A fall risk identification method, system and apparatus

Abstract

This invention proposes a fall risk identification method, system, and device. The fall risk identification method includes: acquiring gait data, X-ray images, and bone density images for continuity and temporal analysis to create a gait description data structure; calculating the offset of the offset line based on the gait description data structure and filtering deformed nodes using X-ray images; obtaining abnormal morphological combination information; generating a gait discontinuity sequence based on the abnormal morphological information and temporal features; filtering propulsion changes based on the gait discontinuity sequence, reference trajectory time, and acceleration data, and constructing a movement performance structure based on movement classification, position, and center of gravity fluctuations; determining movement trigger combinations based on the movement performance structure, and analyzing the displacement direction and pressure point location of the movement trigger combinations to generate a fall risk identification result. By analyzing physiological parameters such as gait, body offset, and support contact, the method accurately reflects an individual's postural changes and skeletal health during movement, improving the accuracy of the assessment.

Description

Technical Field

[0001] This invention relates to the field of health risk assessment technology, and in particular to a method, system and device for identifying fall risk. Background Technology

[0002] The field of health risk assessment technology involves methods and systems for disease prediction, behavioral risk assessment, and health management decision support based on individual health data. Core aspects include physiological parameter collection, quantitative analysis of health status, individual risk level classification, and the construction of risk warning mechanisms for specific health problems. This technology widely integrates wearable device monitoring, medical image processing, biomechanical modeling, and artificial intelligence algorithms in clinical auxiliary assessment. Its overall development trend is towards multimodal data fusion, personalized risk assessment, and dynamic health intervention. Traditional fall risk identification methods refer to the static grading of patients using general fall risk assessment scales. These methods typically set assessment factors based on nursing questionnaire results, such as the patient's past fall history, mobility, cognitive status, and medication use. A total risk value is calculated through scoring to classify fall risk levels. Some methods combine imaging or gait data for single-dimensional analysis.

[0003] Current technologies rely on static assessment scales and subjective questionnaires to classify patients' fall risk, judging solely by factors such as past fall history, mobility, cognitive status, and medication use. This approach is limited to single-dimensional data analysis and cannot reflect the dynamic changes in an individual's health status in real time. Furthermore, traditional risk assessment methods lack monitoring of changes in body posture and structure during movement and fail to adequately consider the integration of multimodal data, resulting in limited accuracy and difficulty in timely identification of potential risks. This assessment method often depends on the subjective judgment of caregivers, which may lead to misjudgments and omissions, and cannot provide personalized health management and intervention plans. Summary of the Invention

[0004] To address the technical problems existing in the prior art, the present invention provides a fall risk identification method, system, and device.

[0005] On the one hand, a method for identifying fall risk is provided, which includes:

[0006] S10: Acquire gait data, X-ray and bone density images for continuous and temporal analysis, gait description data structure; S20: Calculate the offset of the offset line based on the gait description data structure, and filter deformed nodes by combining the X-ray image; to obtain abnormal morphology combination information; S30: Generate a discontinuous gait sequence based on the abnormal morphological information and temporal features; S40: Based on the gait discontinuity sequence, reference trajectory time and acceleration data, filter and propel the changing movements, and construct the movement performance structure by movement classification, position and center of gravity fluctuation markers; S50: Determine the action trigger combination based on the action performance structure, analyze the displacement direction and pressure point location of the action trigger combination, and generate a fall risk identification result.

[0007] Furthermore, the gait description data structure includes positional continuity, support contact timing, X-ray image index, and bone density image index; The abnormal morphology combination information includes the number of times the vertebral body offset line turns, the span of the vertebral body offset line, the contraction range of the vertebral body edge line segment, the number of protruding curve points in the longitudinal direction, the spatial deformation of the support node, and the number of vertebral body structural changes. The gait discontinuity sequence includes the location of changes in the distribution density of the protrusion point, the difference in the path extension direction between the gait interruption point and the adjacent segment, and the arrangement and combination of time delay and spatial offset. The motion performance structure includes motion category name, start and end positions, center of gravity fluctuation mark, motion trajectory initial acceleration segment, path reversal segment, and longitudinal axis bending segment. The risk identification results include action trigger combinations, spatial displacement direction, time location of structural pressure points, and fall risk index.

[0008] Further, step S10 includes: S101: Acquire acceleration data, perform a second integration on the acceleration data according to the time sequence, calculate the trajectory variation frequency by combining the direction change rate, and generate a continuous displacement path sequence and a variation frequency parameter set; S102: Based on the continuous displacement path sequence and the set of changing frequency parameters, combined with pressure data, extract the contact area sequence according to the single-step cycle, calculate the offset angle of the trajectory corresponding to the contact point, and generate a support contact timing index and gait offset angle sequence. S103: Based on the support contact timing index and gait offset angle sequence, match the X-ray and bone density images under time synchronization, filter the imaging data corresponding to the time, and establish a gait description data structure.

[0009] Further, step S20 includes: S201: Obtain the position continuity and support contact timing in the gait description data structure, extract the trajectory segment within the support cycle, calculate the number of trajectory turning changes and the corresponding displacement span, and generate a sequence of the number of turning changes and span within the support cycle; S202; Based on the number of turning cycles and span sequence of the support cycle, and combined with the vertebral edge segments in the X-ray image, the compression amplitude and the number of longitudinal curvature points of each vertebral segment are counted to generate a vertebral structure compression amplitude and curvature point comparison matrix. S203: Based on the comparison matrix of the compression amplitude and curvature points of the vertebral structure, support nodes that exceed the preset benchmark threshold are selected, and correlation analysis is performed in combination with the corresponding number of turns and span to establish an abnormal morphology combination information set.

[0010] Further, step S30 includes: S301: Obtain the offset path in the abnormal morphology combination information, extract the spatial position sequence corresponding to the continuous segment of gait rhythm, detect the change in the temporal distribution density of the sudden point in the trajectory, count the number of sudden point locations and label the index according to the time axis sliding window, and generate a temporal index sequence of rhythm density change. S302: Based on the rhythm density change time index sequence, obtain the time sequence of start and stop between consecutive steps, extract the difference in path direction and start-up / fall sequence according to adjacent segments, calculate the combined difference in time interval and direction transfer of each group of segments, and generate a rhythm segment time and path difference matrix. S303: Based on the rhythm segment time and path difference matrix, filter segments whose interval time deviation exceeds a preset time delay threshold and whose path direction deviation exceeds a preset spatial offset threshold, arrange the abnormal combination segment identifiers according to the sequence recombination method, and establish a gait discontinuity sequence.

[0011] Further, step S40 includes: S401: Based on the differences in the path extension direction and take-off and landing sequence in the discontinuous gait sequence, extract the position indices of the continuous acceleration segment, reverse segment and longitudinal axis bending segment in the trajectory, mark the start and end nodes of the trajectory segment and divide the path into segments, and generate the motion trajectory pattern segment index group. S402: Based on the action trajectory pattern segment index group, obtain the acceleration data sequence corresponding to the segment, extract the acceleration change value according to the time range and calculate the variance, determine whether the range of change concentration exceeds the preset fluctuation benchmark value, and filter the types that constitute the propulsion structure difference in the corresponding path to generate a propulsion structure action mark set. S403: Based on the set of propulsion structure action markers, extract the classification name, start and end time index and center of gravity coordinate sequence change of the action segment, mark the fluctuation state according to the center of gravity axial offset, combine and output the name, position and center of gravity identifier corresponding to the action, and establish the action performance structure.

[0012] Further, step S50 includes: S501: Based on the action category name and center of gravity fluctuation mark in the action performance structure, extract the combination sequence between adjacent actions on the time axis, determine whether there are segments in the continuous action sequence that include center of gravity changes, and mark the action combination that is expected to trigger body imbalance in time order to generate an abnormal action trigger combination index sequence. S502: Invoke the abnormal action trigger combined index sequence, extract the path extension direction vector in the combination and the inertial data change range within the corresponding time window, combine the preset vertebral compression area coordinate mapping rules, locate the time point where the direction shift and impact fluctuation overlap, and obtain the structural compression sensitive spatiotemporal positioning matrix. S503: Based on the structural pressure-sensitive spatiotemporal positioning matrix, filter the positioning indices that simultaneously meet the preset pressure intensity threshold and trigger action combination, classify the risk types formed by the trigger action combination and pressure position change according to the time series, and establish a fall risk identification result set.

[0013] Furthermore, the coordinate mapping rule for the vertebral compression area is based on the vertebral anatomical structure, mapping the pressure-sensitive area to a spatial positioning reference model in a unified coordinate system, and identifying the compression point and the mapping rule obtained from the alignment analysis.

[0014] On the other hand, the present invention also provides a fall risk identification system, which includes: Gait acquisition module: used to acquire gait data, X-ray and bone density images for continuous and temporal analysis, gait description data structure; The morphology recognition module is used to calculate the offset of the offset line based on the gait description data structure, and to filter deformed nodes by combining the X-ray image; thus obtaining abnormal morphology combination information. Rhythm analysis module: used to generate gait discontinuity sequences based on the abnormal morphological information and temporal features; Action identification module: used to filter and promote changes in actions based on the gait discontinuity sequence, reference trajectory time and acceleration data, and construct the action performance structure by action classification, position and center of gravity fluctuation markers; Fall detection module: used to determine the action trigger combination based on the action performance structure, and analyze the displacement direction and pressure point location of the action trigger combination to generate fall risk identification results.

[0015] On the other hand, the present invention also includes a fall risk identification device, which comprises: An inertial sensor, a pressure sensor, an imaging device, and any of the above-mentioned fall risk identification systems, wherein the inertial sensor, pressure sensor, and imaging device are respectively connected to the gait acquisition module; The inertial sensor is used to detect the acceleration of the corresponding part of the object to be detected; the pressure sensor is used to detect the pressure data of the corresponding part of the object to be detected; the imaging device is used to perform continuous X-ray image imaging and bone density image imaging of the image to be detected.

[0016] The beneficial effects of the technical solution provided by this invention include at least the following: By integrating dynamic data collected from inertial sensors, pressure sensors, and imaging equipment, this system can comprehensively and in real-time track and analyze a patient's gait, body deviation, support contact, and other physiological parameters, accurately reflecting an individual's postural changes and skeletal health during movement. Through multimodal data fusion, it can identify potential fall risks and health abnormalities, overcoming the limitations of traditional single-dimensional analysis and improving the accuracy of risk assessment. This solution provides dynamic and personalized health management plans, enabling precise intervention based on real-time monitoring, avoiding misjudgments and delays, and enhancing the timeliness and scientific rigor of assessment and intervention.

[0017] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0018] Figure 1 This is a structural block diagram of the fall risk identification device of the present invention; Figure 2 This is a structural block diagram of the fall risk identification system of the present invention. Figure 3 To execute Figure 2 The flowchart shown is for the risk identification method of the fall risk identification system. Figure 4 This is a flowchart illustrating the specific execution process of the gait acquisition module. Figure 5 This is a flowchart illustrating the specific execution process of the shape recognition module; Figure 6 This is a flowchart illustrating the specific execution process of the rhythm analysis module; Figure 7 Here is a flowchart illustrating the specific execution process of the action identification module; Figure 8 This is a flowchart illustrating the specific execution process of the fall detection module. Detailed Implementation

[0019] To address the limitations of traditional risk assessment methods, which lack monitoring of body posture and structural changes during movement and fail to adequately consider the fusion of multimodal data, thus restricting the accuracy of assessment results, this invention combines image processing and gait data analysis, and based on this, proposes... Figure 1 The fall risk identification device shown includes an inertial sensor, a pressure sensor, an imaging device, and a fall risk identification system. The inertial sensor, pressure sensor, and imaging device are all connected to the gait acquisition module. The inertial sensor is installed on the lower back of the subject to acquire acceleration data of the lower back. The pressure sensor is installed on the sole of the subject's foot to detect pressure contact areas during the walking cycle and to collect pressure data during testing. The imaging device acquires X-ray images and bone density images at the same time. Please refer to [link to relevant documentation]. Figure 2The fall risk identification system includes: a gait acquisition module 10, a morphology recognition module 20, a rhythm analysis module 30, a movement discrimination module 40, and a fall identification module 50. The gait acquisition module 10 acquires continuous gait data of the subject during the testing process and establishes a corresponding gait description data structure. Subsequently, the morphology recognition module 20 identifies the cone offset of the subject at different time points based on the gait description data structure and constructs abnormal morphological combination information. The rhythm analysis module 40 performs time delay and spatial offset permutations based on the abnormal morphological combination information to generate a gait discontinuity sequence. Finally, the fall identification module 50 determines whether the gait discontinuity sequence is combined with body movement triggers, generating a fall risk identification result for the subject. For details, please refer to [link to relevant documentation]. Figure 3 The working process of a fall risk identification system includes: The gait acquisition module 10 is used to perform step S10: acquiring gait data, X-ray and bone density images for continuity and temporal analysis, and gait description data structure. The gait description data structure includes positional continuity, support contact timing, X-ray image index, and bone density image index.

[0020] The system acquires the continuous positional change trajectory along the front-back and up-down directions generated by a human inertial sensor installed on the lower back. Combined with the pressure contact area of ​​a foot pressure sensor within the walking cycle, the distribution pattern of body offset trajectory and support contact sequence in each step is analyzed. Simultaneously, X-ray images and bone density images acquired at the same time as the imaging device are extracted. By analyzing positional continuity, support contact timing, and imaging time index, a gait description data structure is constructed. Specifically, the gait acquisition module 10 includes a data trajectory acquisition submodule, a support sequence recognition submodule, and an imaging data synchronization submodule. Please refer to [link to relevant documentation]. Figure 4 At this point, step S10 includes S101-S103.

[0021] The data trajectory acquisition submodule is used to execute step S101: read the front-back and up-down acceleration data output by the waist and back inertial sensor, integrate them in time order, first integrate the acceleration to obtain the velocity, then perform a second integration to estimate the displacement, combine the direction change rate to calculate the trajectory change frequency, and generate a continuous displacement path sequence and change frequency parameter set. The inertial measurement unit (IMU), located in the subject's lower back (specifically at the junction of the lumbar vertebrae and sacrum), is connected via hardwired or wireless transmission protocol. This IMU incorporates a triaxial accelerometer and a triaxial gyroscope, with a sampling frequency set to 100 Hz. The submodule first reads the raw data output from the sensors, including acceleration values ​​along the sagittal axis (anteroposterior direction) and along the vertical axis (vertical direction), and performs denoising on the raw signal. The denoising process uses a moving average filtering algorithm with a window size of 5 sampling points. Subsequently, the submodule performs a double integration operation on the denoised acceleration data to obtain displacement data: first, the acceleration data is multiplied by the sampling time interval (0.01 seconds) and accumulated to obtain a velocity sequence; then, the velocity sequence is multiplied again by the sampling time interval and accumulated to obtain a displacement sequence relative to the initial position. During integration, the submodule performs zero-velocity correction based on gravity component compensation and a low-velocity threshold. Specifically, if the magnitude of the triaxial acceleration after removing the gravity component is below a set threshold (e.g., 0.1 m / s²) for five consecutive cycles, and the angular velocity change is less than a threshold (e.g., 5° / s), it is considered stationary, and the current velocity is forcibly zeroed to reduce integration drift error. The submodule further calculates the trajectory variation frequency by monitoring the number of times the direction vector changes along the displacement path and dividing it by the total sampling time. For example, in a 10-second sampling period, if the cumulative forward and backward displacement is 5 meters and the cumulative vertical fluctuation is 0.2 meters, and the direction vector undergoes 20 sign reversals, the trajectory variation frequency is 2 Hz. Finally, the submodule arranges the displacement coordinates, including timestamps, sequentially to generate a continuous displacement path sequence and stores the calculated frequency values ​​in a variation frequency parameter set.

[0022] The support sequence recognition submodule is used to execute step S102: call the continuous displacement path sequence and the variable frequency parameter set, combine the pressure data of the plantar pressure sensor, extract the contact area sequence according to the single-step cycle, calculate the offset angle of the trajectory corresponding to the contact point, and generate the support contact time sequence index and gait offset angle sequence. The system calls upon a continuous displacement path sequence and a set of frequency variation parameters, and simultaneously reads data from a thin-film pressure sensor array embedded in the subject's insole via an analog-to-digital converter interface. This array contains pressure sensing points in three key areas: the forefoot, arch, and heel, with output units in kilopascals (kPa). The submodule sets a pressure trigger threshold of 15 kPa. When the pressure value at any sensing point exceeds this threshold, a contact state is initiated; when the pressure values ​​at all sensing points are below this threshold, a swing state is initiated. The submodule extracts the contact area sequence within a single-step cycle based on the timeline (e.g., "heel-arch-forefoot"), and simultaneously captures the displacement trajectory of the lower back within that time period. Subsequently, the submodule calculates the offset angle of the trajectory corresponding to the contact point. This calculation first establishes an ideal straight-line travel vector, then obtains the endpoint vector of the actual displacement trajectory within the current single-step cycle, and calculates the angle between the two vectors. For example, if the ideal vector points due north (0 degrees), and the actual trajectory endpoint vector is biased towards the northeast (5 degrees azimuth), then the offset angle is 5 degrees. If, within a single-step cycle of 1.2 seconds, the pressure sensor sequentially records a peak pressure of 200 kPa for the heel and a peak pressure of 350 kPa for the forefoot, and the corresponding waist trajectory shifts 3 degrees to the left, the submodule will generate a support contact timing index containing start and end times and pressure distribution characteristics, as well as a gait offset angle sequence recording the 3-degree value.

[0023] The imaging data synchronization submodule is used to perform step S103: based on the support contact timing index and gait offset angle sequence, match the X-ray and bone density images under time synchronization, filter the imaging data corresponding to the time, and establish a gait description data structure.

[0024] The submodule invokes the support contact timing index and gait offset angle sequence, and connects the X-ray imaging equipment to the dual-energy X-ray bone densitometer server via the DICOM (Digital Imaging and Communication in Medicine) protocol interface. The submodule reads the metadata header information of the image file, extracting the acquisition time stamp accurate to milliseconds. It then performs time alignment, comparing the time period in the support contact timing index with the image time stamp, with a tolerance range set to 500 milliseconds. When a specific moment in the index (e.g., 10:05:01.200 milliseconds) falls within the tolerance range of the X-ray image's exposure time, the submodule binds the image file's storage path to the gait data at that moment. Furthermore, the submodule extracts the pixel grayscale value matrix of the region of interest (ROI) in the bone densitometer image. If a match is successful, the submodule constructs a multidimensional array containing gait motion parameters (displacement, frequency, angle), plantar pressure distribution, and corresponding bone image data—the gait description data structure. For example, for the gait segment with index ID 1001, the X-ray image data with the file name "XR_1226_001.dcm" and the bone mineral density data with the file name "BMD_1226_001.dcm" were associated to ensure the spatiotemporal consistency of kinematic and morphological data.

[0025] The morphology recognition module 20 is used to execute step S20: calculate the offset degree of the offset line according to the gait description data structure, and filter deformed nodes in combination with X-ray images; to obtain abnormal morphology combination information. The abnormal morphology combination information includes the number of turns of the vertebral offset line, the span of the vertebral offset line, the contraction range of the vertebral edge line segment, the number of protruding curve points in the longitudinal direction, the spatial deformation of the support node, and the number of vertebral structural changes.

[0026] The morphology recognition module 20 includes a support offset calculation submodule, a structural feature extraction submodule, and an abnormal morphology screening submodule. Please refer to [link / reference]. Figure 5 At this point, step S20 includes steps S201-S203.

[0027] The support offset calculation submodule is used to perform step S201: obtain the position continuity and support contact time sequence in the gait description data structure, extract the trajectory segment within the support cycle, calculate the number of trajectory turning changes and the corresponding displacement span, and generate a sequence of the number of turning changes and span of the support cycle.

[0028] Based on the positional continuity data in the gait description data structure, the time window marked as the "support phase" in the support contact time sequence is located, and the set of waist trajectory coordinate points corresponding to this window is extracted. The submodule traverses this coordinate set, calculates the slope of the vector formed by adjacent coordinate points, and records a trajectory turn when the sign of the slope changes (from positive to negative or vice versa). The submodule counts the total number of turns within a single support cycle and calculates the maximum straight-line distance of the trajectory projection onto the horizontal plane within that cycle, i.e., the displacement span. For example, in a 0.8-second right-foot single support cycle, the trajectory coordinates show three alternating peaks and troughs in the vertical direction, indicating three turns; simultaneously, the starting point coordinates of this cycle are (0, 0), and the ending point coordinates are (0.6, 0.1), which, calculated using the Pythagorean theorem, results in a horizontal displacement span of 0.6 meters (ignoring minor lateral deviations). The submodule stores the calculated three turns and the 0.6-meter span value into a support cycle turn count and span sequence. If 10 gait cycles are continuously monitored, the sequence will contain 10 sets of corresponding turning counts and span values.

[0029] The structural feature extraction submodule is used to execute step S202; based on the number of turning cycles and span sequence of the support cycle, combined with the vertebral edge segments in the X-ray image, the compression amplitude and the number of longitudinal curvature points of each vertebral segment are counted, and a vertebral structure compression amplitude and curvature point number comparison matrix is ​​generated.

[0030] The submodule calls the support cycle turning number and span sequence, and loads synchronously matched X-ray image data. Internally, this submodule integrates a deep convolutional neural network for image analysis. The network first receives a vertebral X-ray grayscale image with a resolution normalized to 512×512 pixels through an input layer; then it passes through a first convolutional layer containing 32 3×3 convolutional kernels with a stride of 1, used to extract primary texture features of the vertebral body edges; next, it passes through a ReLU activation function layer to zero out negative values ​​in the convolutional output to increase non-linearity; after that, it enters a 2×2 max-pooling layer for downsampling. After four similar "convolution-activation-pooling" stacked structures, the network outputs the coordinates of key points on the edges of the upper and lower endplates of the vertebral body through fully connected layers. The submodule calculates the anterior and posterior heights of each vertebral segment based on these key point coordinates. The vertebral compression amplitude is calculated by dividing the absolute value of the difference between the anterior and posterior heights by the posterior height. Simultaneously, the submodule extracts pixels along the longitudinal axis edge of the vertebral body, calculates the second derivative of the fitted curve for these points, and counts the number of times the sign of the second derivative changes as the number of curvature points on the longitudinal axis. For example, if the anterior edge height of the L1 vertebral body is identified as 20 mm and the posterior edge height as 25 mm, the compression amplitude is calculated as (25-20)÷25 = 0.2 (20%); if there are two inflection points on the longitudinal axis edge curve, the number of curvature points is 2. The submodule fills these parameters of each vertebral body into the vertebral structure compression amplitude and curvature point count comparison matrix.

[0031] The abnormal morphology screening submodule is used to perform step S203: based on the comparison matrix of the compression amplitude and curvature points of the vertebral structure, it screens support nodes that exceed the preset benchmark threshold, and performs correlation analysis in combination with the corresponding number of turns and span to establish an abnormal morphology combination information set.

[0032] Based on the matrix comparing vertebral compression amplitude and curvature point count, the submodule scans each vertebral parameter row by row. The submodule presets a compression amplitude threshold of 0.15 (i.e., 15%) and a curvature point count threshold of 1. The submodule performs a logical judgment: if the compression amplitude of a vertebra is greater than 0.15, or the number of curvature points is greater than 1, then the supporting node corresponding to that vertebra is deemed to have a morphological abnormality risk. The submodule further correlates the number of turning cycles and span data corresponding to this abnormal node. For example, when the L1 vertebral compression amplitude is 0.2 (exceeding 0.15), the system searches for the corresponding gait cycle at that moment and finds 5 turning cycles (abnormally high frequency) and a span of 0.4 meters (abnormally short). The submodule packages the combination record "L1 vertebra - compression 0.2 - turning 5 - span 0.4" to establish an abnormal morphological combination information set.

[0033] Table 1 shows the comparison between the baseline values ​​of some parameters involved in the abnormal pattern screening and the actual calculation examples.

[0034]

[0035] Table 1. Comparison of Morphology Recognition Parameters Table 1 lists the key parameter threshold settings of the morphology recognition module when performing anomaly screening and the example data it actually captures. This data shows how the system triggers anomaly determination when the actual value exceeds the threshold.

[0036] The rhythm analysis module 30 is used to execute step S30: generating a gait discontinuity sequence based on the abnormal morphological information and temporal features. The gait discontinuity sequence includes the location of changes in the distribution density of the abrupt change points, the difference in the path extension direction between the gait interruption points and adjacent segments, and the arrangement and combination of time delays and spatial offsets. Specifically, the rhythm analysis module 30 includes a rhythm density detection submodule, an interval difference extraction submodule, and a gait coherence generation submodule. Please refer to [link to relevant documentation]. Figure 6 At this point, step S30 specifically includes S301-S303.

[0037] The rhythm density detection submodule is used to perform step S301: obtaining the offset path in the abnormal morphology combination information, extracting the spatial position sequence corresponding to the continuous gait rhythm segment, detecting the temporal distribution density change of the sudden point in the trajectory, counting the number of sudden point locations and labeling the index according to the time axis sliding window, and generating a rhythm density change time index sequence. The continuous gait rhythm segment is a continuous movement interval in which the gait rhythm remains relatively stable and without significant sudden changes within a certain time window.

[0038] Based on the offset path data in the abnormal pattern combination information set, the spatial location coordinate sequence within the time period marked as abnormal is extracted. This submodule uses a sliding window algorithm to detect abrupt points in the trajectory, with the sliding window duration set to 1 second and a step size of 0.1 seconds. Within each window, the submodule calculates the trajectory curvature radius; when the curvature radius is less than 0.05 meters, the location is determined to be an abrupt point. The submodule counts the total number of abrupt points within each window and calculates the temporal distribution density of the abrupt points, i.e., the number of abrupt points divided by the window duration. For example, if four abrupt turning points with curvature radii less than 0.05 meters are detected within the time window from the 2nd to the 3rd second, the rhythm density of this window is 4 per second. The submodule records the density value of each window sequentially along the time axis and indexes and labels high-frequency abrupt change regions with a density exceeding 3 per second, generating a temporal index sequence of rhythm density changes.

[0039] The interval difference extraction submodule is used to perform step S302: based on the rhythm density change time index sequence, obtain the time sequence of start and stop between consecutive steps, extract the differences in path direction and start and stop order according to adjacent segments, calculate the combined difference in time interval and direction transfer of each group of segments, and generate a rhythm segment time and path difference matrix.

[0040] Based on the temporal index sequence of rhythm density changes, the original gait timestamp is retrieved by backtracking. The submodule calculates the time difference between two consecutive foot strikes, i.e., the gait interval. Simultaneously, the submodule extracts the displacement path direction vectors of two adjacent single-step segments and calculates the angle between the two vectors as the directional transfer difference. Subsequently, the submodule calculates the combined difference. The calculation logic is as follows: first, the absolute value of the difference between the current gait interval and the standard gait interval (preset to 1.0 second) is calculated and recorded as the time deviation; then, the angle value (unit: degrees) of the directional transfer difference is divided by 10 for normalization and recorded as the spatial deviation; finally, the time deviation and spatial deviation are added together to obtain the combined difference. For example, if the current gait interval is 1.3 seconds and the standard interval is 1.0 second, then the time deviation is 0.3 seconds; if the directional transfer angle is 20 degrees, which is 2 after normalization; then the combined difference is 0.3 plus 2, i.e., 2.3. The submodule stores the 0.3-second time deviation, 20-degree directional difference, and 2.3-fold combination difference calculated for each gait segment into the rhythm segment time and path difference matrix.

[0041] The gait coherence generation submodule is used to perform step S303: based on the rhythm segment time and path difference matrix, filter segments whose interval time deviation exceeds a preset time delay threshold and whose path direction deviation exceeds a preset spatial offset threshold, arrange the abnormal combination segment identifiers according to the sequence recombination method, and establish a gait incoherence sequence.

[0042] The submodule performs a filtering operation based on the time and path difference matrix of rhythmic segments. The submodule presets a time delay threshold of 0.2 seconds and a spatial offset threshold of 15 degrees. It iterates through the matrix, filtering segments with time deviations greater than 0.2 seconds and directional shift differences greater than 15 degrees. For the filtered segments, the submodule rearranges them according to their chronological order of occurrence and attaches the corresponding ID identifier (e.g., "L1 vertebral abnormality") from the abnormal morphology combination information to the corresponding segment. For example, the system identifies steps 5 and 8 as meeting the above dual exceedance conditions, with step 5 associated with L1 vertebral compression and step 8 associated with L2 vertebral compression. The submodule sequentially connects "step 5-L1" and "step 8-L2" to establish a gait discontinuity sequence. This sequence reflects not only the temporal discontinuity of gait but also its pathological association in spatial structure. Experimental data shows that this dual threshold filtering mechanism can effectively eliminate occasional non-pathological gait fluctuations.

[0043] The action identification module 40 is used to execute step S40: filtering and advancing movements based on the gait discontinuity sequence and reference trajectory time and acceleration data, constructing an action performance structure from action classification, position, and center of gravity fluctuation markers. The action performance structure includes action classification name, start and end positions, center of gravity fluctuation markers, initial acceleration segment of the action trajectory, reverse path segment, and longitudinal axis curvature segment. Specifically, the action identification module includes: a trajectory pattern recognition submodule, an inertial difference matching submodule, and an action structure generation submodule. Please refer to [link to relevant documentation]. Figure 7 Step S40 specifically includes S401-403.

[0044] The trajectory pattern recognition submodule executes step S401: Based on the differences in the path extension direction and the take-off and landing sequence in the gait discontinuous sequence, the position indices of the continuous acceleration segment, the reverse segment and the longitudinal axis bending segment in the trajectory are extracted, the start and end nodes of the trajectory segment are marked and the path segments are divided, and the action trajectory pattern segment index group is generated.

[0045] Based on discontinuous gait sequences, this submodule performs in-depth analysis of the path coordinate data. It aims to segment continuous trajectory flows into semantically meaningful action segments. The submodule calculates the instantaneous velocity and direction angle between adjacent points on the path. If the velocity of five consecutive sampling points shows a monotonically increasing trend with an increment exceeding 0.1 m / s, it is marked as a "continuous acceleration segment"; if the angular velocity experiences a rapid abrupt change greater than ±400° / s within a certain time period, and the direction changes by more than 120°, it is marked as a "reverse segment"; if the displacement of a trajectory segment along the vertical axis exhibits a sinusoidal wave shape with an amplitude exceeding 0.05 m, it is marked as a "vertical axis bending segment." The submodule records the start and end time indexes of these special segments. For example, if the velocity is detected to increase from 0.5 m / s to 1.2 m / s between 3.5 and 4.0 seconds on the time axis, it is marked as an acceleration segment index. The submodule summarizes all identified special segment indices to generate action trajectory pattern segment index groups.

[0046] The inertial difference matching submodule is used to execute step S402: according to the motion trajectory pattern segment index group, obtain the acceleration data sequence corresponding to the segment, extract the acceleration change value according to the time range and calculate the variance, determine whether the range of change concentration exceeds the preset fluctuation benchmark value, and filter the types that constitute the propulsion structure difference in the corresponding path to generate a propulsion structure motion marker set.

[0047] The submodule invokes the motion trajectory mode segment index group and extracts the corresponding triaxial accelerometer data sequence from the original database for each index segment. The submodule calculates the variance of this sequence along the X, Y, and Z axes to characterize the smoothness of the motion. The calculation logic is as follows: first, calculate the average acceleration within the segment; then, calculate the square of the difference between the acceleration at each sampling point and the average value; sum these values ​​and divide by the number of sampling points to obtain the variance. The submodule presets a fluctuation baseline value of 0.05 meters per second squared. If the calculated variance exceeds this baseline value, the motion is determined to be a violently fluctuating motion. The submodule further filters the motion type based on the fluctuation axis. If the variance of the Z-axis (vertical axis) is significantly greater than that of the X and Y axes, it is classified as "vertical impact type"; if the variance of the X-axis (forward and backward axis) is dominant, it is classified as "forward and backward propulsion type". For example, if the calculated variance of the Z-axis acceleration in a certain segment is 0.08, exceeding the baseline value of 0.05, and the variances of the X and Y axes are only 0.01, then this segment is marked as "vertical impact type". The submodule associates these tags with the original index to generate a set of propulsion structure action tags.

[0048] The motion structure generation submodule extracts the classification name, start and end time index and center of gravity coordinate sequence change of the motion segment based on the motion mark set of the propulsion structure, marks the fluctuation state according to the center of gravity axial offset, and combines and outputs the name, position and center of gravity identifier corresponding to the motion to establish the motion performance structure. Based on the set of propulsion structure action markers and combined with the human kinematics knowledge base, the action is semantically named. The submodule extracts the start and end time indices of the action segment and calculates the three-dimensional coordinate change trajectory of the center of gravity within that segment. The submodule calculates the maximum offset of the center of gravity on the vertical axis, i.e., the difference between the highest and lowest Z-coordinates. If this offset is greater than 0.4 meters and accompanied by a "vertical impact" marker, the system classifies the action as "Stand Up" or "Sit Down"; if the offset is less than 0.1 meters and accompanied by a "continuous acceleration segment" marker, it is classified as "Walking Propulsion". The submodule combines and encapsulates the classification name (e.g., "Stand Up"), position information (e.g., "coordinate interval [2.5, 3.0]"), and center of gravity identifier (e.g., "center of gravity rises 0.45 meters"). For example, the system outputs the structure: {Name: "Stand Up", Time: 4.5s-5.2s, Center of Gravity Change: +0.48m, State: Unstable}. Finally, the action performance structure is established.

[0049] The fall detection module 50 is used to execute step S50: determining the action trigger combination based on the action performance structure, analyzing the displacement direction and pressure point location of the action trigger combination, and generating a fall risk identification result. The fall risk identification result includes the action trigger combination, spatial displacement direction, temporal location of the structural pressure point, and fall risk index. The fall detection module 50 includes: an action combination screening submodule, a pressure point location submodule, and a fall risk output submodule. Please refer to [link / reference]. Figure 8 Step S50 specifically includes S501-503.

[0050] The action combination screening submodule is used to perform step S501: Based on the action category name and center of gravity fluctuation mark in the action performance structure, extract the combination sequence between adjacent actions on the time axis, determine whether there are segments in the continuous action sequence that include center of gravity changes, mark the action combinations that are expected to trigger body imbalance in time order, and generate an abnormal action trigger combination index sequence.

[0051] Based on the action performance structure, the module scans the sequence patterns of adjacent actions on the timeline. The submodule focuses on identifying specific sequences that may lead to imbalance, such as a "standing up" action immediately followed by a "reverse segment" or "sudden deceleration." The submodule calculates the continuity of center of gravity changes in consecutive action sequences. If the center of gravity rises in the previous action and then drops vertically by more than 0.2 meters within 0.5 seconds in the next action, it is considered to have a characteristic of center of gravity instability. The submodule marks this high-risk combination. For example, if the sequence is detected as: Action A (standing up, end time 5.2s) immediately followed by Action B (turning around, start time 5.3s), and the center of gravity drops vertically by 0.3 meters at 5.4s, the submodule marks "Action A + Action B" as an estimated action combination that will trigger body imbalance, records its timestamp, and generates an index sequence of abnormal action trigger combinations.

[0052] The compression point location submodule is used to execute step S50: based on the abnormal action triggering the combined index sequence, extract the path extension direction vector in the combination and the range of inertial data change within the corresponding time window, combine the preset vertebral compression area coordinate mapping rules, locate the time point where the direction shift and impact fluctuation overlap, and obtain the structural compression sensitive spatiotemporal location matrix. The abnormal action triggers a combined index sequence and introduces acceleration data. The submodule extracts the path extension direction vector (i.e., the main direction of human movement) and acceleration impact vector within the combined event period. The submodule applies a preset vertebral compression area coordinate mapping rule. This rule, based on the vertebral anatomy, maps the pressure-sensitive area to a spatial positioning reference model in a unified coordinate system, identifying the compression point and aligning it with the mapping rule obtained from the analysis. This rule, based on a human biomechanical model, maps the external impact vector to the vertebral coordinate system. The calculation logic is as follows: calculate the dot product of the impact vector and the vertebral normal vector to estimate the vertical compression component. For example, when the acceleration impact vector is 15 m / s², and its direction forms a 30-degree angle with the normal of the L2 vertebral endplate, the vertical compression component is 15 multiplied by the cosine of 30 degrees (approximately 0.866), i.e., 12.99 m / s². The submodule locates the time point where this compression component overlaps with the weak point in the vertebral structure (the high-compression vertebra identified by the morphological recognition module). If the L2 vertebral body itself has been identified as having an abnormal shape, and it has been subjected to a compressive impact of 12.99 meters per second squared at 5.4s, the submodule generates a structural compression-sensitive spatiotemporal positioning matrix to accurately locate "5.4s-L2 vertebral body-12.99 intensity".

[0053] The fall risk output submodule is used to execute step S503: based on the structural compression sensitive spatiotemporal positioning matrix, filter the positioning indexes that simultaneously meet the preset compression intensity threshold and trigger action combination, classify the risk types formed by the trigger action combination and compression position change according to the time series, distinguish the risk sources of thoracic and lumbar vertebral compression fracture and osteoporosis, and establish a fall risk identification result set.

[0054] The final decision is made based on the structural compression-sensitive spatiotemporal positioning matrix. A preset compression intensity threshold of 1.5 m / s² is used. Indexes in the matrix with intensity values ​​exceeding this threshold are filtered out. For the filtered items, the submodule distinguishes the source of risk based on associated bone mineral density data (T-score). If the T-score is less than -2.5, it is considered high risk; if the T-score is greater than -1.0 but the compression intensity is extremely high (e.g., exceeding 3.0 m / s²), it is also considered high risk. Simultaneously, combined with vertebral compression history, a "proneness to thoracolumbar compression fractures" is identified. For example, for the aforementioned 5.4s L2 vertebral event, the compression intensity of 3.25 exceeds the threshold of 1.5, and the found T-score is -2.8; the system will output the identification result: "High Risk - L2 Vertebral Compression Fracture Warning." This result set is formatted and output, establishing a fall risk identification result set to provide quantitative evidence for clinical intervention. To further improve this quantitative evidence and achieve multi-dimensional risk grading, the system introduces a comprehensive scoring process. The process references the Morse Fall Risk Assessment Scale, first assigning weighted scores to the patient's six key indicators: 1. History of Falls: 0 points for no fall history, 25 points for a fall history; 0 points for no more than one medical diagnosis, 15 points for one diagnosis. 2. Assistive Devices and Treatment: 0 points for not using or being bedridden / assisted by others, 15 points for using a cane / walking stick, 30 points for walking with furniture support; 0 points for no intravenous infusion / indwelling catheter, 20 points for one catheter. 3. Physical and Cognitive Status: 0 points for normal gait / bedridden / wheelchair-bound, 10 points for weakness, 20 points for functional impairment; 0 points for normal cognitive ability, 15 points for overestimating one's abilities or forgetting limitations. The system sums up the scores from the above items and determines the risk level based on the total score, then matches preventative measures accordingly: a total score of 0 to 24 points indicates low risk (low fall risk), requiring general preventative measures (such as environmental information and wearing non-slip shoes); a total score of 25 to 44 points indicates medium risk (moderate fall risk), requiring standard preventative measures (such as placing warning signs at the bedside and having someone accompany the patient); a total score of 45 points or higher indicates high risk (high fall risk), requiring high-risk protective measures (such as raising bed rails, restricting movement, or appropriately restraining the patient). This comprehensive score, combined with sensor data, constitutes a complete risk defense system.

[0055] Table 2 shows the risk assessment logic and calculation examples of the fall detection module:

[0056] Table 2. Fall Risk Assessment Logic Table Table 2 illustrates how the system integrates physical impact parameters and physiological bone parameters to output the final fall and fracture risk category through three typical combinations of actions and data.

[0057] In summary, this invention, by integrating dynamic data collected from inertial sensors, pressure sensors, and imaging devices, can comprehensively and in real-time track and analyze a patient's gait, body deviation, support contact, and other physiological parameters, accurately reflecting an individual's postural changes and skeletal health during movement. Through multimodal data fusion, it can identify potential fall risks and health abnormalities, overcoming the limitations of traditional single-dimensional analysis and improving the accuracy of risk assessment. This solution provides a dynamic and personalized health management plan, enabling precise intervention based on real-time monitoring, avoiding misjudgments and delays, and improving the timeliness and scientific rigor of assessment and intervention.

[0058] Based on the same inventive concept described above, the present invention also provides an electronic device, which may be a server, a desktop computing device, or a mobile computing device (e.g., a laptop computing device, a handheld computing device, a tablet computer, a netbook, etc.). This device includes one or more processors and a memory, wherein the processor is used to execute a program to implement the aforementioned fall risk identification method; and the memory is used to store a computer program executable by the processor.

[0059] Based on the same inventive concept, the present invention also provides a computer-readable storage medium corresponding to the aforementioned embodiments of the fall risk identification method, wherein the computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps described in any of the above embodiments.

[0060] This application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a fall risk identification method as described in any of the above embodiments.

[0061] This invention can take the form of a computer program product implemented on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing program code. Computer-readable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to: phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0062] The embodiments described above are merely examples of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and the present invention also intends to include these modifications and variations.

Claims

1. A method for identifying fall risk, characterized in that, include: S10: Acquire gait data, X-ray and bone density images for continuous and temporal analysis, gait description data structure; S20: Calculate the offset of the offset line based on the gait description data structure, and filter deformed nodes by combining the X-ray image; Obtain information on abnormal morphological combinations; S30: Generate a discontinuous gait sequence based on the abnormal morphological information and temporal features; S40: Based on the gait discontinuity sequence, reference trajectory time and acceleration data, filter and propel the changing movements, and construct the movement performance structure by movement classification, position and center of gravity fluctuation markers; S50: Determine the action trigger combination based on the action performance structure, analyze the displacement direction and pressure point location of the action trigger combination, and generate a fall risk identification result.

2. The fall risk identification method according to claim 1, characterized in that: The gait description data structure includes positional continuity, support contact timing, X-ray image index, and bone density image index. The abnormal morphology combination information includes the number of times the vertebral body offset line turns, the span of the vertebral body offset line, the contraction range of the vertebral body edge line segment, the number of protruding curve points in the longitudinal direction, the spatial deformation of the support node, and the number of vertebral body structural changes. The gait discontinuity sequence includes the location of changes in the distribution density of the protrusion point, the difference in the path extension direction between the gait interruption point and the adjacent segment, and the arrangement and combination of time delay and spatial offset. The motion performance structure includes motion category name, start and end positions, center of gravity fluctuation mark, motion trajectory initial acceleration segment, path reversal segment, and longitudinal axis bending segment. The risk identification results include action trigger combinations, spatial displacement direction, time location of structural pressure points, and fall risk index.

3. The fall risk identification method according to claim 2, characterized in that, Step S10 includes: S101: Acquire acceleration data, perform a second integration on the acceleration data according to the time sequence, calculate the trajectory variation frequency by combining the direction change rate, and generate a continuous displacement path sequence and a variation frequency parameter set; S102: Based on the continuous displacement path sequence and the set of changing frequency parameters, combined with pressure data, extract the contact area sequence according to a single-step cycle, calculate the offset angle of the trajectory corresponding to the contact point, and generate a support contact timing index and gait offset angle sequence. S103: Based on the support contact timing index and gait offset angle sequence, match the X-ray and bone density images under time synchronization, filter the imaging data corresponding to the time, and establish a gait description data structure.

4. The fall risk identification method according to claim 3, characterized in that, Step S20 includes: S201: Obtain the position continuity and support contact timing in the gait description data structure, extract the trajectory segment within the support cycle, calculate the number of trajectory turning changes and the corresponding displacement span, and generate a sequence of the number of turning changes and span within the support cycle; S202; Based on the number of turning cycles and span sequence of the support cycle, and combined with the vertebral body edge segments in the X-ray image, the compression amplitude and the number of longitudinal curvature points of each vertebral body are counted to generate a vertebral body structure compression amplitude and curvature point number comparison matrix. S203: Based on the comparison matrix of the compression amplitude and curvature points of the vertebral structure, support nodes that exceed the preset benchmark threshold are selected, and correlation analysis is performed in combination with the corresponding number of turns and span to establish an abnormal morphology combination information set.

5. The fall risk identification system according to claim 4, characterized in that, Step S30 includes: S301: Obtain the offset path in the abnormal morphology combination information, extract the spatial position sequence corresponding to the continuous segment of gait rhythm, detect the change in the temporal distribution density of the sudden point in the trajectory, count the number of sudden point locations and label the index according to the time axis sliding window, and generate a temporal index sequence of rhythm density change. The continuous segment of gait rhythm includes a continuous movement interval in which the gait rhythm remains relatively stable and there are no significant sudden changes within a certain time window. S302: Based on the rhythm density change time index sequence, obtain the time sequence of start and stop between consecutive steps, extract the difference in path direction and start-up / fall sequence according to adjacent segments, calculate the combined difference in time interval and direction transfer of each group of segments, and generate a rhythm segment time and path difference matrix. S303: Based on the rhythm segment time and path difference matrix, filter segments whose interval time deviation exceeds a preset time delay threshold and whose path direction deviation exceeds a preset spatial offset threshold, arrange the abnormal combination segment identifiers according to the sequence recombination method, and establish a gait discontinuity sequence.

6. The fall risk identification method according to claim 5, characterized in that, Step S40 includes: S401: Based on the differences in the path extension direction and take-off and landing sequence in the discontinuous gait sequence, extract the position indices of the continuous acceleration segment, reverse segment and longitudinal axis bending segment in the trajectory, mark the start and end nodes of the trajectory segment and divide the path into segments, and generate the motion trajectory pattern segment index group. S402: Based on the action trajectory pattern segment index group, obtain the acceleration data sequence corresponding to the segment, extract the acceleration change value according to the time range and calculate the variance, determine whether the range of change concentration exceeds the preset fluctuation benchmark value, and filter the types that constitute the propulsion structure difference in the corresponding path to generate a propulsion structure action mark set. S403: Based on the set of propulsion structure action markers, extract the classification name, start and end time index and center of gravity coordinate sequence change of the action segment, mark the fluctuation state according to the center of gravity axial offset, combine and output the name, position and center of gravity identifier corresponding to the action, and establish the action performance structure.

7. The fall risk identification method according to claim 6, characterized in that, Step S50 includes: S501: Based on the action category name and center of gravity fluctuation mark in the action performance structure, extract the combination sequence between adjacent actions on the time axis, determine whether there are segments in the continuous action sequence that include center of gravity changes, and mark the action combination that is expected to trigger body imbalance in time order to generate an abnormal action trigger combination index sequence. S502: Invoke the abnormal action trigger combined index sequence, extract the path extension direction vector in the combination and the inertial data change range within the corresponding time window, combine the preset vertebral compression area coordinate mapping rules, locate the time point where the direction shift and impact fluctuation overlap, and obtain the structural compression sensitive spatiotemporal positioning matrix. S503: Based on the structural pressure-sensitive spatiotemporal positioning matrix, filter the positioning indices that simultaneously meet the preset pressure intensity threshold and trigger action combination, classify the risk types formed by the trigger action combination and pressure position change according to the time series, and establish a fall risk identification result set.

8. The fall risk identification method according to claim 7, characterized in that: The coordinate mapping rule for the vertebral compression area is based on the vertebral anatomical structure, mapping the pressure-sensitive area to a spatial positioning reference model in a unified coordinate system, and identifying the compression point and the mapping rule obtained from the alignment analysis.

9. A fall risk identification system, characterized in that, include: Gait acquisition module: used to acquire gait data, X-ray and bone density images for continuous and temporal analysis, gait description data structure; The morphology recognition module is used to calculate the offset of the offset line based on the gait description data structure, and to filter deformed nodes by combining the X-ray image; thus obtaining abnormal morphology combination information. Rhythm analysis module: used to generate gait discontinuity sequences based on the abnormal morphological information and temporal features; Action identification module: used to filter and promote changes in actions based on the gait discontinuity sequence, reference trajectory time and acceleration data, and construct the action performance structure by action classification, position and center of gravity fluctuation markers; Fall detection module: used to determine the action trigger combination based on the action performance structure, and analyze the displacement direction and pressure point location of the action trigger combination to generate fall risk identification results.

10. A fall risk identification device, characterized in that, include: The system includes an inertial sensor, a pressure sensor, an imaging device, and the fall risk identification system as described in claim 9, wherein the inertial sensor, pressure sensor, and imaging device are respectively connected to the gait acquisition module; The inertial sensor is used to detect the acceleration of the corresponding part of the object to be detected; the pressure sensor is used to detect the pressure data of the corresponding part of the object to be detected; the imaging device is used to perform continuous X-ray image imaging and bone density image imaging of the image to be detected.

Images