A human walking posture motion capture method

Abstract

The application discloses a human walking posture action capture method and belongs to the field of pedestrian positioning. The method establishes an analysis model by using a simplified human model, a coordinate system defined based on the simplified human model and defined gait parameters, and uses two corrected sensors to collect data of the analysis model to obtain a motion posture. An improved Kalman filtering algorithm is used to solve the motion posture of the foot and calculate the motion posture of the lower limbs to obtain a human walking posture result. The sensor arrangement and calibration method are simple, the number of sensors used is small, and the Kalman filtering technology of zero speed update and foot landing correction effectively eliminates the cumulative error problem that cannot be solved by the Kalman filtering.

Description

Technical Field

[0001] This invention relates to the field of pedestrian positioning, and in particular to a method for capturing human walking postures and movements. Background Technology

[0002] Currently, mainstream motion capture technologies mainly include infrared photography and wireless sensor measurement. Infrared photography uses high-speed cameras to record human motion images, combined with computer vision recognition technology. By capturing the movement of marker points on the body surface, it first estimates the displacement of human joints, then further estimates the displacement and tilt angle of limbs, and obtains motion parameters such as acceleration and angular acceleration through differentiation. Wireless sensor measurement technology mainly uses wearable accelerometers, gyroscopes, and magnetometers to observe acceleration, angular velocity, and angle, and then uses integration algorithms to inversely calculate displacement information. Existing wireless sensor measurement technology is widely used in human factors engineering or industrial production fields. It mainly uses independent sensors worn on the waist, feet, or upper limbs to achieve local motion capture, posture pattern recognition, or overall positioning in indoor spaces. This technology is being extended to the motion capture of various parts and joints of the lower limbs during walking.

[0003] However, the current infrared photography equipment is expensive and the experimental environment is demanding. The experimental operation and data processing are complicated, making it difficult to carry out field observations in normal production and living environments. Wireless sensor measurement has the following problems: (1) There are limited considerations for the sources of error and limited means of error reduction. Although Kalman filtering or improved Kalman filtering based on deep learning algorithms has been adopted, it has eliminated the drift of sensor acquisition data, noise in data transmission and cumulative error in data processing to a certain extent. However, no sensor placement scheme for the torso and lower limbs during walking has been proposed to reduce the error caused by the sensor placement position. Secondly, the difference in body posture between each step cycle during steady-state walking is small. It is a narrow-band random process with relatively fixed distribution characteristics. The practice of predicting and correcting the motion posture through improved Kalman filtering technology works well in multi-mode and unstable motion, but it is not entirely applicable to single-mode stable motion such as walking. (2) Most existing wireless observation technology experimental schemes are proposed for single observation objects, but they do not regard the human body as a multi-body system composed of upper limbs, trunk and lower limbs. For multi-body coupled motion observation, multi-part posture capture is proposed, but the problem of difficulty in synchronizing the data collected by sensors of different parts is not solved. (3) In order to obtain satisfactory prediction accuracy, the technology based on the combination of artificial intelligence algorithms and wireless observation often requires a large number of training samples and high computing power, and it takes a long time. This is not conducive to realizing the problem of simple observation of a large number of experimental groups in civil engineering applications. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method for capturing human walking posture motion.

[0005] To achieve the above technology, the specific steps are as follows: S1. Establish an analysis model and obtain motion posture by collecting data from the analysis model using calibrated sensors. The analysis model includes: a simplified human body model, a coordinate system defined based on the simplified human body model, and gait parameters. The steps for establishing the analysis model are as follows: S1.1. A seven-rigid-body model is obtained by simplifying the human body model; The simplification methods include: (1) When the human body walks, the center of mass is concentrated at the hip joint. The center of mass of the human body during walking is set at the lumbar spine at the midpoint of the line connecting the left and right hip joints; (2) Because the head, upper limbs and trunk move synchronously during walking, the influence of the local movement of the head and upper limbs on the overall center of mass movement and the swing state of the lower limbs is ignored. The head, upper limbs and trunk are combined and regarded as a rigid body HAT; (3) The lower limbs are divided into six rigid bodies: the left and right thighs, the left and right calves and the left and right feet. The center of mass of each rigid body remains unchanged during walking by ignoring the deformation of muscles and the bending of the metatarsophalangeal joints of the feet; (4) The hip joint, knee joint and ankle joint are simplified into hinges, connecting seven rigid bodies. Only the relative rotation of the rigid bodies on both sides of the joint hinge is observed. By simplifying the human body model, the rigid bodies in the seven-rigid-body model are: HAT, left and right thighs, left and right calves, and left and right feet. Thus, the two ends of any rigid body are defined as follows: the end closer to the hip joint is called the proximal end, and the end farther from the hip joint is called the distal end. The posture calculation process requires the use of the dimensional parameters of the human body model, including length and center of mass position, which can be measured for each experimental individual using anthropometry methods.

[0006] S1.2. A coordinate system defined based on a simplified human body model, comprising a first coordinate system, a second coordinate system, and a third coordinate system; the first coordinate system is the Global Coordinate System (GCS), where the origin of the GCS is defined as a foot placement point of the right foot during the standard straight-line walking experiment. G The axis points in the direction of travel, y G The axis points upwards along the normal direction of the ground, z G The first coordinate system points horizontally towards the right side of the body; the second coordinate system is the body coordinate system (BCS), with the origin of the BCS as the center of mass of the rigid body, x... B The axis is the direction of the rigid body axis, pointing from the proximal joint to the distal joint, y B The axis is the direction of the rigid body's normal, in the global coordinate system x G OyG In the plane, pointing towards the front of the body, z B The first axis points horizontally towards the right side of the human body; the third coordinate system is the sensor local coordinate system (SCS), with the coordinate axes based on the measurement direction defined by the sensor measurement unit. The sensor coordinate system integrating a three-axis measurement unit is typically a Cartesian coordinate system that satisfies the right-hand screw theorem, with the three axes being x, y, and g. S y S and z S ; S1.3 Define gait parameters to obtain step frequency and stride length; Gait parameters include the support phase, swing phase, phase difference, bipedal support phase, and unipedal support phase. Gait parameters are the foundation of lower limb posture recognition. The standard gait phase of human walking, using one lower limb as the subject, divides the complete gait phase into two parts: the support phase and the swing phase. The support phase, taking the right foot as an example, is from the moment the right heel lands until the moment the right toe leaves the ground, during which the right foot remains in contact with the ground; the swing phase, also taking the right foot as an example, is from the moment the right toe leaves the ground until the right heel lands again, during which the right foot is in a suspended swinging state; the left... The left and right foot gait phases are considered to have a fixed phase difference, alternating with the right foot. If both feet are considered, the same gait phase can be divided into two parts: the phase where both feet are simultaneously in the support phase is called the two-foot support phase, and the phase where only one lower limb is in the support phase is called the one-foot support phase. A complete gait consists of alternating two-foot and one-foot support phases. The gait period is defined as the sum of the durations of the two-foot and one-foot support phases, and the stride frequency is the reciprocal of the period. The stride length is defined as the straight-line distance between the two feet during the two-foot support phase. The expression for stride length is as follows: In the formula, l ob The length of the observation area; N stp The number of steps is calculated using the following formula: In the formula, N f The number of landing points within the observation area; The formula for calculating step frequency is: ; The steps for obtaining the motion posture by acquiring data from the analysis model using calibrated sensors are as follows: S1.4 Select and arrange the sensors; the sensor arrangement should meet the following conditions, including: 1) Based on the seven-rigid-body model, the sensor positions are set as follows: HAT (Hip-Thigh) area, left and right thighs, left and right calves, and left and right feet; among them, the left and right thighs, left and right calves, and left and right feet are all set on one side; the HAT area sensor is fixed at the lumbar spine; the thigh sensor is placed on the outer thigh, the calf sensor is placed on the plane of the anterior edge of the tibia, and the foot sensor is placed on the upper side of the metatarsal bones of the foot; 2) To reduce observation errors caused by relative movement between the sensor and the body parts, the sensor is fixed with straps or fixing ropes and in contact with the body surface, avoiding thick clothing; 3) The fixing position should be on the skeleton, avoiding muscles, to avoid observation errors caused by muscle deformation; S1.5. Calibrate according to the location of the sensor; The sensors that need calibration are located in the thigh and calf areas; the SCS (Self-Calibrated Sensors) in the thigh and calf areas only need to be calibrated around the y-axis. S Rotating the axis converts it to BCS; the expression is as follows: In the formula, This represents the measured data under SCS; This represents the measured data under BCS. The rotation matrix is ​​expressed as follows: In the formula, , and The process of transforming SCS to BCS involves sequentially rotating the x-axis around the sensor's local coordinate system. S y S and z S The angle of rotation of the axis; S1.6 Arrange the calibrated sensors and obtain the acceleration value through the arranged sensors; The acceleration values ​​include: measured acceleration and measured angular velocity values ​​in the foot coordinate system; Tests revealed that using two sensors simultaneously resulted in good data synchronization, high data quality, and a sufficient sampling frequency. The arrangement scheme for the calibrated sensors is shown in Table 2; Table 2: Sensor Layout Scheme The deployed sensors were subjected to three sets of experiments. Each set of experiments included three walking observations, with the only variable being the sensor deployment scheme, while other experimental settings remained unchanged. S1.7 Establish the discriminant between the support phase and the swing phase in the lower limb gait parameters using acceleration values; According to the definition of the lower limb support phase, the foot is in contact with the ground during this phase. Compared with the swing phase, the foot motion is close to stillness, which means that the foot acceleration count is close to the gravitational acceleration, and the angular velocity measured by the gyroscope is close to zero. Therefore, the following discriminant can be established: In the formula, Indicates the supporting phase; Indicates the oscillation phase; These are the measured acceleration values ​​in the foot coordinate system. , and These are the x-axis of the foot. B y B and z B Acceleration components in the direction; This is the measured value of angular velocity. , and These are the coordinate axes around the foot, x and x respectively. B y B and z B Angular velocity component in the direction of rotation; Represents modulo operation; This indicates the absolute value operation; g is the value of gravitational acceleration; and These are the acceleration and angular velocity thresholds, respectively.

[0007] S1.8. The motion posture is obtained by aligning the time axis through the lower limb support phase and swing phase and constructing full-body synchronous data; The time axis alignment of the lower limb support and swing phases is performed as follows: Using a two-sensor (i.e., sensor arrangement) experimental scheme, three tests are conducted sequentially in one set of experiments. To obtain synchronized data for all parts of the body, the time axes of the three tests need to be aligned. Specifically, using the time axis of one foot as a reference frame, the endpoints of the support and swing phases of each part are aligned, and intermediate data are interpolated linearly or using spline interpolation. The construction of synchronized data for the whole body is based on normal gait characteristics, indicating that the movements of the left and right lower limbs have the same period, and the only difference between them is a phase delay. Therefore, the synchronization data of the left lower limb can be shifted based on the data of the right lower limb. The result can be obtained using the following calculation expression: In the formula, The sampling frequency; Sampling time interval; superscript L Indicates the left side, superscript LF The left foot. LL The left calf. LT Left thigh; t For a specific moment; express t Measured data of the left lower limb at any given time. express Real-time measured data of the right lower limb. The result for the right foot support phase; The result is for the left foot support phase.

[0008] S2. Solve for the motion posture; The spatial motion trajectories of each part can be solved sequentially from the distal to the proximal part based on the two branches of the human body structure. First, calculate the spatial position of the foot's center of mass. Then, calculate the ankle joint position based on the foot's tilt angle. Next, calculate the knee joint position based on the lower leg tilt angle. Finally, calculate the hip joint position based on the thigh tilt angle. The angular acceleration of each part's rotation can be obtained by numerically differentiating the angular velocity observation results. The solution steps are as follows: S2.1. The improved Kalman filter algorithm is used to solve the motion posture of the foot; The spatial displacement of the foot's center of mass is obtained by integrating the acceleration. However, due to observation and data processing errors, the integration result suffers from unacceptable data drift due to the cumulative effect of errors. To address this issue, this invention proposes an improved Kalman filter-based foot displacement calculation algorithm. This algorithm utilizes the characteristic that the foot velocity in the support phase is approximately zero and the prior condition that the stride length is known. Based on the Unscented Kalman Filter (UKF), it corrects the velocity and foot landing point in each stride support phase, promptly eliminating the cumulative errors generated in the oscillating phase. The steps are as follows: S2.1.1 Establish an improved Kalman filter algorithm for calculating foot displacement; definition t Time-of-flight foot motion status measurement The expression is as follows: In the formula, ,in ; p t Represents displacement in the global coordinate system (GCS); v t Represents the velocity in the global coordinate system (GCS); a tAcceleration in the global coordinate system (GCS); acceleration observation bias b A,t The inclination angle of the foot coordinates (BCS) relative to the GCS is given by quaternions. q t express, e t and q 4,t These are the vector and scalar parts of the quaternion, respectively; superscript T Indicates transpose; q 1,t , q 2,t , q 3,t These are the coefficients of the first, second, and third imaginary terms of the quaternion, respectively. The basic idea of ​​Kalman filtering is... t Initial value of the state variable at time step Obtained through the motion state theory model t Prior estimate at time +1 And obtained through observation model t Prior estimate of the measured value at time +1 By combining residual analysis between the estimated values ​​and the actual experimental values, the prior estimates are updated to obtain the final result. t Posterior estimation at time The algorithm design includes: theoretical model and observation model.

[0009] The theoretical model of this invention is established through the basic kinematic state equations, and its acceleration state equation in the global coordinate system (GCS) is as follows: In the formula, and They are respectively t +1 and t Acceleration in the global coordinate system (GCS) at any given time; The acceleration model error is represented by random Gaussian noise with zero expectation. The expression for acceleration observation bias is: In the formula, b A,t+1 and b A,t They are respectively t +1 and t Acceleration observation bias at any given time; The acceleration observation bias model error follows a normal distribution with zero expectation. Velocity in Global Coordinate System (GCS) v t In this process, a correction factor is introduced based on the acceleration integral. The expression is: In the formula, This refers to the time step or sampling time interval; This is the speed correction amount; The weighting factor for the two velocities is 0 for the support phase and 0 for the oscillating phase. ; The model error for velocity follows a normal distribution with zero expectation. Correction amount The speed is corrected using the Zero-velocity Update (ZUPT) method, as shown in the following expression: This expression is based on the support phase discrimination, that is, when IND t When =1 is the support phase, the velocity state variable is corrected using zero velocity. IND t When =0 is the swing phase, use the average velocity of the previous step swing phase. Make corrections, subscripts k This represents the number of steps, updated with the corrected speed. ; Displacement in Global Coordinate System (GCS) p t The calculation is as follows: In the formula, p t+1 and p t They are respectively t +1 and t Displacement at any given moment; The displacement error also follows a normal distribution with a zero expectation. The quaternion combined with the observed angular velocity is calculated using the following formula, as shown in the expression below: ; ; In the formula, These are the observed angular velocity values. It is a skew-symmetric matrix constructed from angular velocities. ] is the cross product operator. The model error for quaternions; The theoretical model can be represented by state equations as follows: ; ; ; ; In the formula, E 0 represents the identity matrix, and 0 represents the zero matrix.

[0010] The observation model is based on the geomagnetic field and gravitational acceleration. Combined with the results of the theoretical model, the measurement results of the magnetometer and accelerometer are predicted.

[0011] The magnetometer observations can be predicted using the following formula: In the formula, It is the rotation matrix from GCS to SCS, by t Theoretical model prediction at time +1 express, ; This is the predicted value from the magnetometer; The magnetic field of GCS is reversed; The observation noise of the magnetic field follows a normal distribution with a mean of zero; The accelerometer observations consist of gravitational acceleration, motion acceleration, and biases caused by experimental setup. Its predictive formula is: In the formula, The predicted value from the accelerometer. For gravitational acceleration under GCS, The acceleration observation noise follows a normal distribution with zero expectation; The geomagnetic direction reflects the global coordinate system orientation and is relatively fixed, while the magnetometer data reflects the sensor's coordinate system orientation, which changes with motion. Accelerometer observations include both gravitational acceleration and body motion acceleration; these two values ​​are only equal when the sensor is stationary. Combining the two equations above, we obtain the following expression for the observation model: ; In the formula, For theoretical models t Prediction results of state variables at time +1 This represents the mapping relationship of the observation model. This represents the observation noise vector.

[0012] The steps of the Unscented Kalman Filter (UKF) algorithm are as follows: 1) t Optimal estimate of time As an initial value, 1) Construct a set of 2n+1 Sigma sample points in the vicinity; 2) Substitute the constructed sample points into the state equation of the theoretical model to calculate... t The theoretical predicted value and its covariance matrix at time +1; 3) Substitute the calculation results from 2) into the observation model, and... t 4) Calculate the Kalman coefficients based on the observations at time +1. 5) Calculation t Optimal estimate at time +1 6) Update the covariance matrix; 7) ... and Substitute the initial value into step 1) to perform the calculation for the next time step, and repeat steps 1) to 6) until the last time point.

[0013] S2.1.2, Correct the landing point; To further reduce the cumulative error generated during displacement calculation, the foot displacement is corrected at the end of each step (i.e., the end of the swing phase and the beginning of the support phase), and a correction is defined as follows: t The displacement observation value of the foot landing point at any given moment is the sum of the position at the beginning of the previous step swing phase and the displacement experienced by the foot at the end of the swing phase (i.e., the step length), which can be expressed as follows: In the formula, That is, the first k -1 Foot position at the start of the stepping swing phase; The step length is calculated based on the distance of the observation area and the number of steps. For the first k -1 The horizontal movement direction of the stepping foot and the global coordinate x G The angle between the axes; since the experimental setup is along the x-axis. G The linear movement along the axis is taken as follows. .

[0014] The above definition t The displacement observation value of the landing point at any given time is compared with the optimal prediction value of the improved Kalman filter algorithm, as shown in the following formula. If the deviation between the two is large (exceeding the threshold) If the predicted value is not found, the actual measured value is used instead, and the actual measured step length is used to calculate the value. k The average speed of -1 step is expressed as follows: In the formula, Indicates the measured value of the foot position. Alternative model predictions are used to eliminate accumulated errors exceeding a threshold. For the first k -1 The average speed of the horizontal movement of the foot during the stepping and swinging phase. Step size, For the first k -1 The horizontal movement direction of the stepping foot and the global coordinate x G The included angle of the axis, That is, the first k -1 is the vector representation of the horizontal distance of the step. For the first k -1 When stepping and swinging in a stalemate.

[0015] Otherwise, if the prediction accuracy is considered acceptable, the average velocity is calculated using the prediction step size, as shown in the following expression: In the formula, the first k -1 Average speed of foot movement within the stepping and swinging phase Similarly, dividing the distance of movement by the swing duration... The calculation is performed, where the distance traveled is determined by the optimal estimate of the foot position at the end of the swing phase. Subtract the foot position at the start of the swing phase calculate.

[0016] This leads to the optimal estimate of the foot position. With average speed The state variables are updated by substituting them into the unscented Kalman filter algorithm and using the zero-rate update method.

[0017] S2.2 Calculate the lower limb movement posture; The basic attitude quantities used for human kinematics and dynamics calculations include linear displacement, linear acceleration, tilt angle, and angular acceleration. Among these, linear acceleration and tilt angle are direct observations from experiments, while displacement and angular acceleration need to be further calculated from the direct observations.

[0018] The calculation steps are as follows: S2.2.1, The displacement is calculated using the following expression: In the formula, p Indicates the location of the center of mass and joints of the lower limbs, indicated by superscript. RF Indicates the right foot. RA Indicates the right ankle joint. K Indicates the knee joint. H Indicates the hip joint, that is This is the location of the center of gravity of the right foot. This refers to the right ankle joint. This refers to the position of the right knee joint. Position of the hip joint; r The axial vectors representing each part are all unit vectors, where This is the direction vector from the center of mass of the right foot to the right ankle joint, i.e., the axial vector of the right foot. This is the directional vector from the right ankle joint to the right knee joint, i.e., the axial vector of the right lower leg. This is the directional vector from the right knee joint to the hip joint, i.e., the axial vector of the right thigh. The unit vector representing the normal direction of each body part, i.e., the body coordinates of each body part. The unit vector of the axis, where The normal vector of the right foot. Let be the normal vector of the right lower leg. Here is the normal vector of the right thigh; the axial vectors of each part can be obtained by rotating the normal vector 90 degrees counterclockwise, that is, multiplying the normal vector by a rotation matrix. , This is the distance from the center of mass of the right foot to the ankle joint; This is the distance from the ankle joint to the knee joint on the right lower leg. This is the distance from the ankle joint to the knee joint on the right thigh. The calculation process for the left lower limb is the same as that for the right.

[0019] S2.2.2, Linear Acceleration a This is the acceleration observation value after Kalman filtering correction, expressed as follows: ; S2.2.3, The inclination angle of each part is defined as the normal direction. With x G The included angle is calculated as follows: In the formula, and These represent the ordinate and abscissa values ​​of the unit normal vector, respectively. S2.2.4 Calculation of angular acceleration; Angular acceleration can be obtained from the first difference of the observed angular velocity, and the calculation formula is: ; In the formula, and Adjacent t and t Angular velocity observation at time -1 This represents the sensor sampling time interval.

[0020] S2.2.5 Obtain the attitude calculation results.

[0021] The beneficial effects of this invention are: This invention combines a multi-rigid-body human body model, a multi-sensor wireless observation method, and Kalman filtering technology based on zero-speed update and foot placement correction to capture the motion of various parts of the human body in a walking posture, obtain human kinematic parameters, and accurately reflect the human kinematic laws of the walking posture through wireless observation methods, including the time history curves of displacement, tilt angle, acceleration and angular acceleration of various parts. The zero-velocity criterion proposed in this invention, based on the walking posture characteristics of the foot, can not only serve as a reference time axis for multi-sensor data synchronization, but also improve Kalman filtering by zero-velocity updates. It effectively eliminates the cumulative error problem that Kalman filtering is difficult to solve, without increasing the amount of computation. Compared with Kalman filtering technology based on artificial intelligence, it saves more time and lowers the technical threshold. The sensor placement and calibration method proposed in this invention is simple, the experiment does not require a strict laboratory environment, and can be carried out in various practical scenarios. The equipment used is only a small number of wearable, low-cost sensors and a PC with ordinary configuration. The experimental process is quick and the training cost for testers is low.

[0022] The posture information obtained from the experiments in the embodiments of the present invention can reflect the dynamic laws of the human body as a system and can serve as the basis for calculating joint forces and human-induced loads. Attached Figure Description

[0023] Figure 1 This is a flowchart of the present invention; Figure 2 This is a schematic diagram of a seven-rigid-body model; Figure 3 A diagram showing the length parameters of a human body model; Figure 4 This is a schematic diagram of the gait phases; Figure 5 This diagram shows the coordinate system, arrangement, and calibration of the sensors in this embodiment. Part a is a schematic diagram of the sensor size; part b is a schematic diagram of the sensor in the HAT region; part c is a schematic diagram of the sensor in the thigh region; part d is a schematic diagram of the sensor in the calf region; part e is a schematic diagram of the sensor in the calf region; part f is a calibration diagram of the sensor in the thigh region; and part g is a calibration diagram of the sensor in the calf region. Figure 6 A schematic diagram is provided for the experiment. Figure 7 The diagram shows the results of the foot support phase and the swing phase discrimination. Figure 8 This is a graph showing the acceleration synchronization data results between the right lower limb and HAT. Figure 9 This is a graph showing the acceleration synchronization data results for the left lower limb; Figure 10 A diagram illustrating the foot movement calculation process; Figure 11 Framework diagram for calculating lower limb movement states; Figure 12 The trajectory and direction diagram of the right foot's center of mass; Figure 13 This is a diagram showing the movement trajectory of the lower limbs. Figure 14 The diagram shows the tilt angles of various parts of the lower limbs. Figure 15 The graph shows the angular acceleration results for different parts of the lower limbs. Detailed Implementation

[0024] The present invention will be further described in detail below with reference to specific embodiments.

[0025] like Figure 1 As shown, a method for capturing human walking posture includes the following steps: S1. Establish an analysis model and obtain motion posture by collecting data from the analysis model through calibrated sensors; wherein, the analysis model includes: a simplified human body model, a coordinate system defined based on the simplified human body model, and gait parameters defined; The steps to establish the analysis model are as follows: S1.1. A seven-rigid-body model is obtained by simplifying the human body model; The simplification methods include: (1) When the human body walks, the center of mass is concentrated at the hip joint. The center of mass of the human body during walking is set at the lumbar spine at the midpoint of the line connecting the left and right hip joints; (2) Since the head, upper limbs and trunk move synchronously during walking, the influence of the local movement of the head and upper limbs on the overall center of mass movement and the swing state of the lower limbs is ignored. The head, upper limbs and trunk are combined and regarded as a rigid body (Head-Arms-Trunk, HAT); (3) The lower limbs are divided into six rigid bodies: the left and right thighs, the left and right calves and the left and right feet. The center of mass of each rigid body remains unchanged during walking by ignoring the deformation of muscles and the bending of the metatarsophalangeal joints of the feet; (4) The hip joint, knee joint and ankle joint are simplified as hinges, connecting seven rigid bodies. Only the relative rotation of the rigid bodies on both sides of the joint hinge is observed. like Figure 2As shown, by simplifying the human body model, the rigid bodies in the seven-rigid-body model are: HAT, left and right thighs, left and right calves, and left and right feet. Thus, the two ends of any rigid body are defined as follows: the end closer to the hip joint is called the proximal end, and the end farther from the hip joint is called the distal end. like Figure 3 As shown, the posture calculation process requires the use of the dimensional parameters of the human body model, including length and center of mass position. These parameters can be measured for each experimental individual using anthropometry methods. If the measurement conditions are insufficient, they can also be estimated based on existing statistical results and proportions of human height. For Asian human bodies, the following proportions are recommended: thigh length is 0.245 times the total height, calf length is 0.246 times the total height, and foot length is 0.152 times the total height. The center of mass positions of each rigid body are recommended to be calculated according to the proportions shown in Table 1. Table 1. Relative positions of the rigid body centers of mass of the human body model In the table, l prox This is the distance from the center of mass of the rigid body to the proximal joint; l dist The distance from the center of mass of the rigid body to the distal joint; l sec The total length of the joint is the distance from the proximal joint to the distal joint.

[0026] S1.2. A coordinate system defined based on a simplified human body model, comprising a first coordinate system, a second coordinate system, and a third coordinate system; the first coordinate system is the Global Coordinate System (GCS), where the origin of the GCS is defined as a foot placement point of the right foot during the standard straight-line walking experiment. G The axis points in the direction of travel, y G The axis points upwards along the normal direction of the ground, z G The first coordinate system points horizontally towards the right side of the body; the second coordinate system is the body coordinate system (BCS), with the origin of the BCS as the center of mass of the rigid body, x... B The axis is the direction of the rigid body axis, pointing from the proximal joint to the distal joint, y B The axis is the direction of the rigid body's normal, in the global coordinate system x G Oy G In the plane, pointing towards the front of the body, z B The horizontal axis points towards the right side of the human body; the third coordinate system is the sensor local coordinate system (SCS), with the coordinate axes based on the measurement direction defined by the sensor measurement unit. The sensor coordinate system integrating a three-axis measurement unit is typically a Cartesian coordinate system that satisfies the right-hand screw theorem, for example... Figure 5 The coordinate system shown in (a) has three axes: x, y, and g. S y S (perpendicular to x in the figure) S Oz S (outward from the plane) and z S ; S1.3, such as Figure 4 As shown, gait parameters are defined to obtain step frequency and step length; Gait parameters include the support phase, swing phase, phase difference, bipedal support phase, and unipedal support phase. Gait parameters are the foundation of lower limb posture recognition. The standard gait phase of human walking, taking one lower limb as the object, divides the complete gait phase into two parts: the support phase and the swing phase. The support phase, taking the right foot as an example, is from the time the right heel lands to the time the right toe leaves the ground, during which the right foot remains in contact with the ground. The swing phase, taking the right foot as an example, is from the time the right toe leaves the ground to the time the right heel lands again, during which the right foot is in a suspended swinging state. The left foot has the same gait phase, alternating with the right foot, and is considered to have a fixed phase difference between the left and right foot gait phases. If both feet are taken as the object of discussion, the same gait phase can be divided into two parts: the gait phase in which both feet are in the support phase is called the bipedal support phase, and the phase in which only one lower limb is in the support phase is called the unipedal support phase. A complete gait consists of alternating bipedal and unipedal support phases. The gait period is defined as the sum of the durations of the bipedal and unipedal support phases, and the stride frequency is the reciprocal of the period. The stride length is defined as the straight-line distance between the two feet during the bipedal support phase. The expression for the stride length is as follows: In the formula, l ob The length of the observation area; N stp The number of steps is calculated using the following formula: In the formula, N f The number of landing points within the observation area; The formula for calculating step frequency is: ; The steps for obtaining the motion posture by acquiring data from the analysis model using calibrated sensors are as follows: S1.4 Select and arrange the sensors; This embodiment employs a nine-axis sensor, which integrates three measurement units: a three-axis accelerometer, a three-axis magnetometer, and a three-axis gyroscope; used to acquire acceleration, angle, and angular velocity signals, respectively; the acceleration range is [-16g0, 16g0], with a measurement accuracy of 0.01g0 (g0 represents gravitational acceleration); the angle range is [-180, 180], with an accuracy of 0.01; and the angular velocity range is [-2000 / s, 2000 / s], with an accuracy of 0.05 / s; the axis of the measurement unit is the SCS direction, such as... Figure 5 As shown in sections b, c, d, and e; the sensor supports a sampling frequency of up to 200Hz, and the collected data is transmitted to a PC for storage and processing via Bluetooth 5.0. The sensor placement must meet the following conditions: 1) Based on the seven-rigid-body model, the sensor positions are set as follows: HAT (Hip-Thigh) area, left and right thighs, left and right calves, and left and right feet; among them, the left and right thighs, left and right calves, and left and right feet are all set on one side; the HAT area sensor is fixed at the lumbar spine; the thigh sensor is placed on the outer thigh, the calf sensor is placed on the plane of the anterior edge of the tibia, and the foot sensor is placed on the upper side of the metatarsal bones of the foot; 2) To reduce observation errors caused by relative movement between the sensor and the body parts, the sensor is fixed with straps or fixing ropes and in contact with the body surface, avoiding thick clothing; 3) The fixing position should be on the skeleton, avoiding muscles, to avoid observation errors caused by muscle deformation; S1.5. Calibrate according to the location of the sensor; like Figure 5 As shown in sections f and g, the sensors that need to be calibrated are those in the thigh and calf regions; the SCS in the thigh and calf regions only needs to be calibrated around y. S Rotating the axis converts it to BCS; the expression is as follows: In the formula, This represents the measured data under SCS; This represents the measured data under BCS. The rotation matrix is ​​expressed as follows: In the formula, , and The process of transforming SCS to BCS involves sequentially rotating the x-axis around the sensor's local coordinate system. S y S and z S The angle of axis rotation; based on the sensor placement, the SCS at the thigh and calf only needs to rotate around y. S Rotating the axis converts it to BCS, thus allowing the above equation to... and It is zero; This requires a calibration experiment. The calibration method is as follows: Have the subject stand upright at rest. First, place the sensor in its actual position and record the direction of the magnetic field. Then, rotate the sensor to align it with the BCS direction and record the direction of the magnetic field again. The angle of rotation recorded twice is the magnetic field angle. .

[0027] S1.6 Arrange the calibrated sensors and obtain the acceleration value through the arranged sensors; The acceleration values ​​include: measured acceleration and measured angular velocity values ​​in the foot coordinate system; Tests revealed that using two sensors simultaneously resulted in good data synchronization, high data quality, and a sufficient sampling frequency. The arrangement scheme for the calibrated sensors is shown in Table 2; Table 2: Sensor Layout Scheme The deployed sensors were subjected to three sets of experiments. Each set of experiments included three walking observations, with the only variable being the sensor deployment scheme, while other experimental settings remained unchanged. Based on the height of the subject and the proportional relationships in Table 1, the human body model parameters are shown in Table 3: Table 3: Human Model Parameters of the Subjects In this embodiment, the experiment was conducted in an open, unobstructed indoor corridor. A walkway was set up using marked lines. Based on the fact that the test subject would experience three processes along the walkway from the starting point to the end point—initial acceleration, constant speed stabilization, and deceleration to a stop—the walkway was divided into three areas, such as... Figure 6 As shown, the starting and ending areas are buffer zones, and the middle area is the observation zone. The experiment mainly observes steady-state walking in the middle area. A smartphone is set up on one side of the walkway to record the walking process, and a laptop computer is set up nearby as a host computer to receive sensor data in real time via Bluetooth. The experiment requires the subject to walk from the starting point to the ending point of the walkway at a typical cadence, guided by a metronome. During this process, in addition to recording sensor information and video, a stopwatch is used to record the time taken for the subject to traverse the observation zone. T ob The observation period is based on the subject's foot placement, starting from the first step the subject takes into the observation area and continuing until the foot placement point exceeds the observation area.

[0028] S1.7 Establish the discriminant between the support phase and the swing phase in the lower limb gait parameters using acceleration values; According to the definition of the lower limb support phase, the foot is in contact with the ground during this phase. Compared with the swing phase, the foot motion is close to stillness, which means that the foot acceleration count is close to the gravitational acceleration, and the angular velocity measured by the gyroscope is close to zero. Therefore, the following discriminant can be established: In the formula, Indicates the supporting phase; Indicates the oscillation phase; These are the measured acceleration values ​​in the foot coordinate system. , and These are the x-axis of the foot. B y B and z B Acceleration components in the direction; This is the measured value of angular velocity. , and These are the coordinate axes around the foot, x and x respectively. B y B and z B Angular velocity component in the direction of rotation; Represents modulo operation; This indicates the absolute value operation; g is the value of gravitational acceleration; and These are the acceleration and angular velocity thresholds, respectively, in this embodiment. and Take 2, the result is as follows Figure 7 As shown.

[0029] S1.8. The motion posture is obtained by aligning the time axis through the lower limb support phase and swing phase and constructing full-body synchronous data. The time axis alignment of the lower limb support and swing phases is performed as follows: Using a two-sensor experimental setup, three tests are conducted sequentially in one set of experiments. To obtain synchronized data from all parts of the body, the time axes of the three tests need to be aligned. Specifically, using the time axis of one foot as a reference, the endpoints of the support and swing phases of each part are aligned, and intermediate data are interpolated linearly or using spline interpolation. The result after aligning the three test data is as follows: Figure 8 As shown, this is the synchronization of data between the right lower limb and the HAT site. Figure 8 The acceleration data for each part is displayed, with superscript numbers. B This indicates that the data in the graph is under BCS. RF Indicates the right foot. RL Indicates the right calf. RTThe figure shows the acceleration of the right thigh, derived from three observations, with the right foot's stance phase used as the basis for judgment in each observation. Using the time points as a basis, the accelerations were aligned, and the results showed that the accelerations of the lower limbs were relatively small in the support phase and relatively large in the swing phase, which is consistent with the actual gait pattern.

[0030] The whole-body synchronization data is constructed as follows: Based on normal gait characteristics, it can be seen that the movements of the left and right lower limbs have the same period, and the only difference between them is the phase delay. Therefore, the synchronization data of the left lower limb can be shifted based on the data of the right lower limb. The result can be obtained using the following calculation expression: In the formula, The sampling frequency; Sampling time interval; superscript L Indicates the left side, superscript LF The left foot. LL The left calf. LT Left thigh; t For a specific moment; express t Measured data of the left lower limb at any given time. express Real-time measured data of the right lower limb. The result for the right foot support phase; The result is for the left foot support phase.

[0031] like Figure 9 The figure shows the translation results at a step frequency of 2.0Hz, thus obtaining the synchronization data of the whole body.

[0032] S2. Solve for the motion posture; The spatial motion trajectory of each part can be solved one by one according to the two branches of the human body, from the distal part to the proximal part. First, calculate the spatial position of the foot's center of mass, then calculate the ankle joint position according to the foot's tilt angle, then calculate the knee joint position according to the lower leg tilt angle, and finally calculate the hip joint position according to the thigh tilt angle. The angular acceleration of each part's rotation can be obtained by numerical differentiation from the angular velocity observation results.

[0033] The solution steps are as follows: S2.1, such as Figure 10 As shown, an improved Kalman filter algorithm is used to solve the motion posture of the foot. The spatial displacement of the foot's center of mass is obtained by integrating the acceleration. However, due to observation and data processing errors, the integration result suffers from unacceptable data drift due to the cumulative effect of errors. To address this issue, this invention proposes an improved Kalman filter-based foot displacement calculation algorithm. This algorithm utilizes the characteristic that the foot velocity in the support phase is approximately zero and the prior condition that the stride length is known. Based on the Unscented Kalman Filter (UKF), it corrects the velocity and foot landing point in each stride support phase, promptly eliminating the cumulative errors generated in the oscillating phase. The steps are as follows: S2.1.1 Establish an improved Kalman filter algorithm for calculating foot displacement; definition t Time-of-flight foot motion status measurement The expression is as follows: In the formula, ,in ; p t Represents displacement in the global coordinate system (GCS); v t Represents the velocity in the global coordinate system (GCS); a t Acceleration in the global coordinate system (GCS); acceleration observation bias b A,t The inclination angle of the foot coordinates (BCS) relative to the GCS is given by quaternions. q t express, e t and q 4,t These are the vector and scalar parts of the quaternion, respectively; superscript T Indicates transpose; q 1,t , q 2,t , q 3,t These are the coefficients of the first, second, and third imaginary terms of the quaternion, respectively. The basic idea of ​​Kalman filtering is... t Initial value of the state variable at time step Obtained through the motion state theory model t Prior estimate at time +1 And obtained through observation model t Prior estimate of the measured value at time +1 By combining residual analysis between the estimated values ​​and the actual experimental values, the prior estimates are updated to obtain the final result.t Posterior estimation at time The algorithm design includes a theoretical model and an observational model. The theoretical model of this invention is established through basic kinematic state equations, and its acceleration state equation in the global coordinate system (GCS) is: In the formula, and They are respectively t +1 and t Acceleration in the global coordinate system (GCS) at any given time; The acceleration model error is represented by random Gaussian noise with zero expectation. The expression for acceleration observation bias is: In the formula, b A,t+1 and b A,t They are respectively t +1 and t Acceleration observation bias at any given time; The acceleration observation model error follows a normal distribution with zero expectation. Velocity in Global Coordinate System (GCS) v t In this process, a correction factor is introduced based on the acceleration integral. The expression is: In the formula, The time step or sampling time interval is defined in this embodiment, which uses a sampling frequency of 100Hz and is set to 0.005s. This is the speed correction amount; The weighting factor for the two velocities is 0 for the support phase and 0 for the oscillating phase. ; The model error for velocity follows a normal distribution with zero expectation. Correction amount The speed is corrected using the Zero-velocity Update (ZUPT) method, as shown in the following expression: This expression is based on the support phase discrimination, that is, when IND t When =1 is the support phase, the velocity state variable is corrected using zero velocity. IND t When =0 is the swing phase, use the average velocity of the previous step swing phase. Make corrections, subscripts kThis represents the number of steps, updated with the corrected speed. ; Displacement in Global Coordinate System (GCS) p t The calculation is as follows: In the formula, p t+1 and p t They are respectively t +1 and t Displacement at any given moment; The displacement error also follows a normal distribution with a zero expectation. The quaternion combined with the observed angular velocity is calculated using the following formula: In the formula, These are the observed angular velocity values. It is a skew-symmetric matrix constructed from angular velocities. ] is the cross product operator. The model error for quaternions; The theoretical model can be represented by state equations as follows: ; ; ; ; In the formula, E 0 represents the identity matrix, and 0 represents the zero matrix.

[0034] The observation model is based on the geomagnetic field and gravitational acceleration. Combined with the results of the theoretical model, the measurement results of the magnetometer and accelerometer are predicted.

[0035] The magnetometer observations can be predicted using the following formula: In the formula, It is the rotation matrix from GCS to SCS, by t Theoretical model prediction at time +1 express, ; This is the predicted value from the magnetometer; The magnetic field of GCS is reversed; The observation noise of the magnetic field follows a normal distribution with a mean of zero; The accelerometer observations consist of gravitational acceleration, motion acceleration, and biases caused by experimental setup. Its predictive formula is: In the formula, The predicted value from the accelerometer. For gravitational acceleration under GCS, The acceleration observation noise follows a normal distribution with zero expectation; the geomagnetic direction reflects the global coordinate system orientation and is relatively fixed, while the magnetometer data reflects the sensor coordinate system orientation and changes with motion. The accelerometer observations include gravitational acceleration and body motion acceleration; these two are only equal when the sensor is stationary. Combining these two equations, we obtain the following expression for the observation model: ; In the formula, For theoretical models t Prediction results of state variables at time +1 This represents the mapping relationship of the observation model. This represents the observation noise vector.

[0036] The steps of the Unscented Kalman Filter (UKF) algorithm are as follows: 1) t Optimal estimate of time As an initial value, The set of 2n+1 Sigma sample points is constructed nearby, as shown in the following expression: In the formula, Represents the Sigma point set; j Indicates the sample point number. j ∈ n ,in Click t Optimal estimation of time variables The superscript "+" sign indicates the optimal estimate of each variable; WM This represents the weighting coefficients of the mean of a point set; WC The weighting coefficients for the covariance of the point set are represented by ; other points are weighted by . Construct around the center; It is the scaling factor for the distance between the construction point and the center point; To determine the dispersion of the construction points, take... ; Let be the second-order scaling factor, and take . ; Let the prior distribution parameters of the sample points be taken as follows: ; The first root of the mean square of the matrix is ​​represented by the first root of the mean square. j The root mean square of the matrix is ​​calculated using Cholesky decomposition. express t The covariance matrix of the optimal estimate of the state variables at time step; 2) Substitute the constructed sample points into the state equations of the theoretical model to calculate... t The theoretical predicted value and its covariance matrix at time +1 are calculated as follows: In the formula, superscript j The Sigma point number is indicated by the "-" sign, which represents the prior estimate of each variable obtained through the theoretical model. for j Theoretical model predictions for the points; These are the weighted theoretical model predictions; for j The weight coefficients of the points; This represents the covariance matrix of the predicted values ​​based on the theoretical model. The covariance matrix represents the computational error; 3) Substitute the calculation results from 2) into the observation model, and... t The prediction is based on the observation at time +1, calculated as follows: In the formula, the superscript "-" also represents the prior estimate of each variable calculated through the observation model; for t +1 moment j The observed model predicts the value of the point. It is a weighted average of the model predictions for each observation point; The covariance matrix represents the predicted values ​​from the observation model; This is the covariance matrix between the predicted values ​​from the theoretical model and the predicted values ​​from the observation model. The covariance matrix of the observation noise; 4) Calculate the Kalman coefficients The expression is as follows: 5) Calculation t Optimal estimate at time +1 The expression is as follows: In the formula, the superscript "+" indicates the optimal estimate after correction, and the second term on the right side of the equation is called the Kalman gain, which is equal to the Kalman coefficient. Multiply by the sensor's measured value Compared with the predicted values ​​of the observation model The difference; 6) Update the covariance matrix using the following expression: 7) and Substitute the initial value into step 1) to perform the calculation for the next time step, and repeat steps 1) to 6) until the last time point.

[0037] S2.1.2, Correct the landing point; To further reduce the cumulative error generated during displacement calculation, the foot displacement is corrected at the end of each step (i.e., the end of the swing phase and the beginning of the support phase), and a correction is defined as follows: t The displacement observation value of the foot landing point at any given moment is the sum of the position at the beginning of the previous step swing phase and the displacement experienced by the foot at the end of the swing phase (i.e., the step length), which can be expressed as follows: In the formula, That is, the first k -1 Foot position at the start of the stepping swing phase; The step length is calculated based on the distance of the observation area and the number of steps. For the first k -1 The direction of walking with each step is the direction of foot movement relative to x. G The angle between the axes; since the experimental setup is along the x-axis. G The linear movement along the axis is taken as follows. .

[0038] The above definition t The displacement observation value of the landing point at any given time is compared with the optimal prediction value of the improved Kalman algorithm, as shown in the following formula. If the deviation between the two is large (exceeding the threshold) If the predicted value is not found, the actual measured value is used instead, and the actual measured step length is used to calculate the value. kThe average speed of -1 step is expressed as follows: In the formula, Indicates the measured value of the foot position. Alternative model predictions are used to eliminate accumulated errors exceeding a threshold. For the first k -1 The average speed of the horizontal movement of the foot during the stepping and swinging phase. Step size, For the first k -1 The horizontal movement direction of the stepping foot and the global coordinate x G The included angle of the axis, That is, the first k -1 is the vector representation of the horizontal distance of the step. For the first k -1. During the step-swing phase, if the phase is stalemate, the prediction accuracy is considered acceptable. The average velocity is calculated using the predicted step length, as shown in the following expression: In the formula, the first k -1 Average speed of foot movement within the stepping and swinging phase Similarly, the distance of movement is divided by the swing duration. The calculation is performed, where the distance traveled is determined by the optimal estimate of the foot position at the end of the swing phase. Subtract the foot position at the start of the swing phase Calculation. Then, the optimal estimate of the foot position is obtained. With average speed The state variables are updated by substituting them into the unscented Kalman filter algorithm and using the zero-rate update method.

[0039] S2.2 Calculate the lower limb movement posture; The basic attitude quantities used in human kinematics and dynamics calculations include linear displacement, linear acceleration, tilt angle, and angular acceleration. Linear acceleration and tilt angle are direct observations from experiments, while displacement and angular acceleration require further calculations based on these direct observations. The calculation steps are as follows: S2.2.1 Calculation of displacement, the expression is as follows: In the formula, pIndicates the location of the center of mass and joints of the lower limbs, indicated by superscript. RF Indicates the right foot. RA Indicates the right ankle joint. K Indicates the knee joint. H Indicates the hip joint, that is This is the location of the center of gravity of the right foot. This refers to the right ankle joint. This refers to the position of the right knee joint. Position of the hip joint; r The axial vectors representing each part are all unit vectors, where This is the direction vector from the center of mass of the right foot to the right ankle joint, i.e., the axial vector of the right foot. This is the directional vector from the right ankle joint to the right knee joint, i.e., the axial vector of the right lower leg. This is the directional vector from the right knee joint to the hip joint, i.e., the axial vector of the right thigh. The unit vector representing the normal direction of each body part, i.e., the body coordinates of each body part. The unit vector of the axis, where The normal vector of the right foot. Let be the normal vector of the right lower leg. Here is the normal vector of the right thigh; the axial vectors of each part can be obtained by rotating the normal vector 90 degrees counterclockwise, that is, multiplying the normal vector by a rotation matrix. , This is the distance from the center of mass of the right foot to the ankle joint; This is the distance from the ankle joint to the knee joint on the right lower leg; This is the distance from the ankle joint to the knee joint on the right thigh. The calculation process for the left lower limb is the same as that for the right.

[0040] like Figure 11 As shown, under the basic assumptions of the human body model, the lower limbs are treated as rigid bodies connected by joints, which are considered hinges. The parts on either side of a joint only undergo relative rotation, not relative translation. Therefore, the axis of each part can be defined as the line connecting the two joints. Once the spatial position of the joint is determined, the position of each part can be determined. The position of the foot's center of mass has been calculated. Based on this, the solution order for the lower limb joints is ankle, knee, and hip joints, and the solution framework is as follows: Figure 10 As shown, based on the perpendicular relationship between the directions of the body coordinate system axes and the axial direction, first calculate the axial vector from the distal end to the proximal end of each part, and then combine it with the part lengths given above to obtain the position of each joint. For example, the right ankle joint can be calculated by adding the direction vector from the right foot's center of mass to the right ankle using the above formula, and other joints can be calculated in the same way.

[0041] S2.2.2, Linear Accelerationa This is the acceleration observation value after Kalman filtering correction, expressed as follows: In the formula, Z A Acceleration observations corrected by Kalman filtering; S2.2.3, The inclination angle of each part is defined as the normal direction. With x G The included angle is calculated as follows: In the formula, and These represent the ordinate and abscissa values ​​of the unit normal vector, respectively. S2.2.4 Calculation of angular acceleration; Angular acceleration can be obtained from the first difference of the observed angular velocity, and the calculation formula is: ; In the formula, and Adjacent t and t Angular velocity observation at time -1 This represents the sampling time interval.

[0042] S2.2.5 Obtain the attitude calculation results; The attitude calculation process was implemented using a self-written Matlab program, and the calculation results are as follows: Figure 12-15 As shown. Figure 12 This displays the foot displacement and tilt angle during two consecutive stepping cycles, as shown by the GSC values ​​on the horizontal and vertical axes. The depicted trajectory of the center of mass and its vector representation. In the diagram, the starting point of the vector is the center of mass, and the two vectors represent the motion state over a time interval of 0.02s. The foot undergoes two stepping cycles, with the foot's center of mass experiencing up-and-down reciprocating motion. During the support phase, the center of mass approaches zero, indicating contact with the ground. The horizontal distance traveled by the foot's center of mass in each cycle is approximately equal to twice the experimentally measured step length, demonstrating that the attitude calculation method proposed in this application can effectively eliminate errors and ensure high observation accuracy. Figure 13 The diagram shows the spatial trajectory of lower limb movement. Blue and red line segments represent the axes of various parts of the right and left lower limbs, respectively. The connection points of the line segments represent joints. Two adjacent sets of line segments represent the movement state at a time interval of 0.05s. The diagram shows that the right lower limb went through two complete gait cycles starting from the support phase. The line segments are denser in the support phase, indicating that the lower limb movement is slower, while the line segments are sparser in the swing phase, indicating that the lower limb movement is faster, which is consistent with the discrimination results of the support phase.

[0043] Figure 14 The figure shows the curves illustrating the change in inclination angles of different parts of the lower limbs over time. Throughout the walking process, the inclination angles of the foot and lower leg exhibit similar trends. During the stance phase, the inclination angles of each part decrease slightly, indicating that as the body's center of gravity shifts forward, the pressure exerted by the body weight on the sole of the foot moves from the heel to the toes, resulting in a slight forward tilting tendency in each part. During the swing phase, the inclination angles of the foot and lower leg show a characteristic of initially decreasing, rapidly increasing to a peak in the middle phase, and then decreasing again at the end phase. Figure 13 The trend of leg movement posture changes is consistent with the experimental video, indicating that this application can effectively observe the kinematic information of various parts of the body.

[0044] Figure 15 This is a curve showing the change of angular acceleration over time. The graph illustrates the time history of angular acceleration. Figure 8 and Figure 9 The linear acceleration time history curves shown exhibit similar changing patterns, based on... Figure 15 The foot lift-off and landing motions can be considered as the transition between the support phase and the swing phase. During the middle of the support and swing phases, the lower limbs are in a relatively stable state of motion (e.g., the foot is nearly stationary in the support phase and nearly moves at a constant speed in the swing phase), resulting in a relatively small angular acceleration. However, during the transition between lift-off and landing, the angular acceleration at various points changes significantly as the lower limbs accelerate after pushing off and decelerate upon landing. The trends in these accelerations and angular accelerations conform to the dynamic characteristics of typical walking, indicating that the observation results of this application can be directly used in related application fields such as pedestrian load calculation, joint load calculation, and walking energy consumption calculation, demonstrating high application value.

[0045] In summary, this application proposes a motion capture method for walking posture by combining wireless observation technology and an improved Kalman filter algorithm, solving the problems faced by existing technologies. Firstly, this invention effectively reduces errors and reflects the randomness of actual walking posture. By combining a sensor placement scheme and position correction algorithm, observation errors are minimized. Furthermore, the Kalman filter algorithm is improved through zero-velocity update and foot-landing point correction algorithms, reducing the cumulative error generated during acceleration integration. Figure 12 As shown in the foot's center of mass motion, during the support phase of both consecutive steps, the foot is in the same horizontal plane, and the stride lengths are similar, which is consistent with the laws and observations of steady-state walking in actual experiments. Figure 8 and Figure 15 It can be seen that the amplitudes of acceleration and angular acceleration change between two steps, reflecting the randomness present in actual walking. Secondly, a scheme using two sensors for multiple observations is employed to achieve motion capture of a multi-rigid-body human model, based on... Figure 7 The support phase discrimination results shown demonstrate that time synchronization of sensor observation data from various locations has been achieved, such as... Figure 8 , Figure 9 , Figure 14 and Figure 15 As shown. Finally, the experimental requirements for this application are relatively low, and the experimental setup is as follows. Figure 6 As shown, the device only requires two wearable sensors, a smartphone, and a computer. Compared to infrared photography technology, this method has low experimental costs and simple preparation. In addition to the flat ground walking experiment shown in the example, it can also be applied to the capture of walking motions in everyday indoor and outdoor spaces such as stairs and ramps, without being limited by the experimental environment. Furthermore, data processing and attitude calculation can be directly calculated in seconds using a simple program written on platforms such as Matlab, eliminating the time spent on building deep learning models and sample training, which is conducive to repeating experiments in a short period of time to obtain a large amount of data.

Claims

1. A method for capturing human walking posture and motion, characterized in that, Includes the following steps: S1. Establish an analysis model and obtain the motion posture by collecting data from the analysis model through the calibrated sensors; The establishment of the analysis model includes: establishing a simplified human body model, establishing a coordinate system based on the simplified human body model, and defining gait parameters; S2. Solve the motion posture to obtain the human walking posture result; The steps in S1 are as follows: S1.

1. A seven-rigid-body model is obtained by simplifying the human body model; The simplification methods of the simplified human body model include: (1) When the human body walks, the center of mass is concentrated at the hip joint. The center of mass of the human body during walking is set at the lumbar spine at the midpoint of the line connecting the left and right hip joints; (2) Because the head, upper limbs and trunk move synchronously during walking, the influence of the local movement of the head and upper limbs on the overall center of mass movement and the swing state of the lower limbs is ignored. The head, upper limbs and trunk are combined and regarded as a rigid body HAT; (3) The lower limbs are divided into six rigid bodies: the left and right thighs, the left and right calves and the left and right feet. The center of mass of each rigid body remains unchanged during walking by ignoring the deformation of muscles and the bending of the metatarsophalangeal joints of the feet; (4) The hip joint, knee joint and ankle joint are simplified into hinges, connecting seven rigid bodies. Only the relative rotation of the rigid bodies on both sides of the joint hinge is observed. By simplifying the human body model, the rigid bodies in the seven-rigid-body model are: HAT, left and right thighs, left and right calves, and left and right feet. Thus, the two ends of any rigid body are defined as follows: the end closer to the hip joint is called the proximal end, and the end farther from the hip joint is called the distal end. S1.2, A coordinate system defined based on a simplified human body model, wherein the coordinate system includes a first coordinate system, a second coordinate system, and a third coordinate system; The first coordinate system is the Global Coordinate System (GCS). In the standard experiment of the subject walking in a straight line, the origin of the GCS is defined as a point where the right foot lands. G The axis points in the direction of travel, y G The axis points upwards along the normal direction of the ground, z G The axis points horizontally towards the right side of the body; The second coordinate system is the body part local coordinate system BCS, with the origin of BCS as the rigid body's center of mass, x B The axis is the direction of the rigid body axis, pointing from the proximal joint to the distal joint, y B The axis is the direction of the rigid body's normal, in the global coordinate system x G Oy G In the plane, pointing towards the front of the body, z B The axis points horizontally towards the right side of the body; The third coordinate system is the sensor local coordinate system (SCS), with the coordinate axes based on the measurement directions defined by the sensor measurement unit. The sensor coordinate system of an integrated three-axis measurement unit is typically a Cartesian coordinate system that satisfies the right-hand screw theorem, with the three axes being x, y, and g. S y S and z S ; S1.3 Define gait parameters to obtain step frequency and stride length; Gait parameters include the support phase, swing phase, phase difference, two-legged support phase, and one-legged support phase; The standard gait phases of human walking, taking one lower limb as the subject, divide the complete gait phase into two parts: the support phase and the swing phase. The support phase is defined as the period from the moment the heel of one foot lands to the moment the toe of the other foot leaves the ground, during which the foot remains in contact with the ground. The swing phase is defined as follows: with one foot as the target, from the moment the toe of one foot leaves the ground until the heel of one foot touches the ground again, during which the one foot is in a suspended swinging state, which is called the swing phase of one foot. The left foot alternates with the right foot in the same gait phase, which is considered to be a fixed phase difference between the left and right foot gait phases; If we take both feet as the object, it can be divided into two parts: the gait phase in which both feet are in the support phase at the same time is called the two-foot support phase, and the phase in which only one lower limb is in the support phase is called the one-foot support phase. A complete gait consists of alternating bipedal and unipedal support phases. The gait period is defined as the sum of the durations of the bipedal and unipedal support phases, and the stride frequency is the reciprocal of the period. The stride length is defined as the straight-line distance between the two feet during the bipedal support phase. The expression for the stride length is as follows: In the formula, l ob The length of the observation area; N stp The number of steps is calculated using the following formula: In the formula, N f The number of landing points within the observation area; The formula for calculating step frequency is: ； S1.4 Select and arrange the sensors; S1.

5. Calibrate according to the location of the sensors; the expression is as follows: In the formula, This represents the measured data under SCS; This represents the measured data under BCS. The rotation matrix is ​​expressed as follows: In the formula, , and The transformation from SCS to BCS requires sequentially rotating the x-axis around the sensor's local coordinate system. S y S and z S The angle of rotation of the axis; S1.6 Arrange the calibrated sensors and obtain the acceleration value through the arranged sensors; The acceleration values ​​include: measured acceleration and measured angular velocity values ​​in the foot coordinate system; S1.

7. Establish the discriminant between the support phase and the swing phase in the lower limb gait parameters using acceleration values. The expression is as follows: In the formula, Indicates the supporting phase; Indicates the oscillation phase; These are the measured acceleration values ​​in the foot coordinate system. , and These are the x-axis of the foot. B y B and z B Acceleration components in the direction; This is the measured value of angular velocity. , and These are the coordinate axes around the foot, x and x respectively. B y B and z B Angular velocity component in the direction of rotation; Represents modulo operation; This indicates the absolute value operation; g is the value of gravitational acceleration; and These are the acceleration and angular velocity thresholds, respectively. S1.

8. The motion posture is obtained by aligning the time axis through the lower limb support phase and swing phase and constructing full-body synchronous data. Time axis alignment is achieved by using the time axis of one foot as a reference system, aligning the endpoints of the support phase and swing phase of each part, and performing linear interpolation or spline interpolation on the intermediate data. The whole-body synchronization data is constructed as follows: Based on normal gait characteristics, it can be seen that the movements of the left and right lower limbs have the same period, and the only difference between them is the phase delay. Therefore, the synchronization data of the left lower limb can be shifted based on the data of the right lower limb. The result can be obtained using the following calculation expression: In the formula, The sampling frequency; Sampling time interval; superscript L Indicates the left side, superscript LF The left foot. LL The left calf. LT Left thigh; t For a specific moment; express t Measured data of the left lower limb at any given time. express Measured data of the right lower limb at any given time; The result for the right foot support phase; The result is for the left foot support phase.

2. The method for capturing human walking posture and motion according to claim 1, characterized in that, The steps for calculating the motion posture are as follows: S2.

1. The improved Kalman filter algorithm is used to calculate the foot's motion posture. The steps are as follows: S2.1.1 Establish an improved Kalman filter algorithm for calculating foot displacement; definition t Time-of-flight foot motion status measurement The expression is as follows: In the formula, ,in ; p t Represents displacement in the global coordinate system (GCS); v t Represents the velocity in the global coordinate system (GCS); a t Acceleration in the global coordinate system (GCS); acceleration observation bias b A,t The inclination angle of the foot coordinates BCS relative to GCS is given by quaternions. q t express, e t and q 4,t These are the vector and scalar parts of the quaternion, respectively; superscript T Indicates transpose; q 1,t , q 2,t , q 3,t These are the coefficients of the first, second, and third imaginary terms of the quaternion, respectively. The improved Kalman filter foot displacement calculation algorithm includes: a theoretical model and an observational model; The theoretical model is established through the basic kinematic state equations, as follows: In the formula, This represents the state vector for the next time step. This represents the state vector at the current time step. , represents the state transition matrix, where The sampling time interval, These are weighting coefficients; ,in This is the speed correction amount; , representing the error, where, The model error for quaternions is... For displacement error, For the model error of velocity, For acceleration observation model errors, Indicates the error in the acceleration model; ,in These are the observed angular velocity values. It is a skew-symmetric matrix constructed from angular velocities; E 0 represents the identity matrix, and 0 represents the zero matrix; The observation model expression is as follows: ； In the formula, For theoretical models t Prediction results of state variables at time +1 This represents the mapping relationship of the observation model; Represents the observation noise vector; It is the rotation matrix from GCS to SCS, by t Theoretical model prediction at time +1 express; The magnetic field of GCS is reversed; This is observation noise for the magnetic field; For acceleration observation noise; S2.1.2, Correct the landing point; definition t The displacement observation value of the foot landing point at any given moment is the sum of the position at the beginning of the previous step swing phase and the displacement experienced by the foot at the end of the swing phase, which can be expressed as follows: In the formula, That is, the first k -1 Foot position at the start of the stepping swing phase; The step length is calculated based on the distance of the observation area and the number of steps. For the first k -1 The horizontal movement direction of the stepping foot and the global coordinate x G The included angle of the axis; Define t The displacement observation value of the landing point at any given time is compared with the optimal prediction value of the improved Kalman filter algorithm, as shown in the following formula. If the deviation between the two exceeds the threshold Then the measured value is used instead of the predicted value, and the measured step length is used to calculate... k The average speed of -1 step is expressed as follows: In the formula, Indicates the measured value of the foot position. Alternative model predictions are used to eliminate accumulated errors exceeding a threshold. For the first k -1 The average speed of the horizontal movement of the foot during the stepping and swinging phase. Step size, For the first k -1 The horizontal movement direction of the stepping foot and the global coordinate x G The included angle of the axis, That is, the first k -1 is the vector representation of the horizontal distance of the step. For the first k -1 When stepping and swinging in a stalemate; Otherwise, if the prediction accuracy is considered acceptable, the average velocity is calculated using the prediction step size, as shown in the following expression: In the formula, the first k -1 Average speed of foot movement within the stepping and swinging phase Similarly, dividing the distance of movement by the swing duration... The calculation is performed, where the distance traveled is determined by the optimal estimate of the foot position at the end of the swing phase. Subtract the foot position at the start of the swing phase calculate; S2.2 Calculate lower limb movement posture; the calculation steps are as follows: S2.2.1, The displacement is calculated using the following expression: In the formula, p Indicates the location of the center of mass and joints of the lower limbs, indicated by superscript. RF Indicates the right foot. RA Indicates the right ankle joint. K Indicates the knee joint. H Indicates the hip joint, that is This is the location of the center of gravity of the right foot. This refers to the right ankle joint. This refers to the position of the right knee joint. Position of the hip joint; r The axial vectors representing each part are all unit vectors, where This is the direction vector from the center of mass of the right foot to the right ankle joint, i.e., the axial vector of the right foot. This is the directional vector from the right ankle joint to the right knee joint, i.e., the axial vector of the right lower leg. This is the directional vector from the right knee joint to the hip joint, i.e., the axial vector of the right thigh. The unit vector representing the normal direction of each body part, i.e., the body coordinates of each body part. The unit vector of the axis, where The normal vector of the right foot. Let be the normal vector of the right lower leg. Here is the normal vector of the right thigh; the axial vectors of each part can be obtained by rotating the normal vector 90 degrees counterclockwise, that is, multiplying the normal vector by a rotation matrix. The calculation process for the left lower limb is the same as that for the right. This is the distance from the center of mass of the right foot to the ankle joint; This is the distance from the ankle joint to the knee joint on the right lower leg; This is the distance from the ankle joint to the knee joint on the right thigh. S2.2.2, Linear Acceleration a This is the acceleration observation value after Kalman filtering correction, expressed as follows: ； In the formula, Z A Acceleration observations corrected by Kalman filtering; S2.2.3, The inclination angle of each part is defined as the normal direction. With x G The included angle is calculated as follows: In the formula, and These represent the ordinate and abscissa values ​​of the unit normal vector, respectively. S2.2.4 Calculation of angular acceleration; Angular acceleration can be obtained from the first difference of the observed angular velocity, and the calculation formula is: ； In the formula, and Adjacent t and t Angular velocity observation at time -1 The sampling time interval; S2.2.5 Obtain the attitude calculation results.

Images