Human stumble risk identification method based on minimum distance between foot and ground detection

Abstract

This invention provides a method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground. First, markers are placed on the shoe surface, and the point cloud and 3D coordinate information of these markers are acquired. Then, matrix transformations are performed on the point cloud and coordinate information based on the markers undergoing rigid body motion to divide gait cycles. Finally, in each gait cycle, the motion trajectory of the markers is found, and the local minimum cross-sectional area (MFC) of each point in the current gait cycle is calculated. The minimum MFC of this local MFC is subtracted from the ground calibration distance, and the minimum MFC of the swing phase of the current gait cycle is obtained. This process is repeated to obtain the MFC of all gait cycles, thereby determining whether the subject belongs to a high-risk tripping population. This invention can be used to identify individuals at risk of tripping and implement effective tripping prevention strategies. It can also be used for the diagnosis of foot diseases and the evaluation of rehabilitation treatment effects. In gait rehabilitation, it ensures safer walking by reducing the tripping risk of a range of individuals with gait impairments.

Description

Technical Field

[0001] This invention belongs to the field of biomedical engineering, and in particular relates to a method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground. Background Technology

[0002] In my country, falls are one of the leading causes of injury-related death among the elderly. Studies have shown that 53% of falls in healthy older adults are caused by tripping (Blake AJ, Morgan K, Bendall MJ, Dallosso H, Ebrahim SBJ, Arie THD, et al. Falls by elderly people at home: prevalence and associated factors. AgeAgeing 1988; 17: 365–72.). Minimum foot-to-ground distance (MFC) is a gait variable associated with tripping, occurring at the moment the foot moves with maximum horizontal speed. MFC refers to the minimum vertical distance between the lowest point of the swinging foot and the walking surface, occurring approximately in the middle of the swing phase of the gait cycle. A decrease in MFC during walking increases the likelihood of tripping over unseen obstacles or due to undetected changes in surface height.

[0003] Wearable tripping detection systems have been a research hotspot in the field of biomedical engineering. The wearable tripping detection system disclosed in patent application number 201611140459.7 is small in size, low in power consumption, and lightweight. It can measure gait parameters such as heel-to-toe height, toe-to-toe height during the swing phase, and other gait parameters in real time using only two distance sensors, and detect tripping in real time. The average error and root mean square error (MFC) between the measured values ​​and the true values ​​of this wearable tripping detection system are both close to 10 mm. Karst et al. (Karst GM, Hageman PA, Jones TF, Bunner SH. Reliability of foot trajectory measures within and between testing sessions. J Gerontol 1999; 54: 343–7.) reported an average MFC of 12.9 mm. An error close to 10 mm is too large for tripping detection, which should be accurate to the millimeter level.

[0004] Different researchers have proposed different representations of the Minimum Toe Clearance (MFC). Rezaul K. Begg (Oren Tirosh, Amity Campbell, Rezaulk. Begg, and W.A. Sparrow., 2013. Biofeedback Training Effects on Minimum Toe Clearance Variability During Treadmill Walking.) proposed defining the MFC as the first minimum vertical displacement of a marker point at the 5th metatarsal head during the oscillation phase. This estimation method itself cannot represent the point where the shoe is closest to the ground. Rezaul K. Begg (Begg, R., Best, R., Dell'Oro, L., Taylor, S., 2007. Minimum foot clearance during walking: strategies for minimisation of trip-related falls. Gait and Posture 25(2), 191–198.) used a two-dimensional mathematical modeling method to calculate the MFC. The marker location and shoe size were used to predict the lowest point on the shoe using a geometric model. The shoe's geometric model was used to calculate the y-coordinate of the lowest point of the shoe (PTP, a virtual point representing the farthest and lowest edge of the shoe) from markers at the big toe (P1) and fifth metatarsal head (P2) using constant triangle geometry. The geometric model used in this study more accurately represents the lowest point of the foot at the moment of MFC (Majestic Foot Flow). However, this method effectively fixes the position of the PTP on the shoe, while the positions of the farthest and lowest edges of the shoe can vary in different gait cycles at the moment of MFC. Furthermore, foot movement is complex, and foot movements during walking cause shoe deformation. The study's use of averaging distances d1(P1-P2), d2(P2-PTP), and d3(P1-PTP) does not eliminate the calculation error caused by shoe deformation. Because MFC measurements should be accurate to the millimeter level while shoe deformation can be on the centimeter level, the average value strategy results in significant calculation errors. Additionally, this method does not actually find the MFC for every gait cycle; it only predicts the PTP on the shoe using the geometric model, marker positions, and shoe size. This is only an approximate estimate of the swing phase MFC, therefore, 3D shoe modeling techniques are needed to accurately represent the MFC. Summary of the Invention

[0005] To address the aforementioned problems, this invention provides a method for tripping risk identification based on precise representation of MFC. The technical solution adopted in this invention is as follows:

[0006] This invention provides a method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground. It obtains the minimum distance between the subject's foot and the ground (MFC) value, and identifies the risk of tripping based on the MFC value. The method includes the following steps: Step S1, placing a marker point on the subject's shoe surface, ensuring the marker point undergoes rigid body motion without deformation during movement; performing 3D modeling of the shoe to obtain the point cloud information of the marker point; and obtaining the set of three-dimensional coordinate points {m} of the marker point's center based on the point cloud information. i Step S2: Collect test videos of the subject walking. Based on the algorithm for obtaining the rigid body motion transformation matrix, calculate the set of three-dimensional coordinates of the center of the marker points and the set of coordinates of the center of the marker points in each frame of the test video {d}. i The transformation matrix {A} i Step S3: Calculate the set of sphere center coordinates {d} for each frame marker point. i The transformation matrix between every two frames {B i Step S4: Divide the gait cycle according to any marker point, use the transformation matrix to recover all frames of the current gait cycle, obtain the gait trajectory data of each marker point on the shoe in the current gait cycle, and use this to find the lowest point of the foot of each marker point in the current gait cycle. Repeat the gait cycle to obtain the lowest point of the foot MFC of all gait cycles. Step S5: Determine whether the subject belongs to the easy-to-trip population based on the lowest point of the foot MFC of all gait cycles, and mark the subject's gait as healthy gait or tripping gait based on the determination result.

[0007] The present invention provides a method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground, which may also have the following technical features: in step S1, the point cloud information is obtained by scanning the 3D model of the shoe obtained by modeling using 3D image generation and editing software, and segmenting the shoe and all marked points, and reading the point cloud information of the 3D model of the shoe using programming software.

[0008] The present invention provides a method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground, which may also have the following technical features, wherein, in step S1, the set of three-dimensional coordinate points of the center of the ball is calculated by using the minimum bounding sphere algorithm.

[0009] The present invention provides a method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground, which also has the following technical feature, wherein the expression of the rigid body motion change matrix is ​​as follows:

[0010] d i =Rm i +T+V i (1)

[0011] In the formula, {m i} and {d i Let {V} be two corresponding sets of marker points, i = 1...N, such that they are related by the above equation, R is a standard 3×3 rotation matrix, T is a 3-D translation vector, and {V} i} is a noise vector, and the solution is to find the set {m i}Mapped to {d i The optimal transformation is minimized using the least squares error criterion:

[0012]

[0013] As a result of the least squares solution to Eq.2, the point set {d} i} and {m i The centroids should be the same. Using this constraint, a new equation can be generated by definition:

[0014]

[0015] In the formula, and They are the point sets {d i} and {m i The centroid of} is used to rewrite and simplify Eq.2 as follows:

[0016]

[0017] When the last term is maximized, the equation is minimized, that is, the trace is maximized. Where H is the correlation matrix defined as follows:

[0018]

[0019] If the singular value decomposition of H is H=U∧V T Given the optimal rotation matrix that maximizes the desired trace, then... for:

[0020]

[0021] The above formula is The expression when the determinant is +1, when The determinant is -1, meaning that when the point set is planar or contains a lot of noise, the required rotation is:

[0022]

[0023] In the formula, matrix V′=[v1, v2, -v3] consists of columns of V, where v3 is the column corresponding to the singular value of H being zero. The optimal representation of the final rotation matrix is:

[0024]

[0025] The set to be collected soon {d i The centroid of} and the set {m i Aligning the centroids of rotation is the optimal translation:

[0026]

[0027] The present invention provides a method for identifying human tripping risks based on the detection of the minimum distance between the foot and the ground. This method also has the following technical features: the gait cycle is defined as the movement of the same foot from heel strike to heel strike again. In step S4, the process of recovering all frames of the current gait cycle using a transformation matrix is ​​as follows: each time, the first frame of the current gait cycle is retrieved from {A}. i The transformation matrix in} is used to recover the point cloud of the shoe in the 3-D modeling coordinate system in the first frame of the gait laboratory coordinate system, using { B The first transformation matrix in {i} restores the point cloud of the first frame in the gait laboratory coordinate system to the shoe point cloud of the second frame in the gait laboratory coordinate system. The process is repeated, taking {B}... i The transformation matrix in the gait laboratory coordinate system is used to recover the shoe point cloud of the previous frame in the gait laboratory coordinate system and the shoe point cloud of the next frame in the gait laboratory coordinate system.

[0028] The present invention provides a method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground. This method also includes the following technical feature: in step S5, a regression-support vector machine (MFC) system is used to detect whether the subject belongs to the tripping-prone population. The input of this MFC system is the original MFC. i The signal, i, represents the number of consecutive gait cycles. The detection principle is as follows: a linear prediction model estimated by an autoregressive process is used to simulate the object MFC. i Signal variability, model coefficient capture MFC i The signal variability features are identified and labeled to represent healthy gait or stumbling gait. These are then used as input features to train a support vector machine classifier so that it can identify the coefficients of the autoregressive model belonging to healthy gait or stumbling gait.

[0029] The present invention provides a method for identifying human tripping risks based on the detection of the minimum distance between the foot and the ground. This method may also have the following technical feature: in step S4, recovering all frames of the current gait cycle using a transformation matrix can be achieved by: extracting the transformation matrix {A} of all frames of the current gait cycle. i This allows us to recover the point cloud of the shoe in the 3D modeling coordinate system across all frames in the gait laboratory coordinate system.

[0030] Invention Function and Effect

[0031] According to the present invention, a method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground first places markers on the shoe surface and acquires the point cloud and three-dimensional coordinate information of the markers; then, a matrix transformation is performed on the point cloud and coordinate information based on the markers undergoing rigid body motion to divide the gait cycle; finally, in each gait cycle, the motion trajectory of all points on the outer surface of the shoe toe is found, and the local minimum calculus (MFC) of each point in the current gait cycle is calculated. The minimum value of this local MFC is subtracted from the ground calibration distance to obtain the MFC of the swing phase of the current gait cycle. This process is repeated to obtain the MFC of all gait cycles, thereby determining whether the subject belongs to the tripping-prone population. Existing MFC estimation methods do not truly find the MFC of each gait cycle; they only predict the MFC using geometric models, marker positions, and shoe size, thus resulting in significant errors. The MFC calculation method of the present invention, however, has high accuracy and can accurately identify the MFC of each gait cycle. Therefore, the MFC measurement of the present invention can provide early detection of gait features associated with tripping, distinguish the gait of healthy and fall-prone elderly subjects, and determine in advance when gait instability begins to increase the likelihood of accidental foot-to-ground contact leading to a trip.

[0032] In summary, the tripping risk identification method based on minimum distance detection between the foot and the ground in this invention records a large amount of raw gait data, processes it to obtain Minimum Foot-to-Ground (MFC) data, and further analyzes MFC variability. This method can be used to identify individuals at risk of tripping and implement effective tripping prevention strategies. It can also be used for the diagnosis of foot diseases and the evaluation of rehabilitation treatment effects. In gait rehabilitation, it ensures safer walking by reducing the tripping risk of a range of individuals with gait impairments. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating the human tripping risk identification method based on the detection of the minimum distance between the foot and the ground in an embodiment of the present invention.

[0034] Figure 2 This is a schematic diagram of shoe 3D modeling in an embodiment of the present invention;

[0035] Figure 3 This is a comparison chart of the error between the measurement results of the transformation matrix algorithm and the measurement results of the VICON motion capture system in this embodiment of the invention;

[0036] Figure 4 This is a schematic diagram illustrating the vertical displacement change between the big toe marker and the actual lowest point of the foot during one gait cycle in an embodiment of the present invention.

[0037] Figure 5 This is a schematic diagram of the calculation process for MFC values ​​in an embodiment of the present invention; and

[0038] Figure 6This is a flowchart illustrating the tripping risk identification using a hybrid AR-SVM system in an embodiment of the present invention. Detailed Implementation

[0039] To address the accuracy issues of existing technologies, this invention provides a method for accurately representing the MFC (Morphological Flow Factor). This is the only complete solution to date that accurately identifies the MFC for each gait cycle. The method is equipped with specialized data analysis software, boasts high accuracy and precision, and is simple to operate, making it the gold standard for tripping detection.

[0040] Experimental Environment and Equipment: This embodiment was conducted in a gait laboratory. The equipment included a treadmill, a 3D motion capture system, and shoes. The treadmill had an on / off switch and speed adjustment buttons to control gait speed, stride length, and step count. Spherical markers (non-collinear) were also installed on the treadmill for distance calibration. The 3D motion capture system was equipped with specialized data analysis software to accurately record the 3D motion trajectory of each marker point. The shoes were used for 3D modeling.

[0041] To make the technical means, creative features, objectives and effects of the present invention easy to understand, the following describes in detail the human tripping risk identification method based on the detection of the minimum distance between the foot and the ground of the present invention with reference to the embodiments and accompanying drawings.

[0042] <Example>

[0043] The human tripping risk identification method in this embodiment identifies the risk of tripping based on the minimum distance between the subject's foot and the ground (MFC value).

[0044] Figure 1 This is a flowchart illustrating the human tripping risk identification method based on the detection of the minimum distance between the foot and the ground in an embodiment of the present invention.

[0045] like Figure 1 As shown, the method for identifying human tripping risks based on the detection of the minimum distance between the foot and the ground includes the following steps:

[0046] Step S1: Place 10 marker points at different locations on the subject's shoe surface, ensuring that these marker points undergo rigid body motion without deformation during movement. Perform 3D modeling of the shoe to obtain the point cloud information of the marker points. Based on the point cloud information, obtain the set of three-dimensional coordinate points {m} of the center of the sphere of the marker points. i}

[0047] Figure 2 This is a schematic diagram of shoe 3D modeling in an embodiment of the present invention.

[0048] In this embodiment, several markers are placed on relatively hard areas of the shoe surface to ensure that the positions of the markers hardly deform during movement. For example... Figure 2 As shown, the following explanation uses 10 marker points as an example. Marker point 1 is attached to the big toe; marker point 2 is attached to the farthest edge of the toe box; marker point 3 is attached to the fifth toe; marker point 4 is attached to the head of the fifth metatarsal; marker point 5 is attached to the base of the fifth metatarsal; marker point 6 is attached to the navicular bone; marker point 7 is attached to the talus; marker point 8 is attached to the calcaneus; marker points 9 and 10 are attached to the calcaneus in an equilateral triangle with marker point 8.

[0049] Subjects first wore the shoes to adapt for a period of time, such as 30 minutes, and then walked at a comfortable pace for 10 minutes in a gait laboratory, such as VICON, obtaining a set of 10 sphere center coordinates for each marker. i}, a total of n rows of data were obtained, and this data {d i Store the data in an Excel spreadsheet format with n rows and 30 columns (each three columns correspond to the three-dimensional coordinates of a marker point).

[0050] In this embodiment, a .stl model is obtained by scanning the shoe using a CT scan model. The CT scan model is then processed using Mimics software to segment the swing shoe and all marked points. Specifically:

[0051] like Figure 2 As shown, the swing shoe is roughly divided into three parts: forefoot, midfoot, and heel. Marker points 1, 2, and 3 are used for 3D modeling of the forefoot; marker points 4, 5, and 6 are used for 3D modeling of the midfoot; and marker points 7, 8, 9, and 10 are used for 3D modeling of the heel.

[0052] Then, MATLAB is used to read the shoe.stl model and obtain v and f. Among them, v (141209 rows, 3 columns): each row specifies the three-dimensional x, y, z coordinates of a vertex; f (283726 rows, 3 columns): each row is defined as the index of the three vertices that need to be connected.

[0053] Then, using the minimum bounding sphere algorithm written in MATLAB, a set of points {m} with the 3D coordinates of the sphere's center in the CT coordinate system for 10 marked points is fitted. i}, {m i (1 row, 30 columns): Every three columns correspond to the three-dimensional coordinates of a marker point.

[0054] Step S2: Collect test videos of the subject walking. Using the algorithm for obtaining the rigid body motion transformation matrix, calculate the set of three-dimensional coordinates of the center of the marker points and the set of coordinates of the center of the marker points in each frame of the test video {d}. i The transformation matrix {A} i}

[0055] In this embodiment, based on the algorithm for obtaining the rigid body motion transformation matrix written in MATLAB, the set of fitted center coordinates of the sphere {m} for 10 marked points in the forefoot, midfoot, and heel in the CT coordinate system is obtained. i} and the set of sphere center coordinates of the marker points corresponding to each frame obtained by VICON Gait Lab {d i The transformation matrix A (4 rows, 4 columns) of A is formed by the first 3 rows and the first 3 columns of A. The first three rows of the fourth column of A correspond to the translation vectors. The last row and 4 columns are fixed values ​​[0, 0, 0, 1]. The transformation matrix of A for all frames is denoted as DATA and stored sequentially in the form of an Excel spreadsheet with 4n rows and 4 columns (each 4 rows corresponds to the transformation matrix of one frame).

[0056] The algorithmic expression for the rigid body motion transformation matrix is ​​as follows:

[0057] d i =Rm i +T+V i (1)

[0058] Two corresponding sets of marker points {m i} and {d i}, i = 1...N, such that they are related in the above way, where R is a standard 3×3 rotation matrix, T is a 3-D translation vector, {V i} is the noise vector.

[0059] Solve for the set {m i}Mapped to {d i The optimal transformation of} typically requires minimizing the least squares error criterion given by the following equation:

[0060]

[0061] As a result of the least squares solution to Eq.2, the point set {d} i} and {m i The} should have the same centroid. Using this constraint, a new equation can be generated. By definition:

[0062]

[0063]

[0064] In the formula, and They are the point sets {d i} and {m i The center of mass of}.

[0065] Therefore, Eq.2 can be rewritten and simplified as follows:

[0066]

[0067] When the last term is maximized, the equation is minimized, which is equivalent to maximizing the trace. Where H is the correlation matrix defined as follows:

[0068]

[0069] If the singular value decomposition of H is H=U∧V T Given the optimal rotation matrix that maximizes the desired trace, then... for:

[0070]

[0071] Equation (7) above is when When the determinant is +1, the optimal rotation matrix The expression.

[0072] However, when the two sets of points are planar, or when there is a lot of noise, The determinant might become -1, indicating that reflection has been calculated instead of rotation. In this case, the required rotation can be found to be... The matrix V′ = [v1, v2, -v3] consists of columns of V, where v3 is the column corresponding to the singular value of zero in H.

[0073] Therefore, the optimal representation of the rotation matrix is:

[0074]

[0075] As before, the set {d i The centroid of} and the set {m i Aligning the centroids of rotation is the optimal translation:

[0076]

[0077] Figure 3 This is a comparison chart of the error between the measurement results of the transformation matrix algorithm and the measurement results of the VICON motion capture system in this embodiment of the invention.

[0078] In this embodiment, the accuracy of the transformation matrix algorithm was experimentally verified: the measured value (true value) was taken as the vertical displacement of the heel and big toe markers of the VICON optical motion capture system minus the ground reference distance, to verify the reliability of the transformation matrix algorithm. The measured value is the ground clearance height calculated by the transformation matrix algorithm disclosed in this invention, which is the vertical displacement of the heel and toe markers minus the ground reference distance. Figure 3 As shown, the average error and root mean square error between the measured value and the true value do not exceed 0.3 mm.

[0079] Step S3, calculate the set of sphere center coordinates {d} for each frame of marker points. i The transformation matrix between every two frames {B i}

[0080] Step S4: Based on the shoe dot cloud information, the gait cycle is divided by the vertical displacement coordinate of the heel marker. The transformation matrix is ​​used to recover all frames of the current gait cycle, and the gait trajectory data of each marker on the shoe in the current gait cycle is obtained. The lowest point of the foot at each marker in the current gait cycle is then obtained. The gait cycles are iterated to obtain the lowest point MFC of the foot in all gait cycles.

[0081] Figure 4 This is a schematic diagram illustrating the vertical displacement change between the big toe marker and the actual lowest point of the foot during one gait cycle in an embodiment of the present invention.

[0082] like Figure 4 As shown, the Minimum Vertical Distance (MFC) is the minimum vertical distance between the lowest point of the swing leg and the walking surface during the swing phase of the gait cycle. MFC is a gait variable, representing the moment the foot moves at its maximum horizontal speed. It is related to tripping mechanisms because a decrease in MFC during walking increases the likelihood of tripping over unseen obstacles or due to undetected changes in surface height. A gait cycle is defined as the movement of the same foot from heel strike to heel strike again. The change in vertical displacement between the big toe marker (TM) and the walking surface is not the same as the change in vertical displacement between the lowest point of the foot (PTP) and the walking surface during walking, but the trends of both are consistent. The MFC event is now explained using the change in vertical displacement between the big toe marker (TM) and the walking surface during a gait cycle. The vertical displacement of TM is denoted as y(TM), and includes the following process:

[0083] (1) Initial landing phase to midpoint support phase: From the heel landing to the toes landing, TM decreases in the vertical direction, and y(TM) decreases.

[0084] (2) Support phase: From the time the heel leaves the ground until the toes leave the ground, TM continues to decrease in the vertical direction, y(TM) continues to decrease, and y(TM) drops to a minimum value in the vertical direction for the first time. This gait cycle event is toe-off.

[0085] (3) From the early stage of swing to the middle stage of swing: From the moment the toes of the supporting leg leave the ground until the knee joint reaches maximum flexion, TM rises in the vertical direction, and y(TM) increases to a maximum value for the first time.

[0086] (4) From the maximum flexion of the knee joint to the point where the lower leg is perpendicular to the ground, i.e. the leg accelerates forward to pass the opposite supporting leg, TM decreases in the vertical direction, and y(TM) decreases to a minimum value again. This gait cycle event is the lowest point of the foot (MFC), which occurs in the middle of the swing phase.

[0087] (5) Later stage of swing: The leg swings forward and continues to swing forward through the opposite supporting leg until the heel hits the ground, at which point y(TM) increases to a maximum value again.

[0088] (6) After that, the heel lands downwards, TM decreases vertically, and y(TM) decreases. From Figure 4 As can be seen, y(PTP) is always less than y(TM) throughout the entire gait cycle.

[0089] Figure 5 This is a flowchart of the algorithm for obtaining the MFC value of all gait cycles by iteratively taking gait cycles in an embodiment of the present invention.

[0090] like Figure 5 As shown, the specific process of this step is as follows:

[0091] The gait period (locs) is divided by reading the z-coordinate of the heel marker. locs records the frame index corresponding to the heel marker's landing point. Since the gait period is defined as the movement of the same foot from heel strike to heel strike again, each gait period corresponds to two consecutive locs values.

[0092] First, read out the transformation matrix A of the m-th frame (m=locs(i):locs(i+1)) of the i-th (i=1:size(locs)-1) state period.

[0093] Next, calculate the 3D coordinates of each point in the shoe point cloud v in the m-th frame of the i-th gait cycle: z = A * v. Store the third column of z in the m-th frame of the i-th gait cycle: Z = [Z, z(:, 3)], until all frames corresponding to the i-th gait cycle are looped. Thus, each row stored in Z corresponds to the z coordinate value of a point in the shoe point cloud in all frames of the i-th gait cycle, which is the vertical displacement of the point in all frames of the i-th gait cycle.

[0094] Then, based on the characteristics of MFC in the gait cycle, the local MFC of each point j (j=1:size(Z,1)) in the shoe point cloud in the current gait cycle is obtained. The minimum value of the lowest point of the foot in the local shoe point cloud is taken and the ground calibration distance is subtracted to obtain the MFC of the swing phase of the i-th gait cycle.

[0095] The above method of recovering all frames of the current gait cycle using the transformation matrix can also be adopted in the following way: extract the transformation matrix {A} of all frames of the current gait cycle. iThis allows us to recover the shoe point cloud in the 3D modeling coordinate system across all frames in the gait laboratory coordinate system.

[0096] Step S5: Determine whether the subject belongs to the tripping population based on the lowest point of the foot (MFC) of all gait cycles, and mark the subject's gait as healthy gait or tripping gait based on the determination result.

[0097] In this embodiment, a hybrid autoregressive-support vector machine (AR-SVM) system is used for tripping risk identification and judgment. Specifically:

[0098] Figure 6 This is a flowchart illustrating the tripping risk identification using a hybrid AR-SVM system in an embodiment of the present invention.

[0099] like Figure 6 As shown, this hybrid autoregressive-support vector machine system has an autoregressive prediction model and a support vector machine classifier. The input to the autoregressive-support vector machine system is the original MFC. i The signal, i, represents the number of consecutive gait cycles. The main detection idea is to first use a linear prediction model estimated by an autoregressive process to simulate the target MFC. i Signal variability, model coefficient capture MFC i The signal variability features are identified and labeled to represent healthy or stumbling gait. These are then used as input features to train a support vector machine classifier so that it can identify the coefficients of an autoregressive model belonging to either healthy or stumbling gait. Once the system is fully trained, it can be tested by first recording MFCs from new subjects. i Data is collected, the coefficients of the autoregressive model are obtained, and support vector machine decisions are calculated to detect stumbling gait.

[0100] Functions and effects of the embodiments

[0101] According to the tripping risk identification method based on minimum distance detection between the foot and the ground provided in this embodiment, firstly, marker points are placed on the shoe surface, and the point cloud and three-dimensional coordinate information of the marker points are obtained; then, matrix transformation is performed on the point cloud and coordinate information based on the marker points performing rigid body motion to divide the gait cycle; finally, in each gait cycle, the motion trajectory of all points on the outer surface of the toe is found, and the local minimum distance detection (MFC) of each point in the current gait cycle is calculated. The minimum value of the MFC is taken and subtracted from the ground calibration distance to obtain the MFC of the swing phase of the current gait cycle. By iteratively taking the gait phase, the MFC of all gait cycles is obtained, thereby determining whether the subject belongs to the tripping-prone population. Therefore, this method can provide early detection of tripping-related gait characteristics, distinguish the gait of healthy subjects from that of elderly subjects with a tendency to fall, and determine in advance when gait instability begins to increase the possibility of accidental foot-ground contact leading to a trip.

[0102] Furthermore, existing methods for solving the MFC (Mean Functional Capacity) do not actually find the MFC for every gait cycle; they merely predict or estimate the MFC using geometric models, marker positions, and shoe dimensions, resulting in significant errors. In this embodiment, however, by recording a large amount of raw gait data and performing rigid body motion based on marker points, the MFC data is solved. Compared to existing MFC estimation methods, this method has higher accuracy and can accurately identify the MFC for each gait cycle.

[0103] Meanwhile, it has been verified that the detection method of this embodiment has an error of less than 0.3mm, while the wearable tripping detection system in the patent application number 201611140459.7 has a measurement error as high as 10mm.

[0104] In summary, the tripping risk identification method based on minimum distance detection between the foot and the ground in this embodiment records a large amount of raw gait data, processes it to obtain Minimum Foot-to-Ground (MFC) data, and further analyzes MFC variability. This method can be used to identify individuals at risk of tripping and implement effective tripping prevention strategies. It can also be used for the diagnosis of foot diseases and the evaluation of rehabilitation treatment effects. In gait rehabilitation, it ensures safer walking by reducing the tripping risk of a range of individuals with gait impairments.

[0105] The above embodiments are only used to illustrate specific implementations of the present invention, and the present invention is not limited to the scope of the description of the above embodiments.

Claims

1. A method for identifying human tripping risk based on minimum distance detection between the foot and the ground, used to obtain the minimum distance between the subject's foot and the ground (MFC) value, and to identify the risk of the subject tripping based on the MFC value, characterized in that, Includes the following steps: Step S1: Place several marker points on the upper of the subject's shoe, ensuring that all marker points undergo rigid body motion without deformation during movement. Perform 3D modeling of the shoe to obtain the point cloud information of the marker points. Based on the point cloud information, obtain the set of three-dimensional coordinate points {m} of the center of the sphere of the marker points. i }; Step S2: Acquire test videos of the subject walking, and calculate the set of three-dimensional coordinates of the center of the marker points and the set of coordinates of the center of the marker points in each frame of the test video, based on the algorithm for obtaining the rigid body motion transformation matrix. i The transformation matrix {A} i }; Step S3, calculate the set of sphere center coordinates {d} for each frame marker point. i The transformation matrix between every two frames {B i }; Step S4: Divide the gait cycle according to any marker point, use the transformation matrix to recover all frames of the current gait cycle, obtain the gait trajectory data of each marker point on the shoe in the current gait cycle, and use this to find the lowest point of the foot of each marker point in the current gait cycle. Repeat the gait cycle to obtain the lowest point of the foot MFC of all gait cycles. Step S5: Determine whether the subject belongs to the tripping population based on the lowest point of the foot (MFC) of all gait cycles, and mark the subject's gait as a healthy gait or a tripping gait based on the determination result.

2. The method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground according to claim 1, characterized in that: in, In step S1, the point cloud information is obtained by scanning the 3D model of the shoe obtained by modeling using 3D image generation and editing software, segmenting the shoe and all marked points, and reading the point cloud information of the 3D model of the shoe using programming software.

3. The method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground according to claim 1, characterized in that: in, In step S1, the set of three-dimensional coordinate points of the sphere center is calculated using the minimum bounding sphere algorithm.

4. The method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground according to claim 1, characterized in that: in, The expression for the rigid body motion transformation matrix is ​​as follows: d i =Rm i +T+V i (1) In the formula, {m i } and {d i Let {V} be two corresponding sets of marker points, i = 1...N, such that they are related by the above equation, R is a standard 3×3 rotation matrix, T is a 3-D translation vector, and {V} i } is the noise vector. Solve for the set {m i }Mapped to {d i The optimal transformation is minimized using the least squares error criterion: As the result of the least squares solution to Formula 2, the point set {d} i } and {m i The centroids should be the same. Using this constraint, a new equation can be generated by definition: In the formula, and They are the point sets {d i } and {m i The center of mass of} Rewrite and simplify Formula 2 as follows: When the last term is maximized, the equation is minimized, that is, the trace is maximized. Where H is the correlation matrix defined as follows: If the singular value decomposition of H is H=U∧V T Given the optimal rotation matrix that maximizes the desired trace, then... for: The above formula is The expression when the determinant is +1. when The determinant is -1, meaning that when the point set is planar or contains a lot of noise, the required rotation is: In the formula, matrix V′=[v1, v2, -v3] consists of columns of V, where v3 is the column corresponding to the singular value of H being zero. The optimal representation of the final rotation matrix is: The set to be collected soon {d i The centroid of} and the set {m i Aligning the centroids of rotation is the optimal translation:

5. The method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground according to claim 1, characterized in that: in, The gait cycle is defined as the movement of the same foot from heel strike to heel strike again. In step S4, the process of recovering all frames of the current gait cycle using a transformation matrix is ​​as follows: Each time the first frame of the current gait cycle is retrieved, it is in {A} i The transformation matrix in {B} recovers the shoe point cloud in the 3D modeling coordinate system in the first frame of the gait laboratory coordinate system, using {B}. i The first transformation matrix in the first frame of the restored state laboratory coordinate system represents the shoe point cloud in the second frame of the gait laboratory coordinate system. This process is repeated iteratively, taking {B}... i The transformation matrix in the gait laboratory coordinate system is used to recover the shoe point cloud of the previous frame in the gait laboratory coordinate system and the shoe point cloud of the next frame in the gait laboratory coordinate system.

6. The method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground according to claim 1, characterized in that: in, In step S5, a regression-support vector machine (RSM) system is used to detect whether the subject belongs to the tripping-prone population. The input of the RSM system is the original MFC. i The signal, where i represents the number of consecutive gait cycles, is detected based on the following principle: The object MFC is simulated using a linear prediction model estimated by an autoregressive process. i Signal variability, model coefficient capture MFC i The signal variability features are identified and labeled to represent healthy gait or stumbling gait. These are then used as input features to train a support vector machine classifier so that it can identify the coefficients of the autoregressive model belonging to healthy gait or stumbling gait.

7. The method for identifying human tripping risk based on the detection of the minimum distance between the foot and the ground according to claim 1, characterized in that: in, In step S4, the transformation matrix is ​​used to recover all frames of the current gait cycle in the following way: Extract the transformation matrix {A} of all frames in the current gait cycle. i This allows us to recover the point cloud of the shoe in the 3D modeling coordinate system across all frames in the gait laboratory coordinate system.

Images