage.

[0136] Muscle groups whose downstream dependency count exceeds a preset downstream dependency threshold are selected as high-influence muscle groups. The preset downstream dependency threshold is typically dynamically set based on the total size of the target intervention muscle group set, for example, as an integer value corresponding to 30% of the total number of muscle groups in the set, thus filtering out muscle groups with high structural centrality in the cooperative dependency network. For each node identified as a high-influence muscle group, its functional limitation score is assigned. The number of downstream muscle groups it regulates Perform a product operation to obtain the bottleneck impact index of this muscle group. ,Right now: Bottleneck Impact Index It captures information from two dimensions: one reflecting the current functional limitation of the muscle group itself, and the other reflecting the spread of its functional limitation in the co-dependency network. If a muscle group has a high degree of functional limitation but a limited downstream influence, or a wide downstream influence but a low degree of functional limitation, it will not generate an extremely high bottleneck influence index. Only muscle groups with both high levels will be preferentially identified as functional bottleneck muscle groups. This design effectively avoids the one-sidedness of a single-dimensional assessment.

[0137] The bottleneck impact index was used to analyze all muscle groups included in the high-impact node range within the target intervention muscle group set. The muscle groups are sorted in descending order, and those ranking at the top of a preset number are selected as the final identified functional bottleneck muscles. The preset number is typically set to 3 to 5, with the specific value adjusted flexibly based on the complexity of the current training phase and the overall functional status of the target elderly individual. For elderly individuals with weaker functional foundations and a higher risk of simultaneous limitation of multiple muscle groups, the preset number can be appropriately increased to ensure more comprehensive coverage of the identified functional bottleneck muscles. For elderly individuals with relatively good functional status, the preset number can be set to a smaller value, concentrating resources on intervening in the most critical bottleneck points.

[0138] The functional bottleneck muscle groups identified through the above-described process can accurately pinpoint the core weaknesses hindering the overall effectiveness of exercise in the current training phase. Compared to traditional methods that rely solely on a single activation intensity or fatigue index, this method combines the functional limitation of a muscle group with its structural position in a co-dependent network, providing not only physiological evidence but also support at the network topology level. The identified functional bottleneck muscle groups serve as core inputs for subsequent alternative muscle pathway retrieval and the insertion of auxiliary exercise movements, driving the dynamic adjustment of a hierarchical, progressive training program. This helps target elderly individuals overcome current functional limitations and maintain the overall effective progress of the training program by activating compensatory alternative muscle pathways.

[0139] In one optional implementation, retrieving alternative muscle group pathways that have a functional compensatory relationship with the functional bottleneck muscle group from the muscle group function-dependent topology, and inserting the auxiliary exercise movements corresponding to the alternative muscle group pathways into the layered progressive exercise program includes:

[0140] Based on the node position of the functional bottleneck muscle group in the muscle group function-dependent topology, the set of precursor muscle groups that have a direct functional input relationship with the functional bottleneck muscle group are traced upstream, and the set of successor muscle groups that are directly constrained by the functional bottleneck muscle group are traced downstream.

[0141] In the muscle group function-dependent topology, search for bypass muscle group nodes that bypass functional bottlenecks and connect the precursor muscle group set and the successor muscle group set. Quantitatively evaluate the functional compensation ability of each muscle group node in the bypass muscle group chain, and select the bypass muscle group chain whose functional compensation ability meets the preset compensation ability threshold as the alternative muscle group path.

[0142] The set of exercise movements corresponding to each muscle group node in the alternative muscle group path is retrieved from the preset muscle group and exercise movement mapping library. The execution order of the exercise movement set is determined according to the order of each muscle group node in the alternative muscle group path in the bypass muscle group chain, and an auxiliary exercise movement sequence for the alternative muscle group path is generated.

[0143] Analyze the target muscle group coverage of each training stage in the layered progressive training program, and identify the target training stage with the highest functional correlation to the functional bottleneck muscle group.

[0144] The auxiliary exercise sequence is inserted into the target exercise phase, replacing the main exercise movements corresponding to the functional bottleneck muscle groups.

[0145] After identifying the functional bottleneck muscle groups, feasible alternative muscle pathways need to be retrieved from the muscle group function-dependent topology, and corresponding auxiliary exercise movements should be appropriately inserted into a tiered and progressive exercise program to ensure that older adults can continue to receive effective exercise stimulation during periods of functional limitation in the bottleneck muscle groups. The entire retrieval and insertion process is completed using node location information, edge direction information, and quantitative results of functional compensation capacity in the topology.

[0146] For the identified functional bottleneck muscle groups, their corresponding nodes are located in the muscle group function-dependent topology, and single-hop neighborhood searches are performed along both the reverse and forward directions of the directed edges. When tracing in the reverse direction, all upstream nodes that directly output functional influence to the functional bottleneck muscle group nodes are collected, forming a precursor muscle group set; when tracing in the forward direction, all downstream nodes that directly receive output from the functional bottleneck muscle group nodes are collected, forming a successor muscle group set. The precursor muscle group set reflects the functional input sources that the functional bottleneck muscle group depends on under normal conditions, while the successor muscle group set reflects the functional output objects directly affected by the functional bottleneck muscle group's limitation. By simultaneously defining these two sets, the range of the topological region that needs to be bypassed can be precisely defined, providing a clear starting and ending point set for subsequent searches of bypass muscle group chains.

[0147] In the topological structure, bypass muscle group chains are searched, requiring candidate paths to meet the following conditions: the starting point of the path belongs to the set of precursor muscle groups, the ending point belongs to the set of successor muscle groups, and the path does not contain the functional bottleneck muscle group node itself. For candidate bypass muscle group chains that meet the topological connectivity conditions, the functional compensation capacity of each muscle group node is further quantitatively evaluated. The quantification of functional compensation capacity comprehensively considers the current functional reserve level of the muscle group at the node, its out-degree scale in the topological structure, and the activation response efficiency reflected by historical training data. Let the th node in the bypass chain be... The functional compensatory capacity score of each muscle group node is as follows: Then the overall compensatory capacity of the bypass muscle group chain Take the harmonic mean of the compensation capacity scores of each node in the chain, that is,

[0148]

[0149] in This represents the number of nodes in the bypass muscle group chain. The harmonic mean is used instead of the arithmetic mean to make the overall compensatory capacity more sensitive to nodes with weaker compensatory capacity in the chain, avoiding the situation where the entire bypass chain cannot effectively carry out functional transmission due to the excessively low compensatory capacity of individual nodes. With preset compensation capacity threshold Compare and retain only those that meet the requirements. The alternative muscle group chains are considered as effective alternative muscle group pathways. If multiple alternative muscle group chains satisfy the threshold condition, they are selected first. The highest-order chain with the fewest nodes is used to reduce the coordination burden on older adults when performing assistive action sequences.

[0150] After identifying the alternative muscle group pathway, a set of exercises corresponding to each muscle group node in the alternative muscle group pathway is retrieved from a pre-defined muscle group and exercise movement mapping library. The mapping library uses muscle group identifiers as index keys, with each muscle group node corresponding to multiple candidate exercises. Each candidate exercise is accompanied by attribute tags such as difficulty level, required assistive device type, corresponding physiological load intensity range, and applicable functional state range. During the retrieval, combined with the current functional state assessment results of the target elderly person, exercises whose difficulty level matches their current functional state and whose physiological load intensity is within a safe range are selected as representative exercises for that node from the candidate exercises corresponding to each node. The representative exercises for each node are arranged according to the topological order of each node in the alternative muscle group pathway within the bypass muscle group chain, forming an assistive exercise sequence with a clear execution order. The topological order is determined based on the direction of directed edge transmission; that is, the exercise corresponding to the node with the earlier functional output direction is executed before the exercise corresponding to the node with the later functional input direction, thus ensuring that the muscle group activation order during assisted exercise is consistent with the physiological logic of the functional transmission link.

[0151] Before inserting assisted exercise sequences into a hierarchical, progressive training program, it is necessary to analyze the target muscle group coverage of each training phase in the program and identify the target training phase with the highest functional correlation to the functional bottleneck muscle groups. The functional correlation is calculated based on the topological distance distribution between the set of target muscle groups covered by each training phase and the functional bottleneck muscle groups within the topological structure. Let the first... The target muscle groups covered in each training phase are: The functional bottleneck muscle group node is Then the first Functional correlation between each stage and the functional bottleneck muscle groups Defined as

[0152]

[0153] in Represents a node With nodes The shortest path distance in the topology This is the distance attenuation coefficient. For the first The number of target muscle groups covered in each stage. A higher value indicates that the target muscle group in this training phase is topologically closer to the functional bottleneck muscle group, and the stronger the functional correlation. (Selection) The most intense training phase is designated as the target training phase, and auxiliary training sequence is inserted into this phase.

[0154] The insertion operation employs substitution rather than simple superposition: During the target training phase, the main exercise movement corresponding to the functional bottleneck muscle group is identified, removed from the current phase's movement list, and an auxiliary exercise sequence is inserted at the removal location. After the substitution operation is completed, the total physiological load of the target training phase is re-estimated. If the estimated total load after inserting the auxiliary exercise sequence exceeds the target elderly person's current phase's safe load limit, the repetition count or duration of each movement in the auxiliary exercise sequence is proportionally reduced until the total load falls back within a safe range. The adjusted hierarchical progressive training program incorporates the updated target training phase into the overall execution flow, while the start time nodes and transition conditions of subsequent phases remain unchanged, ensuring that the hierarchical progression logic of the overall program does not suffer structural disruption due to local substitution operations. Through the above complete retrieval, evaluation, mapping, and insertion process, even when the functional bottleneck muscle group cannot participate normally in training, the continuity of training for the target functional area can still be maintained by activating the bypass muscle group chain with functional compensation capabilities, effectively reducing the risk of interruption of the overall training program due to limited local muscle group function.

[0155] A second aspect of the present invention provides an exercise system for preventing sarcopenia in the elderly, comprising:

[0156] The feature extraction unit is used to perform time-series correlation analysis on the multi-dimensional physical state data of the target elderly and extract the feature vector of muscle function degeneration process.

[0157] The cascade identification unit is used to construct a muscle group function dependency topology based on the feature vector of the muscle function degeneration process, identify the cascade influence nodes of the muscle group function dependency topology and the functional dependency path between the nodes, and determine the set of target intervention muscle groups with cascade influence effect.

[0158] The program generation unit is used to generate a hierarchical progressive exercise program for the target intervention muscle group set. During the execution of the hierarchical progressive exercise program by the target elderly, real-time execution data is collected to establish a two-way mapping relationship between the quality of movement execution and physiological load response.

[0159] The instant correction unit is used to input the real-time execution data into the bidirectional mapping relationship to perform forward prediction and backward calculation respectively, to obtain the deviation direction and deviation magnitude of the current action execution quality, and to generate an instant correction instruction containing an action adjustment vector based on the deviation direction and the deviation magnitude and to provide real-time feedback to the target elderly person through the interactive interface.

[0160] The bottleneck identification unit is used to identify the functional bottleneck muscle groups that appear in the current training phase based on the real-time execution data.

[0161] The program adjustment unit is used to retrieve alternative muscle group paths that have a functional compensatory relationship with the functional bottleneck muscle group from the muscle group function-dependent topology, and insert the auxiliary exercise movements corresponding to the alternative muscle group paths into the layered progressive exercise program.

[0162] A third aspect of the present invention provides an electronic device, comprising:

[0163] processor;

[0164] Memory used to store processor-executable instructions;

[0165] The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0166] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0167] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0168] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. Exercise methods for preventing sarcopenia in the elderly, characterized in that, include: Temporal correlation analysis was performed on multidimensional physical condition data of the target elderly population to extract feature vectors of muscle function degeneration process; Based on the feature vector of the muscle function degeneration process, a muscle group function dependency topology is constructed. The cascaded influence nodes and functional dependency paths between nodes in the muscle group function dependency topology are identified, and the set of target intervention muscle groups with cascaded influence effects is determined. A tiered, progressive exercise program is generated for the target muscle group set. During the execution of the tiered, progressive exercise program by the target elderly person, real-time execution data is collected to establish a two-way mapping relationship between the quality of movement execution and physiological load response. The real-time execution data is input into the bidirectional mapping relationship for forward prediction and backward calculation to obtain the deviation direction and deviation magnitude of the current action execution quality. Based on the deviation direction and deviation magnitude, an instant correction instruction containing an action adjustment vector is generated and fed back to the target elderly person in real time through the interactive interface. Based on the real-time execution data, identify the muscle groups that are experiencing functional bottlenecks during the current training phase. The alternative muscle group paths that have a functional compensatory relationship with the functional bottleneck muscle group are retrieved from the muscle group function-dependent topology, and the auxiliary exercise movements corresponding to the alternative muscle group paths are inserted into the layered progressive exercise program.

2. The method according to claim 1, characterized in that, A time-series correlation analysis was performed on the multidimensional physical condition data of the target elderly population to extract feature vectors of muscle function degeneration process, including: A multi-scale time window sequence is constructed for the multi-dimensional physical state data of the target elderly. Instantaneous change features, stage evolution features and cumulative degeneration features of the physical state data are extracted in each time window. The stage inflection points of muscle function degeneration at different time scales are identified by the feature transfer relationship across windows. Based on the aforementioned stage turning point, the time series is divided into multiple degeneration stage intervals. A dynamic correlation matrix between body parts is constructed in each degeneration stage interval. The dynamic correlation matrix quantifies the mutual influence of functional decline of body parts in the current degeneration stage. By comparing the differences in the dynamic correlation matrix across different degradation stage intervals, the evolution trajectory features of functional dependencies between body parts as degradation progresses are extracted. The instantaneous change features, stage evolution features, and cumulative degradation features extracted from the multi-scale time window sequence are aggregated with the dynamic correlation matrix and evolution trajectory features of different degradation stage intervals at multiple levels. By quantifying the functional decline acceleration of each body part in different degradation stages and the transformation rate of functional dependence between body parts, a feature vector of the muscle function degradation process containing temporal evolution dimension and cross-part correlation dimension is generated.

3. The method according to claim 1, characterized in that, Based on the feature vector of the muscle function degeneration process, a muscle group function dependency topology is constructed. The cascading influence nodes and functional dependency paths between nodes in the muscle group function dependency topology are identified. The set of target intervention muscle groups with cascading influence effects is determined, including: From the feature vector of the muscle function degeneration process, feature components containing temporal evolution dimension and cross-site association dimension are extracted. The muscle groups corresponding to each body part are used as topological nodes, and the functional dependencies between body parts represented in the feature components are used as topological edges to construct an initial muscle group functional dependency topological structure. The initial muscle group function-dependent topology is directionally labeled and intensity quantized. The directional labeling is determined based on the dominance order of the driving muscle group and the responding muscle group, and the intensity quantization is determined based on the influence coefficient of the change in the functional state of the driving muscle group on the change in the functional state of the responding muscle group. The initial muscle group function-dependent topology is transformed into a weighted directed graph structure. In the weighted directed graph structure, based on the multi-hop path propagation attenuation coefficient of each topological node from the current node to other nodes along the topological edge, the cascaded influence node muscle group that can influence multiple downstream nodes through a finite number of path hops is identified. Extract the functionally dependent paths from the currently described cascaded influence node muscle groups to each downstream node, and calculate the cumulative conduction gain of each functionally dependent path. The cumulative conduction gain of the functionally dependent path from the current cascaded affected muscle group to all downstream nodes is summed to obtain the total conduction gain contribution value of each cascaded affected muscle group. The target intervention muscle group set is selected based on the total conduction gain contribution value.

4. The method according to claim 1, characterized in that, A tiered, progressive exercise program is generated for the target muscle group set. During the execution of the tiered, progressive exercise program by the target elderly individuals, real-time execution data is collected, and a bidirectional mapping relationship between movement execution quality and physiological load response is established, including: Extract the topological hierarchy of each muscle group in the target intervention muscle group set; The layered and progressive training program divides the training movements into a basic stability layer, a functional linkage layer, and a comprehensive coordination layer. Based on the aforementioned topological hierarchy, muscle groups located at the starting position are assigned to the basic stabilization layer, and single-joint isolated activation exercises are generated based on the characteristics of the driving muscle groups; muscle groups located in the middle position that receive transmission from multiple pathways are assigned to the functional linkage layer, and cross-joint joint recruitment exercises are generated based on the characteristics of multi-path convergence; muscle groups located at the end position that integrate multi-muscle group synergy are assigned to the comprehensive coordination layer, and whole-body integrated exercises are generated based on the characteristics of multi-muscle group synergy. The motion trajectory data and physiological monitoring data collected during the execution of the layered progressive training program are used as the real-time execution data, and the motion execution quality characteristics and physiological load response characteristics are extracted respectively. A bidirectional mapping relationship is established between the action execution quality characteristics and the physiological load response characteristics, including a forward prediction mapping and a reverse correction mapping. The forward prediction mapping predicts the expected physiological load response characteristics based on the action execution quality characteristics, and the reverse correction mapping adjusts the target range of the action execution quality characteristics in reverse based on the deviation between the actual physiological load response characteristics and the expected physiological load response characteristics.

5. The method according to claim 4, characterized in that, The real-time execution data is input into the bidirectional mapping relationship for forward prediction and backward calculation to obtain the deviation direction and magnitude of the current action execution quality. Based on the deviation direction and magnitude, an immediate correction instruction containing an action adjustment vector is generated and fed back to the target elderly person in real time through an interactive interface, including: Based on the real-time execution data, the physiological load response characteristics are compared with the expected physiological load response characteristics calculated by the positive prediction mapping in the bidirectional mapping relationship to obtain the physiological load response deviation. The physiological load response deviation is input into the reverse correction mapping in the bidirectional mapping relationship. The contribution of each dimension of the action execution quality feature in the reverse correction mapping to the physiological load response deviation is calculated. The action execution quality feature dimension with the largest contribution is selected as the deviation direction. The magnitude of the physiological load response deviation is extracted as the deviation amplitude. Based on the deviation direction, a preset mapping table of action execution quality feature dimensions and body motion space parameters is queried. The deviation direction is mapped to the specific dimension of the body motion space parameters as the spatial dimension of action adjustment. The action adjustment trend is determined according to the positive or negative value of the deviation amplitude. The degree of quantification of action adjustment is determined according to the magnitude of the deviation amplitude. The spatial dimension, the motion adjustment trend, and the quantification level are combined to generate a motion adjustment vector, and the motion adjustment vector is converted into an instant correction instruction containing motion correction indications; The instant correction instructions are fed back to the target elderly person in real time through an interactive interface that includes visual and voice feedback channels.

6. The method according to claim 1, characterized in that, Based on the real-time execution data, the muscle groups that experience functional bottlenecks during the current training phase include: Extract activation intensity data and fatigue accumulation data of each muscle group during the current training phase from the real-time execution data; For each muscle group in the target intervention muscle group set, the deviation of the activation intensity data from the preset benchmark activation intensity is calculated, the degree of exceeding the fatigue accumulation data from the preset fatigue threshold is calculated, and the deviation and the degree of exceeding the threshold are weighted and fused to obtain the functional limitation score of each muscle group. Construct a collaborative dependency graph among the muscle groups in the target intervention muscle group set, count the number of downstream dependent muscle groups of each muscle group node in the collaborative dependency graph, select muscle group nodes whose number of downstream dependent muscle groups exceeds a preset downstream dependency number threshold as high-impact node muscle groups, and multiply the functional limitation score of the high-impact node muscle group with the number of downstream muscle groups constrained by the high-impact node muscle group to obtain the bottleneck impact index. All muscle groups in the target intervention muscle group set are sorted in descending order according to the bottleneck impact index, and the muscle groups whose bottleneck impact index ranks in the top preset position are selected as the functional bottleneck muscle groups.

7. The method according to claim 1, characterized in that, Retrieving alternative muscle group pathways from the muscle group function-dependent topology that have a functional compensatory relationship with the functional bottleneck muscle group, and inserting the auxiliary exercise movements corresponding to the alternative muscle group pathways into the layered progressive exercise program includes: Based on the node position of the functional bottleneck muscle group in the muscle group function-dependent topology, the set of precursor muscle groups that have a direct functional input relationship with the functional bottleneck muscle group are traced upstream, and the set of successor muscle groups that are directly constrained by the functional bottleneck muscle group are traced downstream. In the muscle group function-dependent topology, search for bypass muscle group nodes that bypass functional bottlenecks and connect the precursor muscle group set and the successor muscle group set. Quantitatively evaluate the functional compensation ability of each muscle group node in the bypass muscle group chain, and select the bypass muscle group chain whose functional compensation ability meets the preset compensation ability threshold as the alternative muscle group path. The set of exercise movements corresponding to each muscle group node in the alternative muscle group path is retrieved from the preset muscle group and exercise movement mapping library. The execution order of the exercise movement set is determined according to the order of each muscle group node in the alternative muscle group path in the bypass muscle group chain, and an auxiliary exercise movement sequence for the alternative muscle group path is generated. Analyze the target muscle group coverage of each training stage in the layered progressive training program, and identify the target training stage with the highest functional correlation to the functional bottleneck muscle group. The auxiliary exercise sequence is inserted into the target exercise phase, replacing the main exercise movements corresponding to the functional bottleneck muscle groups.

8. An exercise system for preventing sarcopenia in the elderly, used to implement the method as described in any one of claims 1-7, characterized in that, include: The feature extraction unit is used to perform time-series correlation analysis on the multi-dimensional physical state data of the target elderly and extract the feature vector of muscle function degeneration process. The cascade identification unit is used to construct a muscle group function dependency topology based on the feature vector of the muscle function degeneration process, identify the cascade influence nodes of the muscle group function dependency topology and the functional dependency path between the nodes, and determine the set of target intervention muscle groups with cascade influence effect. The program generation unit is used to generate a hierarchical progressive exercise program for the target intervention muscle group set. During the execution of the hierarchical progressive exercise program by the target elderly, real-time execution data is collected to establish a two-way mapping relationship between the quality of movement execution and physiological load response. The instant correction unit is used to input the real-time execution data into the bidirectional mapping relationship to perform forward prediction and backward calculation respectively, to obtain the deviation direction and deviation magnitude of the current action execution quality, and to generate an instant correction instruction containing an action adjustment vector based on the deviation direction and the deviation magnitude and to provide real-time feedback to the target elderly person through the interactive interface. The bottleneck identification unit is used to identify the functional bottleneck muscle groups that appear in the current training phase based on the real-time execution data. The program adjustment unit is used to retrieve alternative muscle group paths that have a functional compensatory relationship with the functional bottleneck muscle group from the muscle group function-dependent topology, and insert the auxiliary exercise movements corresponding to the alternative muscle group paths into the layered progressive exercise program.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images