Lower limb abnormal gait tracing method and system combined with distributed joint sensing network

Abstract

The invention discloses a lower limb abnormal gait tracing method and system combined with a distributed joint sensing network, and relates to the technical field of gait detection. The method comprises the steps that according to a distributed joint sensing network, lower limb movement data are collected and uploaded to a monitoring middle table, and the lower limb movement data are reconstructed into a function information flow graph; by activating a gait traceability module embedded in the monitoring middle table, performing two-way joint judgment on the function information flow graph, and determining an abnormal gait traceability result; and performing early warning response and indication response on the abnormal gait traceability result on a terminal interface of the monitoring middle station. The technical problems that in the prior art, original disturbance joints of lower limb abnormal gaits are difficult to accurately position, the misjudgment rate is high, and abnormal source positioning is fuzzy are solved, accurate tracing of the abnormal gaits is achieved based on the distributed joint sensing network, the recognition accuracy and the real-time early warning capacity are improved, and the accuracy of abnormal gaits is improved. The technical effect of quickly judging the lower limb dysfunction is achieved.

Description

Technical Field

[0001] This invention relates to the field of gait detection technology, specifically to a method and system for tracing abnormal lower limb gait using a distributed joint sensing network. Background Technology

[0002] In modern rehabilitation medicine, sports science, and geriatric health monitoring, the accurate detection and analysis of lower limb gait is of paramount importance. Abnormal gait is often a significant external manifestation of neurological diseases, musculoskeletal injuries, or postoperative functional impairments. Traditional gait analysis methods largely rely on optical motion capture systems or single wearable inertial sensors. While these methods provide basic kinematic parameters, they have significant limitations. Optical systems are expensive and limited to laboratory environments, making long-term monitoring in natural living conditions difficult; while single sensors often provide only localized, isolated motion data, lacking a global perspective on the coordinated working patterns of joints and muscles throughout the lower limb biomechanical chain. This makes it difficult for current technologies to effectively distinguish normal physiological fluctuations from true pathological abnormalities from complex motion information, and even more difficult to accurately trace the root cause of abnormal gait. Clinicians and therapists often rely on experience for inference, lacking objective, quantitative tools for tracing the cause, thus affecting the accuracy and timeliness of rehabilitation interventions. Summary of the Invention

[0003] This application provides a method and system for tracing abnormal gait of the lower limbs by combining a distributed joint sensing network, which solves the technical problems of existing technologies that make it difficult to accurately locate the original perturbed joints of abnormal gait of the lower limbs, resulting in a high misjudgment rate and ambiguous location of the abnormal source.

[0004] The first aspect of this application provides a method for tracing abnormal gait in the lower limbs by incorporating a distributed joint sensing network, the method comprising: Based on a distributed joint sensing network, lower limb motion data is collected and uploaded to a monitoring platform. The lower limb motion data is reconstructed into a functional information flow graph. By activating the gait tracing module embedded in the monitoring platform, a two-way joint judgment is performed on the functional information flow graph to determine the abnormal gait tracing result. The two-way joint judgment uses symmetry comparison and self-comparison under gait cycle for the first judgment, and information flow anomalies under the core perturbation mode for the second judgment. The original perturbation source probability distribution based on the judgment is used as the output. On the terminal interface of the monitoring platform, the abnormal gait tracing result is responded to with an early warning and an indication.

[0005] A second aspect of this application provides a lower limb abnormal gait tracing system incorporating a distributed joint sensing network, the system comprising: Data acquisition module: Based on the distributed joint sensor network, it collects lower limb motion data and uploads it to the monitoring platform, reconstructing the lower limb motion data into a functional information flow graph; Anomaly judgment module: By activating the gait tracing module embedded in the monitoring platform, it performs a two-way joint judgment on the functional information flow graph to determine the abnormal gait tracing result. The two-way joint judgment uses symmetry comparison and self-comparison under gait cycle for the first judgment, and information flow anomaly under the core perturbation mode for the second judgment, with the original perturbation source probability distribution based on the judgment as the output; Anomaly warning module: On the terminal interface of the monitoring platform, it provides warning and indication responses to the abnormal gait tracing result.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: First, a distributed joint sensor network is used to synchronously monitor multiple key joints and muscle segments of the lower limbs, acquiring continuous motion data, which is then uploaded to a monitoring platform for processing. The platform analyzes the information flow and intensity of the raw data, reconstructing it into a functional information flow map reflecting the functional coordination of the lower limbs. Subsequently, the platform's built-in gait tracing module is activated, performing a dual-path joint judgment on the generated functional information flow map. On one hand, it identifies abnormal deviations between left and right sides or between cycles through symmetry comparison and self-comparison within the gait cycle; on the other hand, it detects abnormal changes in functional coupling relationships based on core perturbation patterns. The module then fuses the results of these two types of judgments, generating a probability distribution of the original perturbation source as the abnormal gait tracing result, which is output to indicate the functional nodes most likely to cause abnormal gait. Finally, the monitoring platform's terminal interface responds in real time based on this result, including issuing abnormal warnings and providing corresponding intervention or prompt information, so that users or medical personnel can take timely measures. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a schematic diagram of the lower limb abnormal gait tracing method combined with a distributed joint sensing network provided in an embodiment of this application.

[0009] Figure 2 This is a schematic diagram of the structure of a lower limb abnormal gait tracing system that incorporates a distributed joint sensing network, as provided in an embodiment of this application.

[0010] Figure labeling: Data acquisition module 11, anomaly detection module 12, anomaly warning module 13. Detailed Implementation

[0011] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0012] Example 1, as Figure 1 As shown, this application provides a method for tracing abnormal gait in the lower limbs by combining a distributed joint sensing network. The method includes: Based on the distributed joint sensing network, lower limb motion data is collected and uploaded to the monitoring platform, and the lower limb motion data is reconstructed into a functional information flow graph.

[0013] In this embodiment, a distributed joint sensor network is fixedly deployed at major joints of the lower limbs, such as the hip, knee, and ankle, as well as key muscle segments such as the anterior and posterior thigh and posterior and posterior calf muscles. This distributed joint sensor network consists of several micro-motion radar sensors and inertial measurement unit sensors, and establishes a data communication connection with the monitoring platform. During walking, standing, or specific functional movements, the system drives each sensor node to synchronously collect raw lower limb motion data such as acceleration, angular velocity, joint rotation amplitude, and muscle segment micro-motion characteristics at a preset sampling frequency, and synchronizes and calibrates the multi-node data using local timestamps. After collection, each sensor node packages the lower limb motion data into data frames and uploads them to the data access interface of the monitoring platform via Bluetooth, Wi-Fi, or a dedicated communication link. The monitoring platform preprocesses the received raw motion data, including time alignment, coordinate system normalization, noise filtering, outlier removal, and gait cycle-based segmentation for data from different sensor types. This yields time-series signals reflecting the dynamic changes of each joint-muscle functional unit. Time alignment can be achieved through linear interpolation synchronization based on a unified timestamp or multi-sensor clock calibration. Coordinate system normalization can be performed by establishing a unified local human coordinate system and performing posture rotation transformations on the sensor coordinates, such as quaternion transformation. Noise filtering can be achieved through bandpass filtering, wavelet thresholding, or Kalman filtering. Outlier removal can be achieved through outlier detection based on statistical thresholds. Gait cycle-based segmentation can be achieved through feature peak-valley detection. Then, using each joint-muscle functional unit as a functional node, the transfer entropy between any two nodes is calculated. The direction of information transmission from the predecessor node to the successor node is defined as a directed edge, and the transfer entropy value is defined as the weight of the directed edge, thus constructing a directed weighted graph reflecting the causal transmission relationship between the functional nodes of the lower limb. Then, the directed weighted graph is dynamically updated according to the time axis, and the directed weighted relationship under different time periods or different time states is serialized to obtain the functional information flow graph for subsequent gait analysis and source tracing judgment. The functional information flow graph is stored in the database of the monitoring platform for subsequent use.

[0014] Furthermore, lower limb movement data is collected and uploaded to the monitoring platform, and the lower limb movement data is reconstructed into a functional information flow graph, including: A distributed key sensor network is deployed based on the physical structure of the lower limbs, with major joints and key muscle segments as sensor deployment locations and micro-motion radar and inertial measurement units as sensor types. By driving the distributed joint sensor network, lower limb motion data is collected. Through communication interaction between the distributed joint sensor network and the monitoring platform data interface, data is transmitted back and reconstructed based on the transmission flow direction and intensity to generate the functional information flow graph.

[0015] Preferably, based on the skeletal and muscular anatomy of the human lower limbs, several sensor node installation points are planned at major weight-bearing and movement joints such as the hip, knee, and ankle joints, as well as key muscle segments such as the quadriceps, hamstrings, calf muscles, and tibialis anterior. Inertial measurement units (IMUs) are preferentially fixed at each joint to acquire triaxial acceleration, triaxial angular velocity, and spatial attitude information at that joint. Micro-motion radars are then deployed on the surface of each target muscle segment or near the muscle belly to operate in a non-contact or adhesive manner. By emitting and receiving electromagnetic waves, they continuously detect minute displacements and vibrations on the muscle surface, extracting micro-vibration envelope features and spectral features related to muscle contraction, relaxation, and fatigue, thereby obtaining highly sensitive signals reflecting the dynamic activity state of the muscles. The sensor nodes at each joint and muscle are connected via a bus or low-power wireless network to form a distributed joint sensing network. Gateway nodes or edge nodes are set up in the network to perform local data caching, initial packaging, and protocol conversion. Subsequently, when the monitored subject performs preset action tasks such as natural walking, climbing stairs, and rehabilitation training, the data acquisition program of the distributed joint sensor network is activated. This triggers each inertial measurement unit (IMU) and micro-motion radar to synchronously acquire raw data at a uniform sampling frequency, with a unified time reference or global timestamp appended to each frame of data. During the acquisition process, the micro-motion radar outputs raw echo signals containing the micro-vibration morphology of the muscle surface or demodulated displacement / velocity envelope signals. The IMU outputs linear acceleration, angular velocity, and attitude quaternion data corresponding to the joint. The sensor nodes perform preliminary filtering and compression encoding on these multi-source data, and then transmit them to the monitoring center via a pre-configured data interface protocol, such as TCP / IP or MQTT, through a wireless or wired link. After receiving lower limb motion data from various nodes, the monitoring platform performs time alignment and source calibration of the multi-node data based on timestamps and node identifiers. Then, it performs echo demodulation, bandpass filtering, wavelet or frequency domain analysis on the micro-motion radar signals to extract muscle micro-motion features. For the inertial measurement unit data, it performs gravity separation, coordinate system unification, and noise suppression to obtain a normalized time series for each joint-muscle unit. Next, using each joint-muscle unit as a functional node, it calculates the transfer entropy between any two node time series according to the upstream and downstream relationships in sports biomechanics to characterize the direction and intensity of information transmission. Based on this, a directed weighted graph characterizing the causal transmission and synergistic relationships between multiple nodes in the lower limb is constructed. Then, the directed weighted graphs with different time-state cycles and different movement stages are serialized and dynamically updated along the time axis, expanding the static directed weighted structure into an information flow network representation that evolves over time, forming a functional information flow graph. Finally, this functional information flow graph is stored in the monitoring platform as structured data for subsequent gait anomaly identification and source analysis.

[0016] Furthermore, performing reconstruction based on transmission direction and intensity includes: Using joint-muscle functional units as nodes, a directed weighted graph is constructed based on the direction and intensity of information flow by introducing transfer entropy. The transfer entropy is defined by the transfer causality and transfer entropy value. The directed weighted graph is dynamically generated based on the time series to generate the functional information flow graph.

[0017] Optionally, for each joint-muscle functional unit, data collected by its sensor nodes are extracted, and then interpolated, normalized, and time-axis aligned to ensure comparability between data from different nodes. Subsequently, any two joint-muscle functional units are denoted as source node X and target node Y. By introducing the concept of transfer entropy, the information contribution of X's time series to Y's time series under time delay conditions is quantified, yielding the intensity and direction of the information flow from X to Y. The calculation of transfer entropy comprehensively considers conditional probability distribution and time lag relationships. The information transfer quantity is defined by both transfer causality and transfer entropy value. That is, when the past state of X can significantly reduce the uncertainty of Y's current state, an effective information flow from X to Y is considered to exist, and its magnitude is characterized by the transfer entropy value, which is the average quantized value of the difference between these two types of conditional probability distributions over the statistical period. Taking the thigh muscle group-knee joint as an example, the force line changes or micro-motion intensity of the thigh muscle group are used as the source node time series, and the knee joint angle changes are used as the target node time series. Under normal gait conditions, the transfer entropy between the two theoretically exhibits a relatively stable and approximately linear quantitative relationship, which can form a baseline value or baseline range of transfer entropy for normal conditions, used for subsequent comparative analysis of abnormal conditions. By calculating the bidirectional transfer entropy of all joint-muscle functional units in pairs, a directed weighted graph is constructed with functional units as nodes, information flow direction as directed edges, and transfer entropy values ​​as edge weights. The directionality is constrained according to the transfer causal relationship, that is, from the side with significantly larger transfer entropy to the other side, forming a directed edge with causal direction. Subsequently, to reflect the evolution of lower limb functional coordination over time, the constructed directed weighted graphs were dynamically processed using gait cycles or fixed-length time windows as units. Specifically, the global transfer entropy was independently calculated within each time window, generating a corresponding directed weighted graph. Then, multiple frames of directed weighted graphs from consecutive time windows were concatenated or stacked according to the time sequence, forming a functional information flow graph that updates over time. This functional information flow graph not only shows the direction and intensity of information transfer between joint-muscle functional units at any given moment, but also reveals dynamic changes such as deviations, sudden drops, or abnormal oscillations in the transfer entropy from the thigh muscle group to the knee joint from the normal baseline value. This provides a complete data foundation and structured representation for subsequent gait disturbance identification and tracing based on information flow anomalies.

[0018] By activating the gait tracing module embedded in the monitoring platform, a two-way joint judgment is performed on the functional information flow graph to determine the abnormal gait tracing result. The two-way joint judgment is based on the first judgment of symmetry comparison and self-comparison under gait cycle, and the second judgment is based on the information flow anomaly under the core disturbance mode. The original disturbance source probability distribution based on the judgment is used as the output.

[0019] In one embodiment, after the monitoring platform receives and generates the functional information flow graph corresponding to the current gait of the target object, it invokes a gait tracing module pre-deployed within the monitoring platform and uses this module to perform a two-way joint judgment on the functional information flow graph. Specifically, the first gait judge of the gait tracing module performs a symmetry comparison and a self-comparison under gait loops based on the functional information flow graph to obtain a first judgment result. Subsequently, the second gait judge of the gait tracing module executes a second judgment process, that is, analyzes the information flow anomalies based on the core disturbance mode to obtain a second judgment result. Afterwards, the first and second judgment results are input into the third tracing generator of the gait tracing module to perform path backtracking analysis on the marked abnormal functional nodes and their upstream and downstream nodes, calculate the probability value of each functional node as the original disturbance source, and form the original disturbance source probability distribution for all functional nodes. Finally, the probability distribution, along with the gait type, abnormality category, and severity involved in the first and second judgments, is encapsulated into an abnormal gait tracing result and output to the monitoring platform for subsequent early warning prompts and intervention plan generation.

[0020] Furthermore, before activating the gait tracing module embedded in the monitoring platform, the construction of the gait tracing module includes: A first gait judge is deployed based on symmetry comparison and self-comparison under gait cycle; a second gait judge is deployed based on information flow anomalies under core perturbation mode, wherein the core perturbation mode includes at least information flow interruption, information flow compensation and information flow oscillation; a gait tracing module is constructed based on the first gait judge, the second gait judge and the third tracing generator.

[0021] Preferably, based on the anatomical symmetry of the human lower limb structure and the periodicity of gait, a first gait judge is established in the monitoring platform. This first gait judge is based on symmetry comparison and self-comparison under gait cycles. By pairwise matching of left and right functional units in the functional information flow graph, comparing their information flow intensity, critical path, and transmission entropy value in the same gait phase, it determines whether there is a structural or functional shift between the left and right sides. At the same time, it performs self-comparison on multiple consecutive gait cycles on the same side to identify whether the gait periodicity is disturbed. The first gait judge obtains a first-class judgment result reflecting gait symmetry and periodic stability through the above-mentioned dual consistency assessment. Subsequently, based on the biomechanical logic of the core functional links and the common propagation modes of gait abnormalities, a second gait judge is deployed in the monitoring platform. This second gait judge uses the information flow abnormality of the core perturbation mode as the main identification basis. It constructs an information flow template around the key joint links and important muscle group coordination paths in lower limb movement and compares it with the current functional information flow graph node by node and path by path. When the current graph exhibits information flow interruptions (significant decrease in edge weights or transit entropy), information flow compensation (significant enhancement of alternative paths), or information flow oscillations (high fluctuations in information flow over time), the second gait discriminator automatically identifies and labels the corresponding disturbance category and severity, forming a second-class judgment result. Subsequently, to achieve source tracing output for abnormal gait, a third source tracing generator is constructed based on the above two discriminators. For example, the third source tracing generator can be trained using an adversarial network (GAN), where the discriminator learns the differences between the normal gait information flow structure and the abnormal structure, and the generator simulates the graph structure changes that the original disturbance source might cause. After training, only the generator part is retained as the source tracing architecture, enabling it to generate a set of disturbance source probability distributions that best match the current abnormal behavior based on the input first and second judgment results. The generator continuously infers the starting node of the anomaly, the propagation direction of the abnormal information flow, and its diffusion path, thereby outputting the probability value of each functional node as the original disturbance source. Finally, the first gait judge, the second gait judge, and the third gait source generator are integrated in a modular manner to form a complete gait source tracing module. This gait source tracing module can perform structured analysis, anomaly identification, and source inference on the input functional information flow graph, realizing the full-link function from anomaly discovery to anomaly source location, and providing core support for the subsequent early warning prompts and intervention plan generation of the monitoring platform.

[0022] Furthermore, the functional information flow graph is subjected to a two-way joint judgment to determine the abnormal gait source tracing result, including: The functional information flow graph is input into the first gait judge and the first gait judgment result is output. The functional information flow graph is input into the second gait judge and the second gait judgment result is output. The first gait judgment result and the second gait judgment result are input into the third source generator. By analyzing the path of disturbance propagation, the probability distribution is determined as the abnormal gait source result. The probability distribution indicates the probability value of each functional node as the original disturbance source.

[0023] Optionally, after the monitoring platform generates the functional information flow graph corresponding to the current gait of the target object, it first sends this functional information flow graph as input to the first gait judge of the gait tracing module. This first gait judge, based on the pre-stored left and right side functional node mapping relationship, divides the functional information flow graph into a first side image and a second side image under symmetrical pose, representing the left and right sub-images respectively. Subsequently, the first gait judge performs a consistency comparison between the first and second side images to detect whether there is a symmetry shift. Simultaneously, based on gait event annotations, such as heel strike and toe lift, the functional information flow graph is segmented according to the gait cycle, and the first gait cycle image on the same side is compared with the second gait cycle image. By summarizing the results of the two comparisons, a first gait judgment result is formed, including symmetry anomalies, cycle self-comparison anomalies, and corresponding confidence scores. Simultaneously, the same functional information flow graph is input into the second gait judge. The second gait judge extracts core functional links from the functional information flow graph, such as the hip-knee-ankle main chain and muscle group synergistic chains. By comparing the edge weights and propagation entropies of each core link in the current information flow graph with the information flow template one by one, it identifies whether abnormal patterns such as information flow interruption, compensation, and oscillation have occurred, forming the second gait judgment result. After obtaining the first and second gait judgment results, the monitoring platform inputs both into the third source generator to perform source inference of the anomaly source. The third source generator constructs a candidate propagation path set for disturbance events based on the abnormal nodes and abnormal links marked by the two types of judgment results, and combines the time evolution sequence of the functional information flow graph to determine how the disturbance spreads, is transmitted, or is amplified between joint-muscle functional units. The generator scores all possible perturbation paths through an internally established adversarial generation structure. If a node exhibits an anomaly earlier in an abnormal path, the degree of anomaly in its downstream nodes is consistent with its propagation direction, or its edge weight mutation matches the actual propagation path, that node is assigned a higher probability of being a perturbation source. Finally, the third source tracing generator normalizes the probability values ​​of all nodes, generating a probability distribution map representing the probability of each functional node as an original perturbation source. This probability distribution serves as the output of the abnormal gait source tracing result, used for subsequent early warning prompts and intervention plan generation.

[0024] Furthermore, the functional information flow graph is input into the first-state determiner, and the first-state determination result is output, including: By scanning the functional information flow graph, a first side view and a second side view under symmetrical pose are determined; a consistency comparison is performed on the first side view and the second side view to determine one judgment result; by scanning the functional information flow graph, a first gait periodic view and a second gait periodic view on the same side are determined; a consistency comparison is performed on the first gait periodic view and the second gait periodic view to determine two judgment results; the one judgment result and the two judgment results are used as the first gait judgment result, wherein the consistency determination satisfies a preset tolerance range.

[0025] Optionally, in the first gait judge, after the functional information flow graph is input, a global scan of the graph is first performed. Based on the natural anatomical symmetry of the human lower limbs, the functional nodes belonging to the left lower limb and their information flow edges are extracted as the first side image, and the corresponding functional nodes belonging to the right lower limb and their information flow edges are extracted as the second side image. Each side image includes core joints (hip, knee, ankle), major muscle groups (quadriceps, hamstrings, calf muscles, etc.), and the transfer entropy weights and information flow directions between them. Under normal circumstances, the lower limbs should exhibit a balanced control mode in the biomechanical transmission cycle. That is, in the same gait phase, the information flow topology, main transmission path, and key transfer entropy values ​​of the left and right sides should remain approximately consistent; especially in the same movement mode, such as a gait mode with synchronized amplitude, the biomechanical function transmission of the left and right legs should exhibit similar amplitude levels and stability. Based on this, the first gait analyzer performs a consistency comparison between the first and second side graphs. By comparing the two graphs in terms of overall topological similarity, dominant path matching degree, and key edge weight differences, it determines whether the left-right symmetry remains within a preset tolerance range. If the deviation exceeds the allowable range, a corresponding judgment result is generated and the symmetry anomaly type is marked. Subsequently, the first gait analyzer scans the functional information flow graph again. Based on gait events, such as the time annotations of heel strike and toe lift, it divides the information flow graph of the same lower limb into multiple periodic segments within a complete gait cycle, thus obtaining the first gait periodic graph and the second gait periodic graph. Each periodic graph reflects the information flow structure of the lower limb on that side in a specific gait phase. In normal gait, there should be high self-similarity between consecutive gait cycles on the same side, that is, the graph structure, key edge weights, and information flow direction between cycles should remain basically stable. The first-stage determiner compares the consistency of the first-stage periodic graph and the second-stage periodic graph, calculating the differences between them in terms of graph structure similarity, dominant path matching degree, and key edge weights. If the variation during the period exceeds a preset tolerance range, the corresponding periodic anomaly information is recorded, forming a binary judgment result. Finally, the single-stage judgment result and the binary judgment result are merged to form a complete first-stage judgment result. This result not only includes whether a deviation exists, but also the type and degree of deviation, and the corresponding node or link label, providing a clear structural input for subsequent core disturbance mode judgment and source tracing analysis.

[0026] Furthermore, inputting the functional information flow graph into the second gait determiner and outputting the second gait determination result includes: According to the motion pattern, a functional information flow template is matched; according to the functional information flow template, the functional information flow graph is judged under the multi-core perturbation mode based on the transfer entropy to determine the graph structure perturbation mode; the graph structure perturbation mode is used as the second gait judgment result.

[0027] Optionally, in the second gait analyzer, firstly, based on the current movement pattern of the monitored object, such as natural walking, brisk walking, climbing stairs, or specific rehabilitation training movements, a functional information flow template corresponding to the movement pattern is retrieved from a pre-established functional information flow template library. This functional information flow template is a standardized representation of the lower limb functional transmission relationship, used to provide a baseline pattern of the information flow direction, intensity distribution, and key pathways between joint-muscle functional units under normal conditions. Subsequently, the functional information flow graph at the current moment or the current gait cycle is compared with the functional information flow template edge-by-edge and path-by-path: for edges marked as key pathways in the template, the second gait analyzer determines whether an information flow interruption has occurred by comparing the transmission entropy value of the corresponding edge in the current graph. That is, if the information flow weight of a certain key edge, such as the control pathway from the hip abduction angle to the downstream muscle group, is significantly weakened or nearly disappeared compared to the template, it indicates that the pathway is functionally disconnected or severely attenuated. Simultaneously, the gaiter also detects whether there is information flow compensation or enhancement. Specifically, it detects strong information flow edges not present in the template in the current functional information flow graph, or significant enhancement of edge weights that were originally secondary or non-primary pathways. For example, an abnormally enhanced reverse or reverse information flow transmission from the calf muscles to the thigh muscles indicates that the lower limb network compensates for functional failures through bypass or alternative pathways. Under normal physiological control logic, information flow interruption and compensation often occur synergistically. That is, the loss or weakening of a certain function will inevitably be maintained by enhancing the function of other pathways to ensure the completion of the overall motor task. The second gait gaiter comprehensively evaluates the synergistic relationship between interruption and compensation based on this. Subsequently, the second gait gaiter also determines the information flow oscillation. Specifically, by tracking the transfer entropy time series and edge weight change curves of key edges within a complete gait cycle or multiple consecutive time windows, if the weights of certain edges exhibit abnormally large fluctuations, frequent changes in direction, or amplitudes far exceeding the allowable fluctuation range in the template within the gait cycle, they are identified as information flow oscillations. This oscillation reflects instability in motor control or imbalance in central / peripheral regulation. Through the joint identification of three core disturbance patterns—information flow interruption, information flow compensation, and information flow oscillation—the second gait judge analyzes the overall disturbance state of the current lower limb network node by node and path by path at the functional information flow graph level. Finally, the obtained graph structure disturbance patterns are structured and encoded as the output of the second gait judgment result, providing the basic input for the subsequent source tracing module to infer the probability of the original disturbance source.

[0028] On the terminal interface of the monitoring platform, early warning and indication responses are given to the abnormal gait tracing results.

[0029] In one embodiment, after the third source generator outputs the abnormal gait source results corresponding to the current gait, the monitoring platform first reads the original disturbance source probability distribution and corresponding abnormality types (such as symmetry abnormality, periodic instability, information flow interruption, compensation, oscillation, etc.) of each functional node from the abnormal gait source results. Subsequently, the terminal interface calls the visualization module to map the probability distribution into intuitive graphical elements. For example, on the lower limb anatomical diagram or functional network diagram, high-probability original disturbance source nodes are highlighted by color depth, flashing status, or marker size, and the main propagation path of abnormal information flow is marked with lines or arrows. At the same time, the terminal interface triggers corresponding early warning response plans based on the periodic frequent items mining of the temporary database. For example, an early warning window automatically pops up, displaying the currently detected abnormal gait type, suspected original disturbance source location, main abnormal pathways, and possible risks, such as increased risk of fall, joint overload, etc., and can be supplemented with sound prompts, vibration prompts, or message pushes to remind medical staff or users to pay attention. Subsequently, the terminal interface generates an instruction response interface based on the abnormal gait tracing results. This interface provides instructions for the current abnormality through text descriptions, graphical markers, or interactive lists. Examples include, "Suspected weakening of hip abduction control, accompanied by compensatory strengthening of the calf muscles; it is recommended to focus on hip muscle strength and control training," and "Current gait cycle stability is decreased; it is recommended to reduce weight-bearing or walking speed." This allows medical staff or users to quickly understand the source of the abnormality and possible intervention directions based on the interface prompts. Both the warning and instruction responses mentioned above can be recorded and linked to the user's historical test data for easy follow-up and trend analysis.

[0030] Furthermore, the early warning and indication responses to the abnormal gait tracing results include: The abnormal gait tracing results are read and added to a temporary database; based on the temporary database, abnormal gait patterns are determined through periodic frequent item mining; and the abnormal gait patterns are used for early warning management and instruction management.

[0031] Preferably, after the monitoring platform obtains the abnormal gait tracing results of the current analysis, it first reads the original disturbance source probability distribution, corresponding abnormality type, abnormal link, and other structured information of each functional node from the tracing results, and encapsulates it into a record according to additional tags such as timestamp, user identifier, and gait task type, and writes it into a temporary database. The temporary database serves as a real-time analysis data cache, used to store the abnormal tracing results of multiple consecutive gait cycles or multiple monitoring sessions for trend analysis and pattern mining. After the data is stored, the monitoring platform performs periodic frequent item mining on the temporary database according to a preset scheduling strategy, such as after completing N gait acquisitions, at fixed time windows, or when a certain number of abnormal records are detected. During the data mining process, frequent pattern mining algorithms suitable for sequence data, such as Apriori and FP-Growth, are employed. Multiple abnormal records in the database are treated as itemsets composed of abnormal functional nodes, abnormal links, abnormal types, and perturbation patterns. The co-occurrence relationships, repetition frequencies, and correlations of various abnormal features in the time series are identified. For example, if the combination of "hip abduction link interruption + calf muscle group compensation enhancement" appears repeatedly in multiple consecutive gait cycles, this is categorized as a high-support abnormal pattern. If a certain type of oscillatory edge weight repeatedly appears over a long period, a corresponding long-term oscillatory abnormal pattern is generated. Finally, based on the frequently mined items, a structured abnormal gait pattern is automatically formed to characterize the current user's recent typical gait abnormality trend. After acquiring the abnormal gait pattern, the monitoring platform performs early warning and instruction management according to different pattern types and their severity. During this process, a corresponding contingency plan is matched from the contingency plan library based on the pattern type and severity as the first contingency plan. Based on this first contingency plan, a pop-up response is executed on the monitoring platform's terminal interface to achieve continuous monitoring of abnormal gait.

[0032] Furthermore, an early warning instruction is generated based on the pattern type and anomaly level; a first contingency plan is generated by performing a contingency plan library matching based on the abnormal gait pattern; and the early warning instruction and the first contingency plan are executed via a pop-up response on the terminal interface of the monitoring platform.

[0033] Optionally, the system first generates corresponding warning instructions based on the type and severity of the current abnormal gait pattern, and packages them into a structured warning message. Then, the monitoring platform selects or matches an applicable first plan from its built-in plan library based on the abnormal gait pattern and the warning instructions. This plan library stores various standardized response measures built upon expert experience, clinical logic, or rehabilitation guidelines, such as "training suggestions for weakened hip abduction control," "stability training programs for gait oscillations," and "load adjustment measures for compensatory abnormalities." The matching process automatically selects the plan that best matches the current abnormal state based on the abnormal gait pattern and the severity of the warning instructions through rule-based matching. If a complex abnormal pattern occurs, multiple related plans can be combined to generate a comprehensive first plan. For example, for a combination of "moderate periodic oscillations + mild multi-node compensation," a composite plan of "reducing exercise intensity + recommending balance training" might be generated. After generating the warning instructions and selecting the first plan, the monitoring platform pushes both to the terminal interface to execute a pop-up response. Upon detecting push notifications, the terminal interface automatically pops up an alert window. This window includes an anomaly mode title, a description of the main anomaly characteristics, a high-probability node indicator of the original disturbance source, and an anomaly level icon. Simultaneously, the content of the first contingency plan is displayed below. For high-level alerts, the terminal interface can simultaneously trigger audio-visual alerts, vibration reminders, or push instant messages to mobile devices, ensuring users can receive and respond promptly. The pop-up interface allows users to view detailed descriptions, expand contingency plan content, or jump to a historical anomaly record page, facilitating users' monitoring of their gait change trends and risk status. Through this process, the monitoring platform achieves a closed-loop execution of anomaly identification, risk warning, and targeted guidance, providing users with real-time and actionable gait health management support.

[0034] In summary, the embodiments of this application have at least the following technical effects: First, lower limb motion data is collected and uploaded to a monitoring platform based on a distributed joint sensor network. This lower limb motion data is then reconstructed into a functional information flow graph. Next, by activating the gait tracing module embedded in the monitoring platform, a two-stage joint judgment is performed on the functional information flow graph to determine the source of abnormal gait. The two-stage joint judgment uses symmetry comparison and self-comparison under gait cycles for the first judgment, and information flow anomalies under core perturbation modes for the second judgment. The probability distribution of the original perturbation source based on the judgment is used as the output. Finally, on the terminal interface of the monitoring platform, an early warning and indication response are provided for the abnormal gait tracing result. This solves the technical problems of existing technologies, such as difficulty in accurately locating the original perturbation joints of abnormal lower limb gait, high misjudgment rate, and ambiguous abnormal source location. It achieves the technical effect of accurate gait tracing based on a distributed joint sensor network, improving recognition accuracy and real-time early warning capabilities, and enabling rapid determination of lower limb functional abnormalities.

[0035] Example 2, based on the same inventive concept as the lower limb abnormal gait tracing method combined with a distributed joint sensor network in the previous examples, such as... Figure 2 As shown, this application provides a lower limb abnormal gait tracing system that incorporates a distributed joint sensing network, including: Data acquisition module 11: Based on the distributed joint sensor network, it collects lower limb motion data and uploads it to the monitoring platform, reconstructing the lower limb motion data into a functional information flow graph; Anomaly judgment module 12: By activating the gait tracing module embedded in the monitoring platform, it performs a two-way joint judgment on the functional information flow graph to determine the abnormal gait tracing result. The two-way joint judgment uses symmetry comparison and self-comparison under gait cycle for the first judgment, and information flow anomaly under the core perturbation mode for the second judgment, with the original perturbation source probability distribution based on the judgment as the output; Anomaly warning module 13: On the terminal interface of the monitoring platform, it provides warning and indication responses to the abnormal gait tracing result.

[0036] Furthermore, the data acquisition module 11 is used to perform the following methods: A distributed key sensor network is deployed based on the physical structure of the lower limbs, with major joints and key muscle segments as sensor deployment locations and micro-motion radar and inertial measurement units as sensor types. By driving the distributed joint sensor network, lower limb motion data is collected. Through communication interaction between the distributed joint sensor network and the monitoring platform data interface, data is transmitted back and reconstructed based on the transmission flow direction and intensity to generate the functional information flow graph.

[0037] Furthermore, the data acquisition module 11 is used to perform the following methods: Using joint-muscle functional units as nodes, a directed weighted graph is constructed based on the direction and intensity of information flow by introducing transfer entropy. The transfer entropy is defined by the transfer causality and transfer entropy value. The directed weighted graph is dynamically generated based on the time series to generate the functional information flow graph.

[0038] Furthermore, the anomaly detection module 12 is used to perform the following method: A first gait judge is deployed based on symmetry comparison and self-comparison under gait cycle; a second gait judge is deployed based on information flow anomalies under core perturbation mode, wherein the core perturbation mode includes at least information flow interruption, information flow compensation and information flow oscillation; a gait tracing module is constructed based on the first gait judge, the second gait judge and the third tracing generator.

[0039] Furthermore, the anomaly detection module 12 is used to perform the following method: The functional information flow graph is input into the first gait judge and the first gait judgment result is output. The functional information flow graph is input into the second gait judge and the second gait judgment result is output. The first gait judgment result and the second gait judgment result are input into the third source generator. By analyzing the path of disturbance propagation, the probability distribution is determined as the abnormal gait source result. The probability distribution indicates the probability value of each functional node as the original disturbance source.

[0040] Furthermore, the anomaly detection module 12 is used to perform the following method: By scanning the functional information flow graph, a first side view and a second side view under symmetrical pose are determined; a consistency comparison is performed on the first side view and the second side view to determine one judgment result; by scanning the functional information flow graph, a first gait periodic view and a second gait periodic view on the same side are determined; a consistency comparison is performed on the first gait periodic view and the second gait periodic view to determine two judgment results; the one judgment result and the two judgment results are used as the first gait judgment result, wherein the consistency determination satisfies a preset tolerance range.

[0041] Furthermore, the anomaly detection module 12 is used to perform the following method: According to the motion pattern, a functional information flow template is matched; according to the functional information flow template, the functional information flow graph is judged under the multi-core perturbation mode based on the transfer entropy to determine the graph structure perturbation mode; the graph structure perturbation mode is used as the second gait judgment result.

[0042] Furthermore, the anomaly warning module 13 is used to perform the following method: The abnormal gait tracing results are read and added to a temporary database; based on the temporary database, abnormal gait patterns are determined through periodic frequent item mining; and the abnormal gait patterns are used for early warning management and instruction management.

[0043] Furthermore, the anomaly warning module 13 is used to perform the following method: Based on the pattern type and anomaly level, an early warning instruction is generated; a first contingency plan is generated by matching the contingency plan library based on the abnormal gait pattern; and the early warning instruction and the first contingency plan are executed in a pop-up response through the terminal interface of the monitoring platform.

[0044] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for tracing abnormal gait in the lower limbs using a distributed joint sensor network, characterized in that, The method includes: Based on a distributed joint sensor network, lower limb motion data is collected and uploaded to the monitoring platform, and the lower limb motion data is reconstructed into a functional information flow graph. By activating the gait tracing module embedded in the monitoring platform, a two-way joint judgment is performed on the functional information flow graph to determine the abnormal gait tracing result. The two-way joint judgment is based on the first judgment of symmetry comparison and self-comparison under gait cycle, and the second judgment is based on the information flow abnormality under the core disturbance mode. The original disturbance source probability distribution based on the judgment is used as the output. On the terminal interface of the monitoring platform, early warning and indication responses are given to the abnormal gait tracing results.

2. The method for tracing abnormal lower limb gait using a distributed joint sensing network as described in claim 1, characterized in that, Lower limb movement data is collected and uploaded to the monitoring platform. The lower limb movement data is then reconstructed into a functional information flow graph, including: A distributed key sensor network is deployed based on the physical structure of the lower limbs, with major joints and key muscle segments as the sensor deployment locations and micro-motion radar and inertial measurement units as the sensor types. By driving the distributed joint sensing network, lower limb motion data is collected. Through communication interaction between the distributed joint sensing network and the monitoring platform data interface, data is transmitted back and reconstruction based on transmission flow direction and intensity is performed to generate the functional information flow graph.

3. The method for tracing abnormal lower limb gait using a distributed joint sensing network as described in claim 2, characterized in that, Perform reconstruction based on transmission direction and intensity, including: Using joint-muscle functional units as nodes, and by introducing transfer entropy, directed edges are determined based on the direction and intensity of information flow. A directed weighted graph is constructed based on the lower limb motion data, wherein the transfer entropy is defined by the transfer causality and the transfer entropy value. Based on the time series, the directed weighted graph is dynamically modified to generate the functional information flow graph.

4. The method for tracing abnormal lower limb gait using a distributed joint sensing network as described in claim 1, characterized in that, Before activating the gait tracing module embedded in the monitoring platform, the construction of the gait tracing module includes: A first-step discriminator is deployed using symmetric alignment and self-alignment under gait loops; A second gait detector is deployed based on information flow anomalies under the core disturbance mode, wherein the core disturbance mode includes at least information flow interruption, information flow compensation, and information flow oscillation; The first gait judge, the second gait judge, and the third gait generator constitute a gait tracing module.

5. The method for tracing abnormal lower limb gait using a distributed joint sensing network as described in claim 4, characterized in that, The functional information flow graph is subjected to a two-way joint judgment to determine the abnormal gait source tracing result, including: Input the functional information flow graph into the first-state determiner and output the first-state determination result; The functional information flow graph is input into the second gait determiner, and the second gait determination result is output. The first gait judgment result and the second gait judgment result are input into the third source generator. By analyzing the path of disturbance propagation, the probability distribution is determined as the abnormal gait source result. The probability distribution indicates the probability value of each functional node as the original disturbance source.

6. The method for tracing abnormal lower limb gait using a distributed joint sensing network as described in claim 5, characterized in that, The functional information flow graph is input into the first-state determiner, and the first-state determination result is output, including: By scanning the functional information flow graph, the first side view and the second side view under the symmetrical pose are determined; A consistency comparison is performed between the first side image and the second side image to determine a judgment result; By scanning the functional information flow graph, the first gait periodic graph and the second gait periodic graph on the same side are determined; The consistency of the first step periodic diagram and the second step periodic diagram is compared to determine the two judgment results; The result of one judgment and the result of two judgments are used as the first state judgment result, wherein the consistency determination satisfies the preset tolerance range.

7. The method for tracing abnormal lower limb gait using a distributed joint sensing network as described in claim 5, characterized in that, The functional information flow graph is input into the second gait determiner, and the second gait determination result is output, including: Match functional information flow templates based on the motion pattern; Based on the functional information flow template, the functional information flow graph is judged under the multi-core perturbation mode based on the transfer entropy to determine the graph structure perturbation mode. The graph structure perturbation pattern is used as the second gait judgment result.

8. The method for tracing abnormal lower limb gait using a distributed joint sensing network as described in claim 1, characterized in that, The abnormal gait tracing results will trigger both an early warning response and an indication response, including: Read the abnormal gait tracing results and add them to the temporary database; Based on the temporary database, abnormal gait patterns are identified through periodic frequent item mining; Early warning and instruction management are performed based on the aforementioned abnormal gait patterns.

9. The method for tracing abnormal lower limb gait using a distributed joint sensing network as described in claim 8, characterized in that, Generate early warning instructions based on the pattern type and anomaly level; A first contingency plan is generated by performing a contingency plan library matching based on abnormal gait patterns; By monitoring the terminal interface of the platform, the warning command and the pop-up response of the first contingency plan are executed.

10. A lower limb abnormal gait tracing system incorporating a distributed joint sensor network, characterized in that, The system is used to implement the lower limb abnormal gait tracing method incorporating a distributed joint sensing network as described in any one of claims 1-9, the system comprising: Data acquisition module: Based on the distributed joint sensor network, it collects lower limb motion data and uploads it to the monitoring platform, and reconstructs the lower limb motion data into a functional information flow graph; Anomaly Detection Module: By activating the gait tracing module embedded in the monitoring platform, the functional information flow graph is subjected to a two-way joint judgment to determine the abnormal gait tracing result. The two-way joint judgment is based on the first judgment of symmetry comparison and self-comparison under gait cycle, and the second judgment is based on the information flow anomaly under the core disturbance mode. The original disturbance source probability distribution based on the judgment is used as the output. Anomaly warning module: On the terminal interface of the monitoring platform, it provides warning and indication responses to the abnormal gait tracing results.

Images