A patient fall early warning method and system based on pressure detection

Abstract

The application provides a patient falling early warning method and system based on pressure detection, the method comprising: conducting filter calibration and biological force partition mapping on plantar array pressure data, constructing a force flow topology network and generating a time sequence; extracting standard topology connection structure, node conduction time delay and closed loop stability parameters, and establishing a steady state force flow baseline model; real-time calculating topology edge fracture degree, conduction time sequence distortion rate and asymmetric broken propagation speed, obtaining a falling risk deduction value through topology broken coupling deduction, and outputting a graded early warning signal. The application can predict instability risk in advance, improve the recognition accuracy of complex falling scenes, and is suitable for inpatient fall protection.

Description

Technical Field

[0001] This invention relates to the field of human posture early warning technology, and in particular to a patient fall early warning method and system based on pressure detection. Background Technology

[0002] In clinical nursing settings, elderly patients, postoperative patients in recovery, and patients with limited mobility are at higher risk of falling. Falls can easily cause secondary injuries such as fractures and soft tissue contusions, and can also prolong the treatment period and aggravate the condition. This is a key safety issue that clinical nursing care needs to control.

[0003] Existing fall warning solutions based on plantar pressure typically collect pressure data at various points on the soles of the feet using pressure sensors. They then calculate indicators such as the degree of shift in the body's center of gravity and the magnitude of pressure changes. When these values ​​exceed preset thresholds, a fall risk is identified, and an alarm is triggered. While these solutions are simple to implement and have low deployment costs, they have significant shortcomings in practical applications.

[0004] First, the recognition accuracy is insufficient, with prominent issues of false alarms and missed alarms. Everyday actions such as sitting down, squatting, and bending over can also cause a shift in the center of gravity and changes in pressure, which can easily be misjudged as a risk of falling. For "soft falls" such as slow slips and fainting, which do not involve violent pressure impacts, the pressure values ​​change gradually and do not reach the preset threshold, making it easy to miss the detection and failing to effectively cover all fall scenarios.

[0005] Secondly, there is poor individual adaptability, and long-term use will lead to a decrease in accuracy. Most existing solutions use a uniform universal threshold, without taking into account the differences in weight, gait, and limb function among different patients, and cannot automatically adapt to environmental changes such as shoe and sock thickness and ground hardness; data drift caused by long-term use of sensors cannot be automatically corrected, which will cause the warning accuracy to gradually decrease over time. Summary of the Invention

[0006] The purpose of this invention is to provide a patient fall warning method and system based on pressure detection, so as to improve the foresight, accuracy and individual adaptability of fall warning.

[0007] To achieve the above objectives, the present invention provides the following solution: A patient fall early warning method based on pressure detection includes the following steps: The pressure data collected by the plantar array pressure sensor is subjected to conduction filtering calibration and biomechanical partition mapping to construct the output force flow topology network and generate a time sequence according to the sampling time sequence; Based on the time sequence corresponding to the patient's daily steady-state actions, the standard topological connection structure, standard conduction delay between nodes, and closed-loop stability parameters are determined to obtain the steady-state force flow baseline model. The real-time time series is tracked and compared frame by frame, and the topological instability characteristic parameters are calculated. The topological instability characteristic parameters include: topological edge breakage degree, propagation time series distortion rate and asymmetric failure propagation speed. The topological instability characteristic parameters are substituted into the steady-state force-fluid baseline model for matching, and topological breaking coupling is performed to calculate the fall risk estimate. Risk warnings are generated based on the estimated fall risk value, resulting in tiered warning signals.

[0008] Optionally, the pressure data collected by the plantar array pressure sensor is subjected to conduction filtering calibration and biomechanical partitioning mapping to construct an output force flow topology network, and a time series is generated according to the sampling time sequence, including: Pressure data is spatially registered with the plantar biomechanical coordinate system based on plantar anatomical landmarks to obtain pressure distribution data; plantar anatomical landmarks include: heel pressure extreme point, first metatarsal pressure extreme point, and fifth metatarsal pressure extreme point; The pressure distribution data is divided into multiple load-bearing zones by using the biomechanical transmission pathway of the sole of the foot; The defect confidence level is calculated based on the deviation components of the load-bearing zones, and load-bearing zones with defect confidence levels lower than the preset confidence level are eliminated. The deviation components include: spatial amplitude deviation, temporal transmission deviation, and zone coupling deviation. The load-bearing zones include: heel rear load-bearing zone, heel front transition zone, arch support zone, forefoot main load-bearing zone, and toe auxiliary load-bearing zone. Using the load-bearing partition as the topology node and the pressure transmission gradient direction and transmission weight between the load-bearing partitions as directed edges, a single-frame force flow topology network is constructed. The force flow topology networks of all sampling times are then spliced ​​together in chronological order to obtain a time series sequence.

[0009] Optionally, based on the time sequence corresponding to the patient's daily steady-state actions, the standard topological connection structure, standard conduction delay between nodes, and closed-loop stability parameters are determined to obtain a steady-state force flow baseline model, including: The standard connection weight of each directed edge is calculated based on the frequency of occurrence of each directed transmission edge in the time sequence corresponding to the patient's daily steady-state actions. The conduction delay data of pressure peak values ​​between adjacent nodes in the time series are extracted sample by sample, and kernel density estimation is performed to obtain the main conduction delay; The conduction direction consistency coefficient is obtained based on the proportion of samples that conform to the standard conduction direction on the conduction path of the main conduction delay. The main conduction delay is corrected based on the conduction direction consistency coefficient to obtain the standard conduction delay between nodes; The average coefficient of variation of all directed edge weights is calculated based on the standard propagation delay between nodes to obtain the topological stability components. The average coefficient of variation of all propagation path delays is calculated based on the standard propagation delay between nodes to obtain the temporal propagation stability component. The bipedal weight symmetry deviation rate is calculated based on the standard connection weights of the corresponding conduction paths of the left and right feet, and the bipedal symmetry stability components are obtained. Multiplying the topological stability component, the time-series propagation stability component, and the bipedal symmetric stability component yields the closed-loop stability parameters. By encapsulating and integrating the standard topology connection structure, standard inter-node propagation delay, and closed-loop stability parameters, a steady-state force flow baseline model is obtained.

[0010] Optionally, the real-time time series is tracked and compared frame by frame, and topological instability characteristic parameters are calculated, including: In the real-time time series, directed edges whose weight decay rate exceeds the preset breakage threshold are marked as broken edges, and the standard weights and weight decay rates of all broken edges are obtained. The proportion of broken edges on the main transmission link from heel to toe is obtained as the ratio of the number of broken edges on the main link to the total number of edges on the main link. The weighted sum of the weighted attenuation rates is exponentially coupled with the proportion of main link breaks to obtain the topological edge breakage degree of the current frame; the formula for calculating the topological edge breakage degree is: ,in For topological edge breakage, This is the set of broken edges in the current frame. Let be the standard connection weight corresponding to the i-th fracture edge. Let be the weight decay rate of the i-th fracture edge. The total connection weights of the steady-state baseline topology network. The number of broken edges on the main transmission link. This is the total number of edges in the main transmission link.

[0011] Optionally, the process of tracking and comparing real-time time-series sequences frame by frame and calculating topological instability characteristic parameters also includes: The actual propagation delay of each propagation path in the time sequence is extracted frame by frame, and the difference normalization calculation is performed with the standard propagation delay of the corresponding path to obtain the single path delay deviation rate. The proportion of reversed direction paths to the total number of conduction paths is calculated based on the number of paths where the conduction direction is reversed, thus obtaining the proportion of reversed direction paths. The single-path delay deviation rate is coupled with the proportion of reversed-direction paths to obtain the propagation timing distortion rate of the current frame; the formula for calculating the propagation timing distortion rate is: ,in This represents the conduction timing distortion rate. This represents the total number of propagation paths in the topology network. Let j be the actual propagation delay of the j-th propagation path. Let j be the standard propagation delay of the j-th propagation path. This represents the percentage of paths with reversed direction.

[0012] Optionally, the process of tracking and comparing real-time time-series sequences frame by frame and calculating topological instability characteristic parameters also includes: Based on the starting node and propagation path of the topological break in the time series, determine the time duration for the break to propagate from the starting node to all bearing partitions of the entire system, and calculate the average propagation speed of the break by combining the spatial transmission distance of the topological nodes. Calculate the topological edge fracture degree of the left and right feet respectively to obtain the difference in the degree of damage between the two feet; The asymmetric malfunction propagation velocity is obtained by exponentially coupling the average propagation velocity with the bipedal malfunction difference; the formula for calculating the asymmetric malfunction propagation velocity is as follows: ,in For the propagation speed of asymmetric sharding, The average propagation velocity of the break within the initial foot, The topological edge breakage degree of the left foot. The topological edge breakage degree of the right foot.

[0013] Optionally, the topological instability characteristic parameters are substituted into the steady-state force-flow baseline model for matching, and topological breaking coupling deduction is performed to calculate the fall risk deduction value, including: The topological instability characteristic parameters are matched class by class with the preset hierarchical threshold intervals in the steady-state force flow baseline model to obtain the current topological breaking evolution stage and the evolution gain coefficient of the corresponding stage. Divide the topological edge fracture degree in the steady-state force flow baseline model by the topological stability component to obtain the structural instability component; Divide the conduction time-series distortion rate in the steady-state power flow baseline model by the time-series conduction stability component to obtain the time-series instability component. The evolutionary gain coefficient, structural instability component, and temporal instability component are coupled and calculated to obtain the fall risk projection value; the calculation formula for the fall risk projection value is as follows: ,in For the projected risk value of falling, This represents the evolution gain coefficient corresponding to the current topology-breaking evolution stage. For topological edge breakage, This represents the conduction timing distortion rate. These are the closed-loop stability parameters in the steady-state force-flow baseline model. For the propagation speed of asymmetric sharding, This is the preset standard transmission velocity of force flow in the steady-state baseline model.

[0014] Optionally, the topological instability characteristic parameters are matched class-by-class with the preset hierarchical threshold intervals in the steady-state force-flow baseline model to obtain the current topological breaking evolution stage and the evolution gain coefficient of the corresponding stage, including: When the topological edge breakage and the propagation time-series distortion rate are both within the baseline normal fluctuation threshold, and the asymmetric spur propagation speed is zero, it is determined to be the topological perturbation stage, and the corresponding evolution gain coefficient is taken as the baseline value. When any two topological instability characteristic parameters exceed the baseline normal fluctuation threshold and the defect range is limited to a single-leg single-bearing zone, it is determined to be a local defect stage, and the corresponding evolution gain coefficient is amplified by a preset first multiple based on the benchmark value. When all three topological instability characteristic parameters exceed the baseline critical threshold, and the breach exhibits cross-regional diffusion or bipedal asymmetric mutation, it is determined to be in the global instability stage, and the corresponding evolution gain coefficient is amplified by a preset second factor based on the local breach stage.

[0015] Optionally, risk warnings are generated based on the fall risk projection value, resulting in tiered warning signals, including: When the matching is in the topological perturbation stage and the fall risk projection value is within the first-level threshold range, a first-level warning is triggered, and a vibration alert is sent to the patient's wearing terminal. When the matching is in the partial failure stage and the fall risk projection value is in the secondary threshold range, a secondary alarm is triggered, an audio-visual reminder is sent to the patient, and an alarm notification is pushed to the nursing terminal. When the match is in the global instability stage and the fall risk projection value is in the third-level threshold range, a third-level emergency warning is triggered, the highest level of audible and visual alarm is activated, and the patient's location is sent to the nursing system.

[0016] A pressure detection-based patient fall warning system, used to implement the aforementioned pressure detection-based patient fall warning method, includes: The topology construction module is used to perform conduction filtering calibration and biomechanical partitioning mapping on the pressure data collected by the plantar array pressure sensor, construct the output force flow topology network, and generate a time sequence according to the sampling time sequence; The parameter base generation module is used to determine the standard topological connection structure, standard conduction delay between nodes and closed-loop stability parameters based on the time sequence corresponding to the patient's daily steady-state actions, so as to obtain the steady-state force flow baseline model. The feature parameter calculation module is used to track and compare real-time time series frames by frame and calculate topological instability feature parameters. The topological instability feature parameters include: topological edge breakage degree, propagation time series distortion rate and asymmetric failure propagation speed. The fall risk prediction module is used to substitute the topological instability characteristic parameters into the steady-state force flow baseline model for matching, and to perform topological breaking coupling inference to calculate the fall risk inference value. The risk warning module is used to issue risk warnings based on the fall risk projection value and obtain graded warning signals.

[0017] According to specific embodiments provided by the present invention, the following technical effects are disclosed: The patient fall warning method and system based on pressure detection provided by the present invention, by constructing a plantar force flow topology network, identifies instability signs from the evolution dimension of the force flow transmission structure, and can complete risk prediction in the early stage of instability before the center of gravity shifts significantly, effectively advancing the warning time and reserving sufficient time for patient self-adjustment and nursing intervention; through nonlinear coupling deduction of three types of features, namely topological edge fracture degree, transmission time sequence distortion rate, and asymmetric failure propagation speed, it can accurately distinguish between daily actions such as sitting and squatting and real instability states, and has a good recognition effect on soft fall scenarios without severe pressure impact, such as slow slips and fainting, significantly reducing the probability of false alarms and missed alarms; at the same time, by constructing an individual-specific steady-state force flow baseline model, it can adapt to the differences in weight and gait of different patients, and can dynamically update the baseline through steady-state data to offset the influence of environmental interference such as shoes, socks, ground and sensor drift, ensuring the stability of the warning for long-term use, and better meeting the actual needs of clinical patients for fall protection. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the 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.

[0019] Figure 1 This is a flowchart of a patient fall early warning method based on pressure detection according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the biomechanical bearing zones of the foot in an embodiment of the present invention; Figure 3 This is a bar chart showing the standard connection weights of directed edges according to an embodiment of the present invention. Figure 4 The closed-loop stability parameters and three-component radar diagrams of this invention are shown in the embodiments of the present invention. Figure 5 This is a schematic diagram of the evolution curve of topological edge fracture degree in an embodiment of the present invention; Figure 6 This is a schematic diagram of the evolution curve of the conduction timing distortion rate according to an embodiment of the present invention; Figure 7This is a schematic diagram of the evolution curve of the asymmetric spurious propagation velocity in an embodiment of the present invention; Figure 8 This is a schematic diagram illustrating the evolution of fall risk projection values ​​and the triggering of a three-level early warning system according to an embodiment of the present invention. Figure 9 This is a schematic diagram of the patient fall warning system based on pressure detection according to an embodiment of the present invention. Detailed Implementation

[0020] 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.

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] like Figure 1 As shown, this embodiment of the invention provides a patient fall warning method based on pressure detection, including the following steps: Step 100: Perform conduction filtering calibration and biomechanical partition mapping on the pressure data collected by the plantar array pressure sensor, construct the output force flow topology network, and generate a time sequence according to the sampling time sequence; Step 200: Based on the time sequence corresponding to the patient's daily steady-state actions, determine the standard topological connection structure, standard conduction delay between nodes, and closed-loop stability parameters to obtain the steady-state force flow baseline model; Step 300: Track and compare the real-time time series frame by frame, and calculate the topological instability characteristic parameters; the topological instability characteristic parameters include: topological edge breakage degree, propagation time series distortion rate and asymmetric failure propagation speed; Step 400: Substitute the topological instability characteristic parameters into the steady-state force-fluid baseline model for matching, and perform topological breaking coupling deduction to calculate the fall risk deduction value; Step 500: Based on the fall risk projection value, conduct risk warning and obtain graded warning signals.

[0023] In this embodiment, step 100 uses a matrix-type flexible piezoresistive sensor to collect plantar pressure. The raw data of a single frame includes the pressure amplitude and pixel coordinates of each sensing unit. First, adaptive filtering is performed based on the continuity of force flow transmission. The first step is zero-point baseline drift correction, which takes the average pressure of each sensing unit under no-load conditions for 10 consecutive frames as the zero-point reference and subtracts the zero-point offset frame by frame. The second step is to construct an 8-neighborhood transmission constraint. For each sensing unit, the average pressure change rate of its 8 neighboring sensing units is calculated. When the single-frame pressure change rate of the target sensing unit exceeds 2.5 times the neighborhood average, it is determined to be transmission jump noise, and linear interpolation in the neighborhood pressure gradient direction is used for correction.

[0024] Next, functional zones are divided based on the biomechanical conduction pathways of the foot. Specifically, such as... Figure 2 As shown, the extreme points of heel pressure (i.e., the maximum pressure point of the whole foot, corresponding to the weight-bearing center of the heel bone), the extreme point of first metatarsal pressure (i.e. the peak pressure point on the medial side of the forefoot), and the extreme point of fifth metatarsal pressure (i.e. the peak pressure point on the lateral side of the forefoot) are located in the single-frame pressure distribution.

[0025] Using the extreme point of the heel as the origin and the angle bisector of the line pointing from the extreme point of the heel to the extreme points of the first and fifth metatarsal bones as the vertical axis, a two-dimensional biomechanical coordinate system for the sole of the foot is established. The original pixel coordinates of all sensing units are mapped to this coordinate system to complete the coordinate normalization. The normalization range of the vertical axis is 0~1, corresponding to the heel to the toe, and the horizontal axis is -0.5~0.5, corresponding to the outer side of the foot to the inner side of the foot.

[0026] Preferably, the pressure distribution data is divided into five load-bearing zones according to the force transmission path: the heel bearing zone (corresponding to the longitudinal axis 0-0.2) is the initial force bearing and transmission starting point; the heel transition zone (corresponding to the longitudinal axis 0.2-0.35) is the buffer transition section for force transmission from the heel to the arch; the arch support zone (corresponding to the longitudinal axis 0.35-0.6) corresponds to the force dispersion structure of the longitudinal arch; the forefoot main bearing zone (corresponding to the longitudinal axis 0.6-0.85) is the core bearing structure during the push-off phase; and the toe auxiliary bearing zone (corresponding to the longitudinal axis 0.85-1.0) is the auxiliary weight-bearing structure at the end of the gait. Finally, the comprehensive pressure value of each bearing zone is calculated by weighted summation of the area of ​​the sensing units within the zone. The closer the sensing unit is to the center of the zone's transmission, the higher its weight. In this embodiment, the weight of the center unit of the forefoot main bearing zone is 1.2, and the weight of the edge unit is 0.8.

[0027] Furthermore, for each bearing partition, three types of deviation components are calculated. Specifically, the first type is spatial amplitude deviation. In the calculation, the arithmetic mean of the pressure values ​​of all sensing units in the partition is first obtained, and then the sample standard deviation of the corresponding pressure values ​​is calculated. Finally, the ratio of the sample standard deviation to the arithmetic mean is used as the spatial amplitude deviation, which is used to characterize the degree of discrete disorder of the pressure distribution within the partition. The second type is temporal conduction deviation. The conduction delay of the pressure peak of the partition between two adjacent frames is obtained to obtain a set of delay samples. Then, the coefficient of variation of the set of delay samples (in this embodiment, the ratio of the delay standard deviation to the delay mean) is calculated to reflect the temporal stability of the force flow through the partition. The third type is partition coupling deviation. The pressure temporal Pearson correlation coefficient between the current partition and the upstream adjacent partition and the downstream adjacent partition is calculated. The arithmetic mean of the two correlation coefficients is taken, and finally, 1 divided by the average correlation coefficient is used as the partition coupling deviation.

[0028] Then, the defect confidence level is calculated based on the three types of biases. The calculation formula is as follows: ; in, Spatial amplitude deviation, For timing propagation bias, For partition coupling deviation, The deviation weighting coefficients are 0.4, 0.35, and 0.25 in this embodiment.

[0029] In this embodiment, a preset confidence threshold is set to 0.6. When the confidence of a partition defect is lower than the threshold, it is determined to be an invalid bearer partition and is removed from the topology node. The corresponding stress data is merged into the adjacent valid partition. The specific data is shown in Table 1.

[0030] Table 1. Examples of load-bearing zones for single-leg walking under steady-state conditions.

[0031] Finally, the discrete pressure data is transformed into structured topological data, and the selected effective bearing areas are used as topological nodes. Node attributes include the comprehensive pressure value of the area, area weight, and spatial center coordinates. Directed edges are constructed based on the pressure transmission relationships between areas, with the transmission direction determined by the pressure gradient direction, i.e., from high-pressure areas to low-pressure areas. The formula for calculating the transmission weight is: ; in, The weight is propagated to the directed edge from node i to node j. Let be the total pressure value of node i. Let be the comprehensive pressure value of node j, and satisfy . > , Let be the spatial center distance between node i and node j. The conduction direction consistency coefficient ranges from 0.3 to 1.0. It is determined by the angle between the conduction direction and the main conduction direction from the heel to the toe. The smaller the angle, the higher the coefficient. It is 1.0 when the angle is 0° and 0.3 when the angle is 90°.

[0032] Specifically, all valid nodes and directed edges within a single frame together constitute the force-flow topology network at that sampling moment, containing four types of data: node list, directed edge list, edge weights, and node pressure values. Following the sampling time sequence, the single-frame force-flow topology networks from consecutive sampling moments are sequentially spliced ​​together, retaining the timestamp of each frame to form a complete topology time sequence.

[0033] For example, in steady-state walking, the five effective zones of a single foot form a total of eight directed edges. Among them, the heel bearing area → heel transition area → arch support area → forefoot main bearing area → toe auxiliary bearing area are the main transmission links, with a total of four edges, accounting for more than 75% of the total weight. A continuous one-second time sequence contains 100 frames of topology data.

[0034] It should be noted that by using adaptive filtering calibration based on the continuity of force flow transmission, combined with biomechanical functional zoning mapping of plantar anatomical landmarks, and integrating a dynamic screening mechanism for defect confidence based on three types of deviations, as well as constructing a force flow topology network and generating a time series with effective bearing zones as topological nodes and pressure transmission gradients and directional weights as directed edges, this solves the inherent limitations of traditional plantar pressure early warning schemes that directly use discrete pressure point values ​​and uniform grid zoning, resulting in loss of force flow transmission characteristics, poor adaptability to individual foot morphology, and large single-point noise interference. It effectively eliminates sensor temperature drift creep and single-point jump noise, fully preserves the temporal continuity and directional characteristics of force flow transmission, and can adaptively adapt to different physiological states such as flat feet, high arches, and postoperative foot deformities.

[0035] In the specific implementation process, step 200 first collects the plantar pressure topology time series data of patients under typical steady-state actions such as walking slowly on flat ground, standing statically, and turning slowly. The initial baseline collection time for a single patient is set to 5 minutes to obtain steady-state samples.

[0036] like Figure 3 As shown, directed transmission edges appearing in all steady-state sample frames are extracted. The frequency of each directed edge in all steady-state sample frames is counted, and the arithmetic mean of the transmission weights in all sample frames where the edge appears is calculated. The average transmission weight of a single edge is multiplied by its frequency to obtain the standard connection weight of that edge. A weight threshold of 5% of the total standard weights is set, and directed edges with standard weights higher than the threshold are retained to form the standard topology connection structure.

[0037] For each pair of adjacent topological nodes, the time difference between the pressure peak of the upstream node and the pressure peak of the downstream node is extracted periodically from all steady-state gait periodic samples to form a single-path delay sample set. Then, the kernel density of the delay sample set is estimated by using a Gaussian kernel function. In this embodiment, the kernel bandwidth is set to 1.5 times the sampling interval. The delay value corresponding to the probability density peak is taken as the main conduction delay, which can effectively eliminate the interference of occasional abnormal delay samples. The robustness of the result is significantly better than that of the arithmetic mean method.

[0038] The percentage of samples whose conduction direction is consistent with the standard heel-toe main direction is calculated from all samples along the path. The conduction direction consistency coefficient is obtained. The main conduction delay is divided by the conduction direction consistency coefficient to complete the correction, and the standard conduction delay between nodes is obtained.

[0039] Furthermore, such as Figure 4 As shown, the topology stability component is calculated as follows: For each standard directed edge, the coefficient of variation of its weight in all steady-state samples is calculated (in this embodiment, it is the ratio of the standard deviation to the mean). The average value of all edge coefficients of variation is taken, and the topology stability component is obtained by converting it to 1 / (1+average coefficient of variation). The closer the value is to 1, the more stable the topology is.

[0040] Simultaneously, the temporal propagation stability component is calculated: for each propagation path, the coefficient of variation of its time delay sample is calculated, and the average value of the coefficients of variation of all paths is taken. Similarly, the temporal propagation stability component is obtained.

[0041] Next, the bipedal symmetric stability components are calculated: the relative deviation rate of the standard connection weights of the conduction paths corresponding to the left and right feet (i.e., the ratio of the difference between the left and right weights to the mean of the left and right weights) is calculated one by one. The average deviation rate of all paths is taken, and the bipedal symmetric stability components are obtained by subtracting the average deviation rate from 1. The three components are multiplied together to obtain the closed-loop stability parameters. Finally, the standard topological connection structure, the standard conduction delay between nodes of all paths, the closed-loop stability parameters, and their three sub-components are structurally encapsulated to form a patient-specific steady-state force flow baseline model. The parameters of the patient-specific steady-state force flow baseline model in this embodiment are shown in Tables 2 to 4.

[0042] Table 2. Standard Connection Weights of Directed Edges

[0043] Table 3 Standard conduction delay data between nodes along each conduction path

[0044] Table 4 Closed-loop stability parameters and three-component data

[0045] It should be noted that by constructing a dedicated steady-state force flow baseline model that perfectly matches the individual biomechanical characteristics of patients, the inherent defects of existing general threshold schemes, such as poor individual adaptability, high false alarm and false negative rates, and long-term accuracy decay, are solved. The high-quality steady-state samples based on dual-constraint screening also ensure the reliability of the benchmark. Combined with standard topological quantization of occurrence frequency and conduction weight, standard time delay calculation of kernel density estimation plus direction correction, and multi-dimensional coupled closed-loop stability parameter design, the normal force flow conduction law of patients is accurately characterized in all aspects. It provides patients with different weights, gaits, and foot physiological states with the same accuracy of early warning adaptability, effectively improving the individual adaptability and long-term stability of the overall early warning scheme.

[0046] like Figures 5 to 7 As shown, in this embodiment, step 300 first performs frame-by-frame quantization calculation of the topological edge breakage. Specifically, for the real-time single-frame force-flow topology network, the real-time transmission weight of each directed edge is compared one by one with the corresponding standard connection weight in the baseline model, and the weight decay rate of a single edge is calculated, that is, the difference between the standard weight and the real-time weight is divided by the standard weight.

[0047] In this embodiment, the preset fracture threshold is 0.3. When the weighted attenuation rate of a single edge is greater than or equal to 30%, it is marked as a fractured edge and included in the fractured edge set. The main conduction link from the heel bearing area to the toe auxiliary bearing area is located, and the proportion of fractured edges on this link to the total number of edges in the main link is counted to obtain the fracture ratio of the main link. Then, the weighted attenuation total of fractured edges is exponentially coupled with the fracture ratio of the main link, which considers the weighted impact of global fractures and amplifies the core instability contribution of the main conduction link fracture through the exponential term.

[0048] Specifically, the formula for calculating the topological edge breakage is: ; in, For topological edge breakage, This is the set of broken edges in the current frame. Let be the standard connection weight corresponding to the i-th fracture edge. Let be the weight decay rate of the i-th fracture edge. The total connection weights of the steady-state baseline topology network. The number of broken edges on the main transmission link. This is the total number of edges in the main transmission link.

[0049] In this embodiment, the coupled calculation of the conduction timing distortion rate is then performed to characterize the degree of temporal disorder in force flow conduction from two dimensions: time delay deviation and direction reversal. Specifically, the single-path time delay deviation rate is calculated path by path. For each conduction path, the actual conduction delay of that path within the current gait cycle is extracted, i.e., the time difference between the time when the pressure peak occurs at the upstream node and the time when the pressure peak occurs at the downstream node. The absolute value of the difference is calculated with the standard conduction delay of the corresponding path in the baseline model, and then divided by the standard conduction delay to obtain the single-path time delay deviation rate. Then, the arithmetic mean of the single-path time delay deviation rates of all conduction paths is taken to obtain the global average time delay deviation rate.

[0050] Further statistical analysis is conducted on the proportion of direction-reversed paths. Each path is compared with the baseline standard propagation direction; if the two directions are opposite, it is marked as a direction-reversed path. The proportion of direction-reversed paths to the total number of propagation paths is calculated. Furthermore, by combining delay deviation and direction reversal through product coupling, the propagation timing distortion rate of the current frame is obtained. This allows the instability in dual-distortion scenarios to be synergistically amplified, effectively identifying implicit timing disturbances that are difficult to capture with a single indicator.

[0051] Specifically, the formula for calculating the conduction timing distortion rate is: ; in, This represents the conduction timing distortion rate. This represents the total number of propagation paths in the topology network. Let j be the actual propagation delay of the j-th propagation path. Let j be the standard propagation delay of the j-th propagation path. This represents the percentage of paths with reversed direction.

[0052] In this embodiment, the asymmetric failure propagation velocity is finally quantified. This parameter incorporates the spatiotemporal evolution characteristics and bipedal symmetry of the failure, and can capture dynamic diffusion signs before instability occurs. Specifically, the failure initiation node is tracked in consecutive time frames, and the bearing partition where the first edge fracture and abnormal weight decay occur is located as the failure initiation point. Subsequently, the failure propagation process is continuously tracked along the force flow transmission path, and the total time from the occurrence of failure at the initiation node to the occurrence of fracture edges in all bearing partitions of the entire foot is recorded. Combined with the sum of the spatial transmission distances from the initiation node to each partition, the average propagation velocity of the failure within a single foot is calculated.

[0053] Further, the topological edge breakage degree of the left and right feet in the current frame is calculated separately, and the absolute value of the difference between the two is taken to obtain the difference in the degree of breakage of both feet; and the average propagation velocity of a single foot is exponentially coupled with the difference in the degree of breakage of both feet to obtain the asymmetric breakage propagation velocity. The higher the asymmetry of the breakage of both feet, the more the instability contribution of the propagation velocity is amplified exponentially.

[0054] Specifically, the formula for calculating the propagation velocity of asymmetric breaking is: ; in, For the propagation speed of asymmetric breaking, The average propagation velocity of the break within the initial foot, The topological edge breakage degree of the left foot. The topological edge fracture degree of the right foot. Specific instability characteristic parameters for this embodiment are shown in Table 5.

[0055] Table 5 Keyframe Topological Instability Feature Parameters for Slow Sideslip Events

[0056] It should be noted that by constructing a topological instability feature system combining structure, time, and symmetry, the degree of topological edge fracture accurately characterizes the degree of damage to the force-bearing structure through the exponential coupling of the main link fracture; the conduction time distortion rate integrates time delay deviation and direction reversal, enabling the capture of the disordered state of the conduction law; and the asymmetric failure propagation speed introduces spatiotemporal evolution and bipedal asymmetry characteristics, achieving advanced perception of early signs of instability. These three types of features complement and verify each other, enabling the system to not only keenly identify soft fall scenarios with gradual pressure changes, such as slow slips and fainting collapses, but also effectively distinguish the differences in topological evolution between daily actions and true instability, significantly reducing the probability of false alarms and false negatives.

[0057] In this embodiment, step 400 pre-sets a two-level threshold system based on the individual steady-state force flow baseline model. The normal fluctuation threshold is the upper limit of the 95% confidence interval of the corresponding instability characteristic parameter in the patient's steady-state sample, and the critical threshold is the upper limit of the 99% confidence interval. Frame by frame, the real-time calculated topological edge fracturing degree, conduction temporal distortion rate, and asymmetric defect propagation velocity are matched with the corresponding thresholds, and the stage determination is completed by combining the defect diffusion range obtained from temporal tracking.

[0058] Specifically, when the topological edge breakage degree and the propagation time-series distortion rate are both within the normal fluctuation threshold and the asymmetric spurious propagation speed is zero, it is determined to be the topological perturbation stage, and the evolution gain coefficient is taken as the baseline value of 1.0. When any two unstable characteristic parameters exceed the normal fluctuation threshold and the defect range is limited to a single foot and a single bearing zone, it is determined to be a local defect stage, and the evolution gain coefficient is amplified by a preset first multiplier of 1.8 times based on the benchmark value. When all three instability characteristic parameters exceed the critical threshold and the defect exhibits cross-regional diffusion or bipedal asymmetric mutation, it is determined to be in the global instability stage, and the evolution gain coefficient is further amplified by a preset second multiple of 2.2 times based on the coefficient of the local defect stage.

[0059] In this embodiment, the closed-loop stability parameter in the steady-state force-flow baseline model is used as the individual steady-state benchmark quantity to normalize the real-time instability characteristics, transforming the absolute instability value into a relative degree of instability relative to its own steady-state level. Specifically, during the calculation, the topological edge fracture degree is first multiplied by "1 + conduction time-series distortion rate" to obtain the comprehensive absolute quantity of instability resulting from the coupling between structure and time. The time-series distortion rate is coupled by adding 1 before multiplication, which can realize the superposition and amplification effect of time-series distortion on the basis of structural instability.

[0060] For example, when the temporal distortion is zero, the overall instability is equal to the structural instability; when the temporal distortion increases, the overall instability increases proportionally, accurately reflecting the synergistic risk-causing effect of structural fracture and temporal disorder. Dividing the absolute amount of overall instability by the closed-loop stability parameter yields the normalized structural instability component. Patients with poorer steady-state levels exhibit higher instability components corresponding to the same absolute instability.

[0061] In this embodiment, a baseline value for the standard conduction velocity of the force flow is first determined. This value is calculated by dividing the total spatial distance of the main conduction link from the heel to the toe in the steady-state baseline model by the total standard conduction delay of the main link, and is a patient-specific steady-state conduction velocity baseline. The ratio of the real-time asymmetric defect propagation velocity to the standard conduction velocity is calculated, and then the exponent is taken with the natural constant as the base to obtain the propagation velocity gain term.

[0062] Preferably, when the propagation speed of the defect is lower than the standard speed, the gain term is close to 1, and the risk is not significantly amplified; when the propagation speed exceeds the standard speed, the gain term grows exponentially, and the risk of instability rises rapidly.

[0063] Furthermore, by multiplying the evolutionary gain coefficient, the structural and temporal coupling instability component, and the propagation velocity exponential gain term, the estimated fall risk value for the current single frame is obtained. The calculation formula is as follows: ; in, For the projected risk value of falling, This represents the evolution gain coefficient corresponding to the current topology-breaking evolution stage. For topological edge breakage, This represents the conduction timing distortion rate. These are the closed-loop stability parameters in the steady-state force-flow baseline model. For the propagation speed of asymmetric breaking, This represents the preset standard force flow transmission velocity in the steady-state baseline model. The calculated fall risk values ​​for this embodiment are shown in Table 6.

[0064] Table 6: Data Table of Fall Risk Projection Values ​​Frame-by-Frame Calculation Process

[0065] It should be noted that by constructing a three-level risk extrapolation architecture of "stage judgment + individual normalization + nonlinear coupling", it breaks through the technical limitations of existing fall warning schemes that rely on general fixed thresholds and linear weighted summation. Its dynamic threshold system based on individual baselines realizes personalized adaptation of risk benchmarks, enabling the staged evolutionary gain mechanism to accurately match the law of instability from perturbation to outbreak. The exponential coupling mechanism of the failure propagation speed fully restores the nonlinear characteristics of rapid spread of instability. The final output fall risk extrapolation value can not only accurately quantify the relative instability of different patients, but also keenly capture the early risk evolution trend of slow instability scenarios such as soft falls, significantly improving the accuracy and foresight of risk prediction.

[0066] In this embodiment, the Level 1 warning in step 500 is an early reminder for minor instability. Triggering this warning requires meeting two rigid conditions simultaneously: the real-time state matching is in the topological perturbation stage, and the smoothed fall risk projection value falls within the Level 1 threshold range. All threshold ranges are determined based on the individual's steady-state force flow baseline. The lower limit of the Level 1 threshold range is set to the 95th percentile of the steady-state baseline risk value, and the upper limit is set to 1.0, covering only minor risks outside the patient's normal fluctuation range.

[0067] Specifically, after being triggered, the warning signal is only sent to the foot sensor terminal or wristband terminal worn by the patient, and the output prompt is in short pulse vibration mode, specifically 3 sets of pulses with a 200ms vibration and a 300ms interval. Each prompt lasts for 2 seconds and is repeated every 3 seconds until the risk value falls below the lower limit of the first level interval.

[0068] In this embodiment, the secondary alarm warning is for medium-risk scenarios with local instability. The triggering requires matching the local failure stage and the secondary threshold range at the same time. The secondary threshold range is set to [1.0, 5.0), which corresponds to the state where the risk has exceeded the range of slight fluctuations and there is a clear tendency to become unstable.

[0069] Specifically, upon triggering, a dual-terminal linkage alarm is activated: the patient's terminal initiates a moderate audio-visual reminder, guiding the patient to immediately stop their actions and adjust their posture; at the same time, an early warning notification is pushed to the nursing station terminal and the responsible nurse's handheld terminal. The notification includes the patient's bed number, current risk level, duration of risk, and initial instability type, prompting nursing staff to conduct nearby rounds to confirm the status.

[0070] In this embodiment, the Level 3 emergency warning targets high-risk scenarios of overall instability and is the highest level of warning. Triggering it requires matching both the overall instability stage and the Level 3 threshold range. The Level 3 threshold range is set to ≥5.0, which corresponds to an emergency state where instability has spread rapidly and the probability of falling is extremely high.

[0071] Specifically, after triggering, multi-system emergency linkage is activated: the patient terminal activates the highest-level audible and visual alarm to achieve a strong reminding effect; meanwhile, the patient's identity, ward location, risk level, and start time of instability are synchronously pushed to the nursing management system, an emergency disposal work order with the highest priority is automatically generated, dispatched to the nearest nursing staff, and simultaneously copied to the ward head nurse. The evolution process of the fall risk deduction value in this embodiment is as Figure 8 shown.

[0072] As shown in Figure 9 , an embodiment of the present invention further provides a patient fall early warning system based on pressure detection, for implementing the aforementioned patient fall early warning method based on pressure detection, comprising: a topology structure construction module, configured to perform conduction filtering calibration and biomechanical partition mapping on pressure data collected by a plantar array pressure sensor, construct a force flow topology network, and generate a time series according to sampling time sequence; a parameter base generation module, configured to determine a standard topology connection structure, standard conduction delay between nodes, and closed-loop stability parameters according to the time series corresponding to the patient's daily steady-state movements, so as to obtain a steady-state force flow baseline model; a characteristic parameter calculation module, configured to perform frame-by-frame tracking and comparison on real-time time series, and calculate and obtain topological instability characteristic parameters; the topological instability characteristic parameters comprise: topological edge breakage degree, conduction time sequence distortion rate, and asymmetric breakage propagation velocity; a fall risk prediction module, configured to substitute the topological instability characteristic parameters into the steady-state force flow baseline model for matching, perform topological breakage coupling deduction, and calculate and obtain a fall risk deduction value; a risk early warning module, configured to perform risk early warning according to the fall risk deduction value, and obtain graded early warning signals.

[0073] Beneficial effects of the present invention are as follows: 1) By constructing a plantar force flow topology network, instability signals are identified from the evolution dimension of the force flow conduction structure, which can complete risk prediction in the early stage of instability when the human body's center of gravity has not yet shifted significantly, effectively advance the early warning time, and reserve sufficient reaction time for the patient to adjust posture independently and for nursing staff to intervene and dispose; 2) By adopting a nonlinear coupling deduction mechanism of three types of characteristics, i.e., topological edge breakage degree, conduction time sequence distortion rate, and asymmetric breakage propagation velocity, it can accurately distinguish daily movements such as sitting down, squatting, and bending over from real instability states; meanwhile, it has good recognition capability for "soft fall" scenarios such as slow slipping and syncopal collapse without violent pressure impact, has high scene recognition accuracy, and greatly reduces false alarms and missing alarms; 3) By constructing a patient-specific steady-state force flow baseline model, it can adaptively match the differences in weight, gait, and limb function among different patients; at the same time, through the dynamic screening mechanism of defect confidence, it avoids the judgment bias caused by a general fixed threshold and can adapt to different physiological states. 4) Through a three-level early warning system with dual verification of "stage judgment + risk value range", different intervention methods are matched for different risk levels. The system upgrades step by step from vibration prompts at the patient end to emergency linkage of the nursing system. This not only avoids unnecessary alarms from interfering with daily nursing order, but also enables rapid mobilization of nursing resources in high-risk scenarios, reducing secondary injuries caused by falls.

[0074] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0075] Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. Furthermore, those skilled in the art will recognize that, based on the ideas of this invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A patient fall early warning method based on pressure detection, characterized in that, Includes the following steps: The pressure data collected by the plantar array pressure sensor is subjected to conduction filtering calibration and biomechanical partition mapping to construct the output force flow topology network and generate a time sequence according to the sampling time sequence; Based on the time sequence corresponding to the patient's daily steady-state actions, the standard topological connection structure, standard conduction delay between nodes, and closed-loop stability parameters are determined to obtain the steady-state force flow baseline model. The real-time time sequence is tracked and compared frame by frame, and the topological instability characteristic parameters are calculated. The topological instability characteristic parameters include: topological edge fracturing degree, propagation time-series distortion rate, and asymmetric failure propagation speed; The topological instability characteristic parameters are substituted into the steady-state force-flow baseline model for matching, and topological breaking coupling deduction is performed to calculate the fall risk deduction value. Risk warnings are generated based on the predicted fall risk values, resulting in tiered warning signals.

2. The patient fall early warning method based on pressure detection according to claim 1, characterized in that, The pressure data collected by the plantar array pressure sensor is subjected to conduction filtering calibration and biomechanical partitioning mapping to construct the output force flow topology network, and a time series is generated according to the sampling time sequence, including: The pressure data is spatially registered with the foot biomechanical coordinate system based on the anatomical landmarks on the sole of the foot to obtain pressure distribution data; the anatomical landmarks on the sole of the foot include: the extreme pressure point of the heel, the extreme pressure point of the first metatarsal bone, and the extreme pressure point of the fifth metatarsal bone; The pressure distribution data is divided into multiple load-bearing zones through the biomechanical transmission pathway of the sole of the foot; The defect confidence level is calculated based on the deviation components of the bearing zone, and the bearing zones with defect confidence levels lower than the preset confidence level are removed; the deviation components include: spatial amplitude deviation, temporal transmission deviation and zone coupling deviation; the bearing zones include: heel rear bearing zone, heel front transition zone, arch support zone, forefoot main bearing zone and toe auxiliary bearing zone; Using the bearing partition as the topology node and the pressure transmission gradient direction and transmission weight between the bearing partitions as directed edges, a single frame of the force flow topology network is constructed, and the force flow topology networks of all sampling times are spliced ​​together in chronological order to obtain the time series sequence.

3. The patient fall early warning method based on pressure detection according to claim 1, characterized in that, Based on the time sequence corresponding to the patient's daily steady-state actions, the standard topological connection structure, standard inter-node conduction delay, and closed-loop stability parameters are determined to obtain the steady-state force flow baseline model, including: The standard connection weight of each directed edge is calculated based on the frequency of occurrence of each directed transmission edge in the time sequence corresponding to the patient's daily steady-state actions. The conduction delay data of the pressure peak between each adjacent node in the time series is extracted sample by sample, and kernel density estimation is performed to obtain the main conduction delay; The conduction direction consistency coefficient is obtained based on the proportion of samples conforming to the standard conduction direction on the conduction path of the main conduction delay. The main conduction delay is corrected based on the conduction direction consistency coefficient to obtain the standard conduction delay between nodes; The average coefficient of variation of all directed edge weights is calculated based on the standard propagation delay between nodes to obtain the topology stability components. The average coefficient of variation of all propagation path delays is calculated based on the standard propagation delay between nodes to obtain the temporal propagation stability component. The bipedal weight symmetry deviation rate is calculated based on the standard connection weights of the corresponding conduction paths of the left and right feet, and the bipedal symmetry stability components are obtained. The closed-loop stability parameter is obtained by multiplying the topological stability component, the time-series propagation stability component, and the bipedal symmetric stability component. The standard topology connection structure, the standard inter-node propagation delay, and the closed-loop stability parameters are encapsulated and integrated to obtain the steady-state force flow baseline model.

4. The patient fall early warning method based on pressure detection according to claim 1, characterized in that, The real-time time-series sequence is tracked and compared frame by frame, and topological instability characteristic parameters are calculated, including: In the real-time time series, directed edges whose weight decay rate exceeds a preset breakage threshold are marked as broken edges, and the standard weights and weight decay rates of all broken edges are obtained. The proportion of broken edges on the main transmission link from heel to toe is obtained as the ratio of the number of broken edges on the main link to the total number of edges on the main link. The weighted sum of the weighted attenuation rates is exponentially coupled with the proportion of main link breaks to obtain the topology edge breakage degree of the current frame; the calculation formula for the topology edge breakage degree is: ,in For topological edge breakage, This is the set of broken edges in the current frame. Let be the standard connection weight corresponding to the i-th fracture edge. Let be the weight decay rate of the i-th fracture edge. The total connection weights of the steady-state baseline topology network. The number of broken edges on the main transmission link. This is the total number of edges in the main transmission link.

5. The patient fall early warning method based on pressure detection according to claim 4, characterized in that, The process includes frame-by-frame tracking and comparison of the real-time time-series sequence, and calculation of topological instability characteristic parameters, and also includes: The actual conduction delay of each conduction path in the time sequence is extracted frame by frame, and the difference is normalized by comparing it with the standard conduction delay of the corresponding path to obtain the single path delay deviation rate. The proportion of reversed direction paths to the total number of conduction paths is calculated based on the number of paths where the conduction direction is reversed, thus obtaining the proportion of reversed direction paths. The single-path delay deviation rate is coupled with the proportion of the direction-reversed path to obtain the propagation timing distortion rate of the current frame; the formula for calculating the propagation timing distortion rate is: ,in This represents the conduction timing distortion rate. This represents the total number of propagation paths in the topology network. Let j be the actual propagation delay of the j-th propagation path. Let j be the standard propagation delay of the j-th propagation path. This represents the percentage of paths with reversed direction.

6. The patient fall early warning method based on pressure detection according to claim 5, characterized in that, The process includes frame-by-frame tracking and comparison of the real-time time-series sequence, and calculation of topological instability characteristic parameters, and also includes: Based on the starting node and diffusion path of the topology break in the time series, determine the time duration for the break to spread from the starting node to all bearing partitions of the entire system, and calculate the average propagation speed of the break by combining the spatial transmission distance of the topology nodes. The degree of breakage of the topological edge of the left and right feet is calculated separately to obtain the difference in the degree of breakage between the two feet; The asymmetric malfunction propagation velocity is obtained by exponentially coupling the average propagation velocity with the bipedal malfunction difference; the formula for calculating the asymmetric malfunction propagation velocity is as follows: ,in For the propagation speed of asymmetric breaking, The average propagation velocity of the break within the initial foot, The topological edge breakage degree of the left foot. The topological edge breakage degree of the right foot.

7. The patient fall early warning method based on pressure detection according to claim 3, characterized in that, The topological instability characteristic parameters are substituted into the steady-state force-flow baseline model for matching, and topological breaking coupling deduction is performed to calculate the fall risk deduction value, including: The topological instability characteristic parameters are matched class by class with the preset hierarchical threshold intervals in the steady-state force flow baseline model to obtain the current topological breaking evolution stage and the evolution gain coefficient of the corresponding stage. Divide the topological edge fracture degree in the steady-state force flow baseline model by the topological stability component to obtain the structural instability component; Divide the conduction time-series distortion rate in the steady-state force flow baseline model by the time-series conduction stability component to obtain the time-series instability component; The evolutionary gain coefficient, the structural instability component, and the temporal instability component are coupled and calculated to obtain the fall risk projection value; the calculation formula for the fall risk projection value is as follows: ,in For the projected risk value of falling, This represents the evolution gain coefficient corresponding to the current topology-breaking evolution stage. For topological edge breakage, This represents the conduction timing distortion rate. These are the closed-loop stability parameters in the steady-state force-flow baseline model. For the propagation speed of asymmetric breaking, This is the preset standard transmission velocity of force flow in the steady-state baseline model.

8. The patient fall early warning method based on pressure detection according to claim 7, characterized in that, The topological instability characteristic parameters are matched class-by-class with the preset hierarchical threshold intervals in the steady-state force-fluid baseline model to obtain the current topological breaking evolution stage and the corresponding evolution gain coefficient, including: When the topological edge breakage and the propagation time-series distortion rate are both within the baseline normal fluctuation threshold, and the asymmetric spur propagation speed is zero, it is determined to be the topological perturbation stage, and the corresponding evolution gain coefficient is taken as the baseline value. When any two of the topological instability characteristic parameters exceed the baseline normal fluctuation threshold and the defect range is limited to a single-leg single-bearing zone, it is determined to be a local defect stage, and the corresponding evolution gain coefficient is amplified by a preset first factor based on the benchmark value. When all three topological instability characteristic parameters exceed the baseline critical threshold, and the breach exhibits cross-regional diffusion or bipedal asymmetric mutation, it is determined to be a global instability stage, and the corresponding evolutionary gain coefficient is amplified by a preset second factor based on the local breach stage.

9. The patient fall early warning method based on pressure detection according to claim 8, characterized in that, Based on the predicted fall risk values, risk warnings are generated, resulting in tiered warning signals, including: When the match is in the topological perturbation stage and the fall risk projection value is in the first-level threshold range, a first-level warning is triggered and a vibration alert is sent to the patient's wearing terminal. When the match is the local defect stage and the fall risk projection value is in the secondary threshold range, a secondary alarm is triggered, an audio-visual reminder is sent to the patient, and an alarm notification is pushed to the nursing terminal. When the match is in the global instability stage and the fall risk projection value is in the third-level threshold range, a third-level emergency warning is triggered, the highest-level audible and visual alarm is activated, and the patient's location is sent to the nursing system.

10. A patient fall warning system based on pressure detection, used to implement the patient fall warning method based on pressure detection according to any one of claims 1-9, characterized in that, include: The topology construction module is used to perform conduction filtering calibration and biomechanical partitioning mapping on the pressure data collected by the plantar array pressure sensor, construct the output force flow topology network, and generate a time sequence according to the sampling time sequence; The parameter base generation module is used to determine the standard topological connection structure, standard conduction delay between nodes and closed-loop stability parameters based on the time sequence corresponding to the patient's daily steady-state actions, so as to obtain the steady-state force flow baseline model. The feature parameter calculation module is used to track and compare the real-time time series frame by frame, and calculate the topological instability feature parameters. The topological instability characteristic parameters include: topological edge fracturing degree, propagation time-series distortion rate, and asymmetric failure propagation speed; The fall risk prediction module is used to substitute the topological instability characteristic parameters into the steady-state force flow baseline model for matching, and to perform topological breaking coupling inference to calculate the fall risk inference value. The risk warning module is used to issue risk warnings based on the fall risk projection value and obtain graded warning signals.

Images