Method for capturing upper limb motion of two-person kayak based on inertial sensor

Abstract

The application belongs to the field of human motion capture, and proposes a two-person kayak upper limb motion capture method based on an inertial sensor, which is applied to an actual rowing scene to perform kinematic posture analysis on the upper limb motion in a two-person kayak. The motion capture system based on the inertial sensor includes six inertial nodes, a receiving node and an upper computer. Each inertial node can send data to the upper computer through the receiving node; after initial calibration, the extended Kalman filtering method is used to improve the data fusion accuracy, and the posture information at each time is solved; the joint angle is extracted from the motion information, and the dynamic time warping method is used to analyze the posture synchronization effect of two rowing athletes. The application is suitable for kayak training in an actual scene, can monitor the two-person motion posture at the same time, synchronously analyze the joint angle, and help coaches and athletes to clearly understand the training effect and improve the technical action.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of human motion capture, in particular to a two-person kayak upper limb motion capture method based on inertial sensors. BACKGROUND

[0002] Kayak has a long history, and it appeared as a water tool for hunting and fishing in the primitive society. With the development of production and the progress of society, kayak has become a common sport in entertainment and sports competition. A complete rowing motion requires full-body coordination and power, involving leg, hip, lower back, shoulder and arm muscle groups. The rower adopts the posture of single-knee kneeling and single-oar rowing, taking the waist as the axis, fully utilizing the body rotation to drive the shoulder and arm movement to obtain the maximum rowing output power.

[0003] The movement ability of kayak depends on the physical fitness and technical proficiency of the athletes. In the process of body rotation, the arm mainly plays a role of link, so analyzing the posture of the upper limb can help the athletes to establish the movement mode more intuitively. For two-person kayak sports, the best effect can be achieved only by the cooperation of two athletes. Through the analysis of the synchronous posture of two persons, the coach can find the most suitable kayak athletes for training to achieve better results.

[0004] The single coach evaluation method is subjective. When the kayak athletes learn new techniques, the intuitive training results cannot accurately locate the technical action defects. Through motion capture, the human motion of the athletes can be tracked and recorded in real time, and the body posture, technical expertise and action defects of the athletes can be analyzed. According to the summary of "Motion Capture Technology and Its Application Research", due to the high requirement of optical motion capture configuration environment, it is not suitable for outdoor environment, so the inertial motion capture method is selected to solve the motion posture of the athletes. The extended Kalman filter method is selected for data fusion to improve the motion capture accuracy. The time series analysis method is used to analyze the joint angles of two persons to provide reference indexes for the coach and athletes to assist training. SUMMARY

[0005] The purpose of the present application is to provide a two-person kayak upper limb motion capture method based on inertial sensors, which can be applied to actual rowing scenes, a two-person kayak upper limb motion capture and posture synchronous analysis system based on inertial sensor technology, and a kinematic posture analysis of the upper limb movement in the two-person kayak.

[0006] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0007] A kind of based on inertial sensor's double kayak upper limb motion capture method, based on a kind of based on inertial sensor's double kayak upper limb motion capture system implementation, based on inertial sensor's double kayak upper limb motion capture system includes: six inertial nodes, a receiving node and host computer;

[0008] Each inertial node is built-in three-axis gyroscope, three-axis accelerometer and three-axis magnetometer, a wireless module and storage card;

[0009] Receiving node is equipped with 32-bit microprocessor and Lora wireless communication module, controls the data acquisition and upload of inertial node, realizes the data communication of host computer with three-axis accelerometer, three-axis gyroscope data and three-axis magnetometer;

[0010] Host computer is written using Qt;With good man-machine interface.

[0011] Based on inertial sensor's double kayak upper limb motion capture system stores in storage card using offline data acquisition mode;Multiple inertial nodes start collecting simultaneously, ensure synchronization;After athlete goes ashore, the data collected by inertial node is transmitted to host computer by receiving node for further processing and analysis;

[0012] Based on inertial sensor's double kayak upper limb motion capture method includes the following steps:

[0013] A1, six inertial nodes are respectively worn in the waist, chest, left upper arm, left forearm, right upper arm and right forearm of human body, and the inertial motion data of the upper limbs of two athletes are collected synchronously;Each inertial node sends data to host computer by receiving node, and outputs three-axis accelerometer data, three-axis gyroscope data and three-axis magnetometer data;

[0014] A2, athlete first stands still facing the north direction of magnetic field and carries out initial calibration, aligns sensor coordinate system with carrier coordinate system;At the same time, the attitude quaternion of sensor coordinate system to geographic coordinate system is obtained, and finally the athlete posture information in geographic coordinate system is described;

[0015] Generally describe the motion posture of object in space, generally relative to the space motion in geographic coordinate system, and in inertial coordinate system, it is acceleration and gyroscope information measured along carrier coordinate system. Therefore, it is necessary to convert the measured value in carrier coordinate system into the measured value in ground coordinate system using mathematical transformation, so as to finally describe the posture information in geographic coordinate system. The measured value in carrier coordinate system is converted into the measured value in geographic coordinate system using mathematical transformation, and finally the athlete posture information in geographic coordinate system is described;

[0016] In the process of updating the carrier attitude, in order to conveniently describe the motion attitude of the object in the geographic coordinate system, the change of the object attitude angle also needs the coordinate transformation system as a reference. The attitude information is updated by using the quaternion method; in the initial state, the geographic coordinate system coinciding with the carrier coordinate system is rotated to the current attitude by the yaw angle ψ around the Z axis, the pitch angle θ around the Y axis and the roll angle φ around the X axis;

[0017]

[0018]

[0019]

[0020] In the geographic coordinate system, the spatial state of the moving object moves in the order of the heading angle, the roll angle and the pitch angle, and the final synthesis result is consistent with the motion state of the object in space. The rotation matrix from the geographic coordinate system GCS to the carrier coordinate system BCS is expressed as:

[0021]

[0022] According to the rotation transformation order of Z-Y-X, the initial attitude quaternion is expressed as:

[0023]

[0024] Since the sensor is fixed on the surface of the limb, when calibrating the initial attitude, the kayak athlete needs to stand north for a period of time, so the initial alignment of ICS and BCS is roughly equal. The initial alignment of BCS and GCS can be described as Then the initial alignment is:

[0025]

[0026] wherein, represents the rotation quaternion of the sensor coordinate system to the geographic coordinate system; represents the rotation quaternion of the carrier coordinate system to the sensor coordinate system; represents the rotation quaternion of the carrier coordinate system to the geographic coordinate system;

[0027] After the initial alignment, the actual positions of the three-axis accelerometer, the three-axis gyroscope data and the three-axis magnetometer are represented by the quaternion (1, 0, 0, 0) in the geographic coordinate system GCS; the attitude quaternion converges to the actual position in multiple iterations;

[0028] A3, the extended Kalman filter method is used for data fusion, and the body posture of the athlete is solved; the human body is regarded as a group of rigid body models, including a plurality of joints with customized lengths, and each joint part is represented by a posture quaternion; the rigid body model is modeled as a line connected by frictionless joints, taking the waist as the root node, updating the body posture at each time, and restoring the motion information;

[0029] The extended Kalman filter method is used for data fusion, taking the posture quaternion, three-axis gyroscope error, human joint position and human joint speed as the system state quantity, and taking the three-axis magnetometer and the position information and speed information calculated from the whole body posture as the system observation value, the state equation is:

[0030]

[0031] Among them, p x , p y , p z are human joint positions, v x , v y , v y are human joint speeds, ω xb , ω yb , ω zb are three-axis gyroscope errors, q0, q1, q2, q3 are posture quaternions;

[0032] The initial covariance matrix is:

[0033]

[0034] Among them, σ p is the human joint position covariance, σ v is the human joint speed covariance, σ q is the posture quaternion covariance, σ ωb is the three-axis gyroscope error covariance; P is the initial value of the error covariance, indicating the confidence in the current predicted state, and determining the initial convergence speed; with the iteration of the Kalman filter, the value of P changes constantly; when the system enters a steady state, the value of P converges to the smallest estimation variance matrix, and the optimal Kalman gain is obtained;

[0035] The posture quaternion update process is as follows:

[0036]

[0037] The state differential equation of the extended Kalman system is:

[0038]

[0039] Among them, a x , a y , az is the three-axis accelerometer value, g is the gravity acceleration;

[0040] The input noise matrix is:

[0041]

[0042] where σ ω is the three-axis gyroscope error variance, σ a is the three-axis accelerometer error variance;

[0043] The system observation equation is:

[0044] z = [p x p y p z v x v y v z ω xm ω ym ω zm ] T

[0045] The gravity component is defined as [0 0 -1], and the observation matrix is obtained as:

[0046]

[0047]

[0048] The observation noise matrix is:

[0049]

[0050] where σ m is the three-axis magnetometer error variance;

[0051] R is the measurement noise; if R is too large, the response of the Kalman filter will be slow. If R is too small, the system is not easy to stabilize. During testing, keep the three-axis gyroscope still, record the output data of the three-axis gyroscope for a period of time, and the output data is normally distributed; according to the 3σ principle, take (3σ) 2 as the initial value of R;

[0052] Substitute the above state equation, initial covariance matrix, state differential equation, observation equation and noise matrix into the extended Kalman filter iteration process, and the Kalman gain K k is:

[0053]

[0054] where P k is the initial covariance matrix, H kis the observation matrix, and R is the observation noise matrix;

[0055] The prior matrix and the posterior matrix are updated as follows:

[0056]

[0057]

[0058]

[0059]

[0060] wherein, is the state quantity, u k-1 is the system input value, Φ is the state transition matrix, Q represents the covariance matrix of the system noise, z k is the observation equation, H k is the observation matrix;

[0061] According to the fused attitude position, the human body attitude at each time is updated taking the waist as the root node, and the motion attitude information is restored.

[0062] A4, according to the results of the attitude calculation of A3, the joint angles between adjacent joint segments are extracted; the left shoulder, right shoulder, left elbow and right elbow joint angles of the two athletes are synchronously analyzed;

[0063] The upper body is defined as eleven segments, and the angle between adjacent joint segments is defined as the joint angle; vectors U and V represent the ipsilateral upper arm and ipsilateral forearm of the upper limb body vector respectively; taking the right arm as an example, the joint angle of the right elbow is defined as the included angle between the right upper arm and the right forearm. Vectors U and V are used to represent the right upper arm and the right forearm of the upper limb body vector respectively. The lengths of vectors U and V are defined according to the actual limb length of the human body; the joint angle is calculated as follows:

[0064]

[0065] wherein, represent two adjacent limb segment vectors in the geographic coordinate system respectively; the change of the joint angle during the movement of the athlete is a time sequence, and the dynamic time warping method DTW is used to measure the correlation of the corresponding joint angles; DTW calculates the similarity between two time sequences by stretching and shortening the time sequences, and twists one of the time sequences on the time axis to realize alignment; after dynamic time warping, the time sequence is calculated according to the Euclidean distance, and the DTW distance measuring the similarity of the two time sequences is calculated; the calculation is as follows:

[0066]

[0067] wherein, d(x i , yi ) represents the distance between x i ,y i ; after normalization, the DTW distance is used to measure the similarity between two time series, which is used to provide a quantitative posture synchronization indicator for coaches and athletes.

[0068] The geographic coordinate system GCS is the spherical coordinate system of the earth; the center of gravity of the human body is taken as the origin, the X axis points to the north of the earth, the Y axis points to the east of the earth, and the Z axis is perpendicular to the surface of the earth and points downward; the inertial coordinate system ICS takes the center of the sensor as the origin and takes the three-axis gyroscope coordinate axis direction as the reference; the carrier coordinate system BCS is defined, taking the center of gravity of the human body as the origin, the direction pointing to the right side of the athlete's body as the X axis, the direction pointing to the front of the athlete's body as the Y axis, and the direction perpendicular to the horizontal plane of the athlete's body downward as the Z axis; the sensor coordinate system takes the center of the sensor as the origin and takes the three-axis gyroscope coordinate axis direction as the reference.

[0069] In step A2, the magnetometer is susceptible to magnetic interference during measurement, that is, the earth's magnetic field at the position where the magnetic sensor is placed is shifted due to the presence of magnetic substances or substances that can affect the local magnetic field strength. During measurement, an ellipsoid fitting method based on least squares is used to correct the output of the three-axis magnetometer.

[0070] The three-axis gyroscope, three-axis accelerometer and three-axis magnetometer have consistent internal coordinate systems and are strictly orthogonal and aligned, supporting high sampling frequency signal output. The range can cover the maximum measurement value of the canoeing movement.

[0071] The overall size of the inertial node is 4.5cm*3.5cm*2.25cm, and the gravity is 42g, which does not affect the athlete's canoeing. Each inertial node is independently powered by a 3.7V (400mAh) rechargeable lithium battery, which can continuously collect data for two hours.

[0072] The beneficial effects of the present application: the present application provides a two-person canoe upper limb motion capture method based on inertial sensors, which adopts a wireless transmission mode to collect human motion information, uses an extended Kalman filtering method to improve data fusion accuracy, and calculates the human joint angle at each time. The present application is suitable for canoe training in actual scenes, can monitor the motion posture of two people at the same time, synchronously analyze the joint angles, and help coaches and athletes to clarify the training effect and improve the technical action. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 is a flow block diagram of the two-person canoe upper limb motion capture method based on inertial sensors of the present application;

[0074] Figure 2 is a hardware schematic diagram of the two-person canoe upper limb motion capture system based on inertial sensors of the present application;

[0075] Figure 3 is a schematic diagram of a data fusion algorithm of the present application;

[0076] Figure 4 is a schematic diagram of rowing motion capture in Example One of the present application;

[0077] Figure 5 is a schematic diagram of inertial sensor node installation in Example Two of the present application;

[0078] Figure 6(a) is a diagram of joint initial angle analysis in Example Two of the present application;

[0079] Figure 6(b) is a diagram of joint normal angle analysis in Example Two of the present application. DETAILED DESCRIPTION

[0080] In order to make the technical problems solved by the present application, the technical solutions adopted and the technical effects achieved more clear, the present application will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present application, and not to limit the present application. In addition, it should be noted that, in order to facilitate the description, only part of the present application is shown in the drawings, not all.

[0081] Figure 1 The present application provides a flow chart of a two-person kayak upper limb motion capture method based on an inertial sensor. The two-person kayak upper limb posture synchronous analysis method provided by the present application can be realized by the two-person kayak upper limb motion capture system based on an inertial sensor provided by the present application, and includes the following steps:

[0082] A1, six inertial nodes are respectively worn on the waist, chest, left upper arm, left forearm, right upper arm and right forearm of the human body through a strap, and the upper limb inertial motion data of two athletes are synchronously collected. Each inertial sensor node sends data to the upper computer through a receiving node, and outputs three-axis accelerometer, three-axis gyroscope data and three-axis magnetometer data.

[0083] An inertial sensor-based two-person kayak upper limb motion capture system based on inertial sensors, comprising six inertial nodes, a receiving node and a host computer. Each of the inertial nodes is internally provided with a three-axis gyroscope, a three-axis accelerometer and a three-axis magnetometer, a wireless module and a storage card. The overall size of the node is 4.5cm*3.5cm*2.25cm, and the weight is 42g, which does not affect the athlete's canoeing. Each node is independently powered by a 3.7V(400mAh) rechargeable lithium battery, which can continuously collect data for two hours. The receiving node is provided with a 32-bit microprocessor and a Lora wireless communication module, which can control the data collection and uploading of the measurement node, and realize the data communication between the PC and each sensor. The host computer is programmed with Qt, which has a good human-computer interaction interface. The system adopts an offline data collection mode stored in the storage card. Multiple nodes start collecting at the same time to ensure timely synchronization. After the kayaking athlete gets off the shore, the data collected by the measurement node is transmitted to the host computer through the receiving node for further processing and analysis.

[0084] A2, the athlete first stands still facing the north direction of the magnetic field for initial calibration, aligns the sensor coordinate system with the carrier coordinate system, and obtains the rotation quaternion of the sensor coordinate system to the geographic coordinate system, finally describes the attitude information in the geographic coordinate system.

[0085] Preferably, the three-axis magnetometer is susceptible to magnetic interference during measurement, that is, the earth's magnetic field at the position where the magnetic sensor is placed is offset due to the presence of magnetic substances or substances that can affect the local magnetic field strength. The present application selects an ellipsoid fitting method based on least squares to correct the output.

[0086] Generally, the motion posture of an object in space is usually relative to the spatial motion in the geographic coordinate system, while in the inertial coordinate system, the acceleration and gyroscope information are measured along the axes of the carrier coordinate system. Therefore, it is necessary to convert the measured values in the carrier coordinate system to the measured values in the geographic coordinate system by using mathematical transformation, so as to finally describe the attitude information in the geographic coordinate system.

[0087] The coordinate system is defined as follows in the present application:

[0088] Geographic coordinate system (GCS): The geographic coordinate system is a spherical coordinate system of the earth. The X-axis points to the north of the earth, the Y-axis points to the east of the earth, and the Z-axis is perpendicular to the surface of the earth and points downward.

[0089] Inertial coordinate system (ICS): The coordinate of the inertial sensor fixed on the body surface is defined.

[0090] Carrier coordinate system (BCS): The direction pointing to the right side of the plane of the moving object is defined as the X-axis, the direction pointing to the front of the object is defined as the Y-axis, and the direction perpendicular to the horizontal plane of the object is defined as the Z-axis.

[0091] In the process of updating the carrier attitude, in order to conveniently describe the motion attitude of the object in the geographic coordinate system, the change of the object attitude angle also needs the coordinate transformation system as a reference. The application selects the quaternion method to update the attitude.

[0092] The geographic coordinate system coinciding with the carrier coordinate system is rotated to the current attitude by the yaw angle (ψ) around the Z axis, the pitch angle (θ) around the Y axis and the roll angle (φ) around the X axis.

[0093]

[0094]

[0095]

[0096] In the geographic coordinate system, the spatial state of the moving object can be moved in the order of the heading angle, the roll angle and the pitch angle, and the final synthesis result is consistent with the motion state of the object in space. The rotation matrix from GCS to BCS can be expressed as:

[0097]

[0098] According to the rotation transformation order of Z-Y-X, the initial attitude quaternion can be expressed as:

[0099]

[0100] Since the sensor is fixed on the surface of the limb, when calibrating the initial attitude, the kayak player needs to stand north for a period of time, so the initial alignment of ICS and BCS is roughly equal. The initial alignment of BCS and GCS can be described as Then the initial alignment is:

[0101]

[0102] After the initial alignment, the actual position of the sensor is represented by the quaternion (1, 0, 0, 0) in the reference coordinate system GCS. The quaternion converges to the actual position in several iterations.

[0103] A3, an extended Kalman filter method is used for data fusion to solve the body posture of the player. The human body is regarded as a group of rigid body models, including a plurality of custom length joints, and each joint part is represented by a quaternion. These body parts are modeled as lines connected by frictionless joints. Taking the waist as the root node, the body posture of each time is updated to restore the motion information.

[0104] Preferably, the present application selects the extended Kalman filter method with attitude quaternion, gyroscope error, node position and node speed as system state quantity, three-axis magnetometer and position and speed information calculated from the whole body attitude as system observation value for data fusion, and the state equation is:

[0105] x = [p x p y p z v x v y v z q0 q1 q2 q3ω xb ω yb ω zb ] T

[0106] The initial covariance matrix is:

[0107]

[0108] P is the initial value of error covariance, indicating the confidence in the current predicted state, and determines the initial convergence speed. With the iteration of the Kalman filter, the value of P will change. When the system enters a steady state, the value of P will converge to a minimum estimation variance matrix, and the Kalman gain at this time is also optimal.

[0109] The quaternion update process is as follows:

[0110]

[0111] The state differential equation of the extended Kalman system is:

[0112]

[0113] The input noise matrix is:

[0114]

[0115] The system observation equation is:

[0116] z = [p x p y p z v x v y v z ω xm ω ym ω zm ] T The gravity component is defined as [0 0 -1], and the observation matrix is obtained:

[0117]

[0118]

[0119] The observation noise matrix is:

[0120]

[0121] R is the measurement noise. If R is too large, the response of the Kalman filter will be slow. If R is too small, the system will not be stable. When testing, keep the gyroscope still and record the output data of the gyroscope for a period of time. The data is approximately normally distributed. According to the 3σ principle, take the 3σ of the normal distribution as the initial value of R. 2 As the initial value of R.

[0122] Substitute the above formula into the extended Kalman filter iteration process, and the Kalman gain K k is:

[0123]

[0124] The prior and posterior matrix updates are as follows:

[0125]

[0126]

[0127]

[0128]

[0129] According to the fused data, the posture position is taken as the root node of the waist, and the posture of the human body at each time is updated to restore the motion information.

[0130] A4, according to the results of the A3 attitude solution, the joint angles between adjacent joint segments are extracted. The left shoulder, right shoulder, left elbow and right elbow joint angles of the two athletes are synchronously analyzed to provide accurate and intuitive dual posture synchronous analysis results for coaches and athletes.

[0131] Preferably, the body in the present application is defined as an eleven-segment structure, and the angle between adjacent joint segments is defined as a joint angle. Taking the right arm as an example, the joint angle of the right elbow is defined as the included angle between the right upper arm and the right forearm. Vectors U and V are used to represent the right upper arm and the right forearm of the upper limb body vector respectively. The length is defined according to the actual length of the human body. The joint angle is calculated as follows:

[0132]

[0133] wherein, respectively represent two adjacent limb segment vectors in the geographic coordinate system.

[0134] The change of joint angle in the canoeing movement can be regarded as a time series. The dynamic time warping (DTW) method can be used to measure the correlation of the corresponding joint angles. The DTW can be used to calculate the similarity between two time series by stretching and shortening the time series, and twisting one of the sequences in the time axis to achieve better alignment. After the dynamic time warping of the time series, the DTW distance measuring the similarity of the two time series can be calculated according to the Euclidean distance. The calculation is as follows:

[0135]

[0136] Wherein, d(x i ,y i ) represents the distance between x i ,y i . After the DTW distance is normalized, it can be used to measure the similarity between two time series. The parameter can provide a quantitative posture synchronization index for the coach and the athlete.

[0137] The embodiment of the application provides a two-person canoeing upper limb motion capture system based on inertial sensing technology, adopts a wireless transmission mode to collect human motion information, uses an extended Kalman filtering method to improve data fusion accuracy, and calculates human joint angles at each time, so as to perform kinematic synchronization analysis on the two-person canoeing posture.

[0138] The application will be further explained and described below in the form of examples:

[0139] Example one,

[0140] The athlete wears a node on the upper limb, and performs a canoeing movement in an outdoor environment. The experimental results are shown in Figure 4 . Through the inertial motion capture device, the human posture information at each time can be intuitively restored. According to the human skeleton model, the joint angles between adjacent joint segments can be accurately calculated.

[0141] Example two,

[0142] In the two-person canoeing process, two athletes wear inertial nodes to synchronously collect data, and the dynamic time warping method is used to analyze the two-person posture synchronization. The wearing diagram is shown in Figure 5 . The inertial nodes are respectively worn on the waists, chests, left upper arms, left lower arms, right upper arms and right lower arms of the two athletes. The experimental results are shown in FIG. 6(a) and FIG. 6(b). The dynamic time warping method can align the two joint angle time series that are out of synchronization in phase, and the result can be used as an index for measuring the two-person synchronization.

Claims

1. A method for capturing upper limb movements in a two-person kayak based on inertial sensors, characterized in that, This is based on an inertial sensor-based upper limb motion capture system for two-person kayaks. The inertial sensor-based upper limb motion capture system for two-person kayaks includes: six inertial nodes, one receiving node, and a host computer. Each inertial node has a built-in three-axis gyroscope, a three-axis accelerometer, and a three-axis magnetometer, a wireless module, and a memory card; The receiving node is equipped with a 32-bit microprocessor and a LoRa wireless communication module to control the data acquisition and uploading of the inertial node, and to realize data communication between the host computer and the three-axis accelerometer, three-axis gyroscope and three-axis magnetometer. The host computer is written in Qt; The inertial sensor-based upper limb motion capture system for two-person kayaks uses offline data acquisition and stores the data on a memory card; multiple inertial nodes start acquiring data simultaneously to ensure synchronization; after the athlete gets out of the water, the data acquired by the inertial nodes is transmitted to the host computer through the receiving node for further processing and analysis. The method for capturing upper limb motion in a two-person kayak based on inertial sensors includes the following steps: A1. Wear six inertial nodes on the waist, chest, left upper arm, left forearm, right upper arm, and right forearm of the human body respectively, and simultaneously collect upper limb inertial motion data of two athletes; each inertial node sends data to the host computer through the receiving node and outputs three-axis accelerometer data, three-axis gyroscope data, and three-axis magnetometer data. A2. The athlete first stands still facing the north direction of the magnetic field for initial calibration, aligning the sensor coordinate system with the carrier coordinate system; at the same time, the attitude quaternion from the sensor coordinate system to the geographic coordinate system with the human body's center of gravity as the origin and the three coordinate axes pointing to geographic north, geographic east and the ground direction respectively is obtained, and finally the athlete's attitude information is described in the geographic coordinate system. Mathematical transformations are used to convert measurements in the carrier coordinate system to measurements in the geographic coordinate system, ultimately describing the athlete's posture information in the geographic coordinate system. Attitude information is updated using the quaternion method; initially, the geographic coordinate system, which coincides with the carrier coordinate system, is updated by the yaw angle around the Z-axis. Pitch angle around the Y-axis and roll angle around the X-axis Rotate to the current orientation; ； ； ； In the geographic coordinate system, the spatial state of a moving object moves in the order of heading angle, roll angle, and pitch angle. The final composite result is consistent with the object's motion state in space. The rotation matrix from the geographic coordinate system GCS to the carrier coordinate system BCS is represented as: Based on the rotational transformation sequence of ZYX, the initial attitude quaternion is represented as: ； The initial alignment description of the carrier coordinate system BCS and the geographic coordinate system GCS is as follows The initial alignment is then: ； in, Represents the rotation quaternion from the sensor coordinate system to the geographic coordinate system; Represents the rotation quaternion from the carrier coordinate system to the sensor coordinate system; Represents the rotation quaternion from the carrier coordinate system to the geographic coordinate system; After initial alignment, the actual positions of the three-axis accelerometer, three-axis gyroscope, and three-axis magnetometer are represented by quaternions (1, 0, 0, 0) in the geographic coordinate system (GCS); the attitude quaternions converge to the actual positions in multiple iterations. A3. The extended Kalman filter method is used for data fusion to solve the athlete's body posture. The human body is regarded as a set of rigid body models, including several joints of customized length. Each joint part is represented by a posture quaternion. The rigid body model is modeled as a line connected by frictionless joints. The waist is used as the root node to update the human posture at each moment and restore the motion information. Using attitude quaternions, triaxial gyroscope errors, human joint positions, and human joint velocities as system state variables, and employing an extended Kalman filter method to fuse data based on triaxial magnetometer readings and position and velocity information calculated from the whole-body attitude as system observations, the state equation is: ； in, , , These are the locations of human joints. , , These are the speeds of human joints. , , These represent the errors of the three-axis gyroscope. , , , These are attitude quaternions; The initial covariance matrix is: ； in, The variance of human joint position For the variance of human joint velocity, Let the variance be the attitude quaternion. Let P be the variance of the three-axis gyroscope error; P is the initial value of the error covariance, representing the confidence level of the current predicted state and determining the initial convergence speed; as the Kalman filter iterates, the value of P changes continuously; when the system enters steady state, the value of P converges to the minimum estimated variance matrix, obtaining the optimal Kalman gain; The attitude quaternion update process is as follows: ； The state differential equation of the extended Kalman system is: ； in, , , These are the values ​​from the triaxial accelerometer, where g is the acceleration due to gravity. The input noise matrix is: ； in, This represents the variance of the three-axis gyroscope error. The variance of the triaxial accelerometer error; The system observation equation is: ； Define the gravitational component as The observation matrix is ​​obtained as follows: ； ； The observation noise matrix is: ； in, The error variance of the triaxial magnetometer; The R-value represents the measurement noise; during the test, the triaxial gyroscope is kept stationary, and its output data is recorded over a period of time. This output data follows a normal distribution. The principle is to take the normal distribution. As the initial value of R; Substituting the aforementioned state equation, initial covariance matrix, state differential equation, observation equation, and noise matrix into the extended Kalman filter iteration process, the Kalman gain... for: ； in, The initial covariance matrix, R is the observation matrix, and R is the observation noise matrix; The prior and posterior matrices are updated as follows: ； ； ； ； in, For state variables, Input values ​​for the system. Here is the state transition matrix. The covariance matrix represents the system noise. It is the observation equation. The observation matrix; Based on the posture position after data fusion, with the waist as the root node, the human posture at each moment is updated to restore the motion posture information. A4. Based on the posture calculation results of A3, extract the joint angles between adjacent joint segments; perform synchronous analysis on the joint angles of the left shoulder, right shoulder, left elbow, and right elbow of the two athletes; The upper body is defined as an eleven-segment structure, and the angle between adjacent joint segments is defined as the joint angle; vectors U and V represent the ipsilateral upper arm and ipsilateral forearm of the upper limb body vector, respectively; the lengths of vectors U and V are defined according to the actual length of the human limbs; the joint angle is calculated as follows: ； in, , These represent two adjacent limb segment vectors in the geographic coordinate system. The changes in joint angles during an athlete's movement are presented as a time series. Dynamic Time Warping (DTW) is used to measure the correlation of the corresponding joint angles. DTW calculates the similarity between two time series by stretching and shortening them, distorting one of the time series on the time axis to achieve alignment. After dynamic time warping, the DTW distance, which measures the similarity between the two time series, is calculated based on Euclidean distance. The calculation is as follows: ； in, represent The distance between them; after DTW distance normalization, it is used to measure the similarity between two time series, and to provide coaches and athletes with a quantitative indicator of posture synchronization.

2. The method for capturing upper limb movements in a two-person kayak based on inertial sensors according to claim 1, characterized in that, The geographic coordinate system GCS is the Earth's spherical coordinate system; with the human body's center of gravity as the origin, the X-axis points to the Earth's north, the Y-axis points to the Earth's east, and the Z-axis is perpendicular to the Earth's surface and points downwards; the carrier coordinate system BCS is defined with the human body's center of gravity as the origin, the X-axis pointing to the athlete's right side, the Y-axis pointing to the athlete's front, and the Z-axis perpendicular to the athlete's horizontal plane and pointing downwards; the sensor coordinate system has the sensor center as the origin and the coordinate axes of the three-axis gyroscope as the reference.

3. The method for capturing upper limb movements in a two-person kayak based on an inertial sensor according to claim 1 or 2, characterized in that, During the measurement process, the output of the triaxial magnetometer was corrected using an ellipsoid fitting method based on the least squares method.

Images