CN104757976B - A kind of Human Body Gait Analysis method and system based on Multi-sensor Fusion - Google Patents

Abstract

The invention relates to the technical field of gait analysis in biomedical engineering, and provides a human gait analysis method and system based on multi-sensor fusion. The method includes: filtering sensor signals to eliminate signal noise errors according to human motion characteristics , use the improved zero-velocity update algorithm to eliminate the integral error, making it suitable for different walking scenes; use the Dinavit-Hartenberg method to fuse multiple sensor data to reduce the calculation error of the leg position; after error correction, accurate calculation The gait speed, step length, stride frequency, gait cycle and walking trajectory of the subjects during the walking process; the gait database is established, and the gait data of different subjects are statistically analyzed by the quantile regression analysis method. The invention can improve the measurement accuracy of gait parameters, and make the gait parameters of different measured persons comparable through standardized processing.

Description

A human gait analysis method and system based on multi-sensor fusion

technical field

The invention relates to the technical field of gait analysis, in particular to a human gait analysis method and system based on multi-sensor fusion.

Background technique

Human gait is the behavioral characteristic of the coordination relationship of the lower limbs in the process of human walking. It involves factors such as personal exercise habits, health status, gender, age, occupation, etc. Human gait detection has significant significance. For example, in terms of sports rehabilitation, gait analysis can evaluate the recovery of the testee's lower limb exercise ability. In terms of telemedicine, portable gait analysis devices can reduce long hours of caregiver attendance. In terms of personal navigation, the position positioning in an environment without GPS signals can be realized by calculating the trajectory information of the pedestrian's lower limbs.

Existing gait analysis methods include traditional methods based on visual inspection, footprint method, optical signal, ultrasonic signal, pressure signal and so on. The US Patent System and method formotion capture with the patent No. 7457439 uses multiple cameras to restore the three-dimensional motion information of the body. However, the optical-based motion measurement system generally has the problem of signal occlusion, and its use is limited; the patent with the publication number of CN 102670207 introduces a gait analysis method based on plantar pressure, which can be identified by analyzing the change of plantar pressure distribution over time. Human body gait phase and lower limb movement information mode, but this method can only obtain the pressure change when the foot is in contact with the ground, and then cannot obtain the gait information of a complete gait cycle; the patent with the publication number of CN101694499 describes a pedestrian The pace measurement system obtains the motion signal of the subject during the exercise through the biaxial acceleration sensor fixed in front of the abdomen. This method is easily affected by the deformation of the abdominal soft tissue, and the accuracy of system gait detection is low.

With the rapid development of microelectromechanical systems (MEMS) technology, wearable sensors have been widely used in the field of human sports rehabilitation. The existing gait analysis methods based on wearable inertial sensors all use the zero-velocity update algorithm to eliminate error accumulation, but the effect of zero-velocity update is heavily dependent on the selection of the algorithm threshold, and the existing methods for determining the threshold often use three-axis The modulus of the acceleration signal is combined with the rate of change (first derivative) of the angular velocity change to determine the time period during which the zero velocity update algorithm can be used. Since the acceleration and acceleration values are not zero during the stance period, but fluctuate around zero, and from the curves of acceleration values and angular velocity values, there is no obvious peak or trough. Therefore it is difficult to find a generally applicable threshold detection method. The existing gait analysis methods all calculate the absolute pace, step length and stride frequency, without considering the influence of height on the detection results, and the measurement results obtained in this way are not comparable for different measured objects. In addition, due to the deformation of the leg muscles during the movement, the relative relationship between the sensor coordinate system and the ground reference coordinate system will change. However, only the feet of the lower limbs of the human body apply the zero-velocity update algorithm to eliminate errors, and the estimation of the position and azimuth angle information of other parts inevitably has large errors.

Contents of the invention

The present invention mainly solves the technical problem in the prior art that it is difficult to effectively eliminate the integration error of the motion signal, and proposes a human gait analysis method and system based on multi-sensor fusion, in order to improve the accuracy of the calculation of lower limb motion information in the process of human walking. purpose of sex and reliability.

The invention provides a human gait analysis method based on multi-sensor fusion, the human gait analysis method based on multi-sensor fusion comprises:

Step 100, using sensors to collect movement signals of lower limbs and three-dimensional geomagnetic field component signals during human walking, where the movement signals include three-dimensional acceleration signals and three-dimensional angular velocity signals;

Step 200, obtain the initial posture of the human body according to the collected motion signal and the three-dimensional geomagnetic field component signal, the initial posture includes the pitch angle, roll angle and yaw angle of the human body's initial static standing state, and obtain the sensor coordinates according to the initial posture of the human body system and the ground coordinate system, and use the deviation to correct the rotation matrix transformed from the sensor coordinate system to the ground coordinate system, so as to compensate the pitch angle, roll angle and yaw angle, and obtain the initial orientation of the human body after compensation information;

Step 300, according to the collected motion signal and the corrected rotation matrix, obtain the starting moment when the human body switches from the standing state to the walking state, and use the extended Kalman filter to perform sensor data fusion from the starting moment to update the human body orientation information, and detect the gait phase in the human walking process according to the collected motion signals, and then obtain the human gait parameters, wherein, the gait phase includes a support phase and a swing phase, and the support phase is divided into heel phase Ground strike period, middle stance phase, full stance period and heel-off period, swing phase is divided into acceleration period, swing period and deceleration period; described human body gait parameters include pace, step length, step Frequency, walking cycle and walking trajectory;

Step 400, eliminating sensor error accumulation and updating human gait parameters, including:

When the moment when the calf is vertical to the ground in the middle stage of the stance phase is detected, the center of gravity of the subject’s body moves forward with the leg on the ground as the swing axis, and a first-order inverted pendulum model is established for the entire human body, and the zero-velocity update algorithm is executed to eliminate the error and update to obtain human gait parameters;

When the full stance phase is detected, the foot is completely attached to the ground, and the calf and thigh of the foot form a rigid body connected by a hinge joint. The kinematic model of the lower body is established using the Dinavitt-Hartenberg method, and the motion of the leg is fused signal and the motion signal of the foot to eliminate errors and update the obtained human gait parameters.

Further, the collection of motion signals of lower limbs and three-dimensional geomagnetic field component signals during human walking includes:

Calibrate the sensor through a three-dimensional turntable and a three-dimensional guide rail;

Collect movement signals of lower limbs and three-dimensional geomagnetic field component signals during human walking;

Perform denoising processing on the collected motion signals and three-dimensional geomagnetic field component signals;

The collected motion signals and three-dimensional geomagnetic field component signals are stored in a storage device.

Further, after step 400, it also includes:

Step 500, standardize the obtained human gait parameters, and then establish a human gait database, including:

by formula To normalize the pace, through the formula To normalize the step size, through the formula Standardize the stride frequency to obtain standardized human gait parameters, where l ₁ is the height of the subject, l _m is the average height of the age group the subject belongs to; The measurer's pace, V _rel is the standard pace, the unit is m/s, L is the step length obtained by integrating the pace, L _rel is the standard step length, the unit is m/step, and C is calculated through the walking cycle C _rel is the standard stride frequency, and the unit is step/second.

Further, the use of sensors to collect movement signals of lower limbs and three-dimensional geomagnetic field component signals during human walking includes:

The movement signals of the lower limbs during human walking are collected through a three-axis accelerometer and a three-axis gyroscope;

Collect the three-dimensional geomagnetic field component signal during human walking through the three-axis electronic compass;

The three-axis accelerometer, the three-axis gyroscope and the three-axis electronic compass are installed on the mid-thigh, mid-calf and instep of the subject.

Further, the two constraints established by the lower limb kinematics model are:

in, is the knee joint displacement vector based on calf motion, is the ankle position vector calculated by the foot-based sensors, is the hip displacement vector calculated by the sensors placed in the thigh.

Correspondingly, the present invention also provides a human gait analysis system based on multi-sensor fusion, the human gait analysis system based on multi-sensor fusion includes: a data acquisition device and a data analysis and processing device, the data acquisition device is used to utilize The sensor collects the motion signal of the lower limbs and the three-dimensional geomagnetic field component signal during human walking, and the motion signal includes three-dimensional acceleration and three-dimensional angular velocity; the data analysis and processing device includes an initial posture analysis unit, a gait parameter calculation unit and an error correction unit;

The initial posture analysis unit is used to obtain the initial posture of the human body according to the collected motion signal and the three-dimensional geomagnetic field component signal, the motion signal includes the three-dimensional acceleration signal and the three-dimensional angular velocity signal, and obtains the sensor coordinate system and the ground coordinate according to the initial posture of the human body System deviation, using the deviation to modify the rotation matrix transformed from the sensor coordinate system to the ground coordinate system, to compensate the pitch angle, roll angle and yaw angle, and obtain the compensated initial orientation information of the human body;

The gait parameter calculation unit is used to obtain the start moment when the human body switches from the standing state to the walking state according to the collected motion signal and the corrected rotation matrix, and use the extended Kalman filter to perform sensor Data fusion, update the orientation information of the human body, and detect the gait phase in the walking process of the human body according to the collected motion signals, and then obtain the gait parameters of the human body, wherein the gait phase includes the support phase and the swing phase, and the support phase The phases are divided into the heel-strike phase, the middle phase of the stance phase, the complete standing phase and the heel-off phase, and the swing phase is divided into the acceleration phase, the swing phase and the deceleration phase; , step length, stride frequency, walking cycle and walking trajectory;

The error correction unit is used to eliminate the accumulation of sensor errors and update the gait parameters of the human body, including: when the moment when the calf is perpendicular to the ground in the middle stage of the stance phase is detected, the center of gravity of the subject's body moves forward with the leg on the ground as the swing axis, and the whole The human body establishes a first-level inverted pendulum model, executes a zero-velocity update algorithm to eliminate errors, and updates the obtained human gait parameters; when the full stance phase is detected, the foot is completely attached to the ground, and the calf and thigh of the foot are connected through a hinge joint The rigid body of the lower limbs is established using the Dinavitt-Hartenberg method, and the motion signals of the legs and feet are fused to eliminate errors and update the obtained human gait parameters.

Further, the data acquisition device includes an inertial sensor calibration unit, a sensor signal acquisition unit, a data filtering unit and a self-contained data storage unit;

The inertial sensor calibration unit is used to calibrate the sensor through the three-dimensional turntable and the three-dimensional guide rail;

The sensor signal acquisition unit is used to collect the motion signal of the lower limbs and the three-dimensional geomagnetic field component signal during human walking;

The data filtering unit is used for denoising the collected motion signal and three-dimensional geomagnetic field component signal;

The self-capacity data storage unit is used to save the collected motion signal and three-dimensional geomagnetic field component signal into the storage device.

Further, the human gait analysis system based on multi-sensor fusion also includes:

The gait database unit is used to standardize the obtained human gait parameters, and then establish a human gait parameter database, which is specifically used for; through the formula To normalize the pace, through the formula To normalize the step size, through the formula Standardize the stride frequency, where l ₁ is the height of the testee, l _m is the average height of the age group of the testee; V is the pace of the testee obtained by integrating the sensor signal, V _rel is the standard step Speed, the unit is m/s; is the step length obtained by integrating the step speed, L _rel is the standard step length, the unit is m/step; C is the step frequency of the subject per unit time calculated through the walking cycle , C _rel is the standard stride frequency, the unit is step/second.

Further, the data acquisition device includes a three-axis accelerometer, a three-axis gyroscope and a three-axis electronic compass;

The three-axis accelerometer and the three-axis gyroscope collect the movement signals of the lower limbs during human walking;

The three-axis electronic compass collects three-dimensional geomagnetic field component signals during human walking;

A three-axis accelerometer, a three-axis gyroscope and a three-axis electronic compass are installed on the middle part of the thigh, the middle part of the calf and the dorsum of the foot of the subject.

Further, in the error correction unit, the two constraints established by the lower limb kinematics model are:

in, is the knee joint displacement vector based on calf motion, is the ankle position vector calculated by the foot-based sensors, is the hip displacement vector calculated by the sensors placed in the thigh.

A human gait analysis method and system based on multi-sensor fusion provided by the present invention has the following advantages compared with the prior art:

1. The use of auxiliary sensors placed on the lower leg significantly improves the effectiveness of the foot zero velocity update algorithm.

2. Apply the Dynawirt-Hartenberg coordinate transformation method in the field of robotics to the lower limb motion model to reduce the position calculation error and improve the calculation accuracy of the leg position information.

3. Standardize the pace, step length, and stride frequency among the calculated human gait parameters, so as to realize the horizontal comparability of the gait data of different subjects. By performing quantile regression analysis on the gait data of subjects of different genders and ages, the changing trend of gait parameters can be obtained.

4. Data acquisition does not depend on external equipment, and it is completely self-contained storage, which eliminates the disadvantages of data packet loss in wireless transmission mode.

5. It has high measurement accuracy, high sensitivity, low cost, and is convenient for operators to use. It can be used for accurate measurement of gait information and detection of motion information of other parts of the human body.

Description of drawings

Fig. 1 is the implementation flowchart of the human gait analysis method based on multi-sensor fusion provided by the embodiment of the present invention;

Fig. 2 is a schematic diagram of sensor installation in an embodiment of the present invention;

3 is a schematic diagram of data collected by a three-axis acceleration, a three-axis gyroscope and a three-axis electronic compass in an embodiment of the present invention;

Fig. 4 is a schematic diagram of the coordinate system transformation of the human lower limb sensor in the embodiment of the present invention;

Fig. 5 is a schematic diagram of three-dimensional direction angle information of the subject who walks two circles along the rectangular route calculated in Example 1 in the embodiment of the present invention;

6 is a schematic diagram of the three-dimensional angle information of the foot of the measured person walking ten steps along a straight line calculated in Example 1 in the embodiment of the present invention;

Fig. 7 is a schematic diagram of the relationship between lower limb pain and lower limb joint rotation angle;

Fig. 8 is the schematic diagram of the walking track under the three-dimensional coordinate system when the subject climbs two floors of stairs calculated by Example 2 in the embodiment of the present invention;

FIG. 9 is a structural diagram of a human gait analysis system based on multi-sensor fusion provided by an embodiment of the present invention.

detailed description

In order to make the technical problems solved by the present invention, the technical solutions adopted and the technical effects achieved clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention, but not to limit the present invention. In addition, it should be noted that, for the convenience of description, only parts related to the present invention are shown in the drawings but not all content.

Embodiment one

Fig. 1 is a flow chart of the realization of the human gait analysis method based on multi-sensor fusion provided by the embodiment of the present invention. The human gait analysis method based on multi-sensor fusion provided by the embodiment of the present invention can be executed by the human gait analysis system based on multi-sensor fusion provided by the embodiment of the present invention, and the system can be realized by software and/or hardware. As shown in Figure 1, the human gait analysis method based on multi-sensor fusion provided by the embodiment of the present invention includes:

Step 100, using sensors to collect movement signals of lower limbs and three-dimensional geomagnetic field component signals during human walking.

Wherein, the motion signal includes a three-dimensional acceleration signal and a three-dimensional angular velocity signal. Specifically, the sensor is calibrated through the three-dimensional turntable and the three-dimensional guide rail; the motion signal of the lower limbs and the three-dimensional geomagnetic field component signal during human walking are collected; the collected motion signal and the three-dimensional geomagnetic field component signal are de-noised; the collected The motion signal and the three-dimensional geomagnetic field component signal are stored in the storage device. Wherein, the denoising processing may include performing high-pass (0.001 Hz) filtering, low-pass (5 Hz) filtering and notch wave (50 Hz) processing on the detected three-dimensional acceleration signal, three-dimensional angular velocity signal and three-dimensional geomagnetic field component signal. Fig. 2 is a schematic diagram of sensor installation in the embodiment of the present invention. Referring to Figure 2, the movement signals of the lower limbs during human walking can be collected through the three-axis accelerometer and the three-axis gyroscope, and the three-dimensional geomagnetic field component signals during the human walking process can be collected through the three-axis electronic compass; the three-axis accelerometer, three-axis gyroscope The instrument and the three-axis electronic compass are installed on the mid-thigh, mid-calf and instep of the human body. Fig. 3 is a schematic diagram of data collected by a three-axis acceleration, a three-axis gyroscope and a three-axis electronic compass in an embodiment of the present invention. Figure 3 can be obtained by processing the data collected by the three-axis acceleration, three-axis gyroscope and three-axis electronic compass with a computer. In Figure 3, X, Y, and Z represent the three mutually orthogonal coordinate axes of the sensor, VPT represents the energy of the motion signal, which is used to detect the starting moment of walking, and ZUPT represents the switch signal of the zero speed update algorithm.

For the accuracy of data collection, the sensor can be calibrated through the three-dimensional turntable and three-dimensional guide rail before collecting data, in order to avoid errors caused by sensor baseline drift. The specific process of calibration is: the static parameter calibration of the accelerometer, using the gravity g to make each axis of the sensor coincide with the gravity direction, and detecting the deviation between the sensor output and g; the dynamic parameter calibration of the accelerometer, using the three-dimensional guide rail, applying Acceleration, the sensor is respectively fixed on the three axes of the guide rail, and the fixed acceleration generated by the induction guide rail due to the motor traction belt is compared with the output of the sensor to calculate the deviation; the static parameter calibration of the gyroscope, the sensor is static, and the output of the three axes should be If it is not zero, record the deviation of each axis separately; gyroscope dynamic parameter calibration, the sensor is fixed on the turntable, so that the center of gravity of the sensor and the center of the turntable are reconciled, and the three axes of the gyroscope coincide with the rotation axis of the turntable , turn on the motor to apply a fixed speed to the turntable, record the output value of the gyroscope and compare the speed of the turntable to obtain the offset. The offset is eliminated in the subsequent calculation formula, thereby eliminating the inherent error of the sensor.

Specifically, the collected motion signals (three-dimensional acceleration signals and three-dimensional angular velocity signals) and three-dimensional geomagnetic field component signals of the lower limbs during human walking are stored in a mobile memory card (SD card), and then transmitted to the data center through a card reader. in equipment for analytical processing.

Step 200, obtain the initial posture of the human body according to the collected motion signal and the three-dimensional geomagnetic field component signal, obtain the deviation between the sensor coordinate system and the ground coordinate system according to the initial posture of the human body, and use the deviation to correct the transformation from the sensor coordinate system to The rotation matrix of the ground coordinate system is used to compensate the pitch angle, roll angle and yaw angle, and obtain the initial orientation information of the human body after compensation.

Wherein, the initial posture includes the pitch angle, roll angle and yaw angle of the human body's initial static standing state. At the beginning of the measurement, the subject is standing still. The pitch angle and roll angle are obtained by measuring the components of gravity on the sagittal plane and the horizontal plane. The electronic compass measures the components of the earth's magnetic field on the three planes of the sensor to calculate the initial yaw angle. The specific process is: take the gravitational acceleration g as the reference vector, obtain the components of the gravitational acceleration on the sagittal plane and the horizontal plane, and then get the pitch angle and roll angle; take the component H of the geomagnetic field vector on the local horizontal plane as the reference vector, and use the three-axis electronic compass The ratio of the measured values H _x and H _y of the horizontal X-axis and Y-axis to H is used to calculate the angle between the initial azimuth and the true north direction, and then obtain the yaw angle.

Fig. 4 is a schematic diagram of the coordinate system transformation of the human lower limb sensor in the embodiment of the present invention. Referring to Figure 4, when the human body is standing still, in the ground reference coordinate system, the gravitational acceleration vector is [0,0,g] ^T , and the calculated value from the rotation matrix to the sensor coordinate system is [x,y,z] ^T , the measured value of acceleration in the sensor coordinate system is [a,b,c] ^T , [x,y,z] ^T and [a,b,c] ^T all represent the gravitational acceleration vector in the sensor coordinate system, for The error [e _x , e _y , e _z ] ^T can be obtained by doing the vector product of these two vectors, and the rotation matrix can be corrected by using this error vector. The corrected sensor coordinate system is shown in Figure 4. The above correction is only to coincide the XOY plane of the ground reference coordinate system and the sensor coordinate system. For the rotation around the Z axis, that is, the yaw angle, the accelerometer has nothing to do, and its measured value is always [0,0,g] ^T , only Magnetometers can be relied upon for further compensation. The measurement object of the three-axis electronic compass is the magnetic field vector. In a relatively pure electromagnetic environment, the measurement object of the electronic compass is the geomagnetic field. The direction of the geomagnetic field forms an included angle with the horizontal plane. The components of the geomagnetic field in the three planes are recorded as [u, v,w] ^T , if the X axis of the sensor is aligned with the true north direction, then v=0, the geomagnetic component is [u,0,w] ^T , and the output of the electronic compass in the sensor coordinate system is [i,j,k ] ^T , after accelerometer compensation (coordinate system rotation), [i′,j′,k′] ^T is obtained. On the XOY plane of the ground reference coordinate system, the projection of [u,0,w] ^T is u, [ i′,j′,k′] The projection of ^T is The size of the projection vectors of the geomagnetic field on the XOY plane must be the same, so there is At the same time w=k', the processed [i',j',k'] ^T is transferred to the sensor coordinate system through the transposition of the rotation matrix, and the obtained vector is vector product with [i,j,k] ^T to find the error , modify the rotation matrix again to complete the compensation of the yaw angle, and the accurate initial quaternion can be obtained after using the deviation correction. The initial quaternion is used to describe the initial orientation information of the measured object, and the orientation change of the measured object can be calculated by quaternion multiplication on the basis of the initial quaternion. The theoretical value of the initial quaternion is [1,0,0,0] ^T .

Step 300, according to the collected motion signal and the corrected rotation matrix, obtain the starting moment when the human body switches from the standing state to the walking state, and use the extended Kalman filter to perform sensor data fusion from the starting moment to update the human body Orientation information, and detect the gait phase of the human body in the process of walking according to the collected motion signals, and then obtain the human body gait parameters.

Among them, the starting moment when the human body switches from the standing still state to the walking state, that is, the moment when the first step is taken, the way to obtain the starting moment when the human body switches from the standing still state to the walking state is: through the acceleration energy signal threshold detection. . When the subject switches from a stationary state to a walking state, by calculating the energy of the acceleration signal Compared with the preset threshold λ, when the energy of the acceleration signal is greater than the threshold λ for the first time, it is the moment when the human body starts to walk. In the formula, a _x represents the component of the accelerometer output signal on the X-axis of the ground coordinate system, a _y Indicates the component of the accelerometer output signal on the Y-axis of the ground coordinate system, and a _z indicates the component of the accelerometer output signal on the Z-axis of the ground coordinate system.

From the start moment, the extended Kalman filter is used for sensor data fusion, the orientation information of the human body is updated, and the gait phase in the process of human walking is detected according to the collected motion signals, and then the gait parameters of the human body are obtained. The extended Kalman filter is an optimal autoregressive data processing algorithm. In the present invention, the extended Kalman filter is used in the vicinity of the filter value, and the Taylor expansion algorithm is used to expand the nonlinear system, and all the higher-order items above the second order are discarded, so that the original system becomes a linear system, and the reuse The idea of the standard Kalman filter algorithm is to filter and fuse the linearized model of the system to achieve sensor data fusion. After using the extended Kalman filter for sensor data fusion, the orientation information of the human body is updated, that is, the quaternion is updated, and the gait phase in the process of human walking is detected according to the collected motion signals, and then the gait parameters of the human body are obtained. Among them, the gait phase includes the support phase and the swing phase. The support phase is divided into the heel strike phase, the middle stance phase, the full stance phase and the heel off the ground phase. The swing phase is divided into the acceleration phase, the swing phase and the deceleration phase. Expect. For example, the gait phase of the human body can be analyzed through Fig. 3 . Human gait parameters include pace, step length, stride frequency, walking cycle and walking trajectory during human walking. The process of obtaining human gait parameters is as follows: the three-dimensional acceleration vector is integrated to obtain the three-dimensional velocity vector, and then the three-dimensional velocity vector is integrated to obtain the three-dimensional displacement vector; combined with the calculated walking cycle information, the pace and step length can be calculated and step frequency parameters; through the relationship between the quaternion and the Euler angle, the three-dimensional azimuth angle can be calculated; the three-dimensional displacement vector and the azimuth angle can be integrated, and the whole walking process can be described using the three-dimensional coordinate axis to obtain the walking track. The initial three-dimensional position of the foot is set to [X ₀ , Y ₀ , Z ₀ ] ^T , and the foot position information at each data sampling moment is updated to [X _t , Y _t , Z _t ] ^T by integrating the three-dimensional velocity vector , the azimuth angle includes the foot movement roll angle, pitch angle and yaw angle during walking. the initial orientation specified by the initial quaternion Using the angular velocity information detected by the gyroscope, use the quaternion multiplication operation to update the orientation as Combining the three-dimensional position and azimuth angle can obtain the walking trajectory.

Step 400, eliminate sensor error accumulation, and update human gait parameters.

Since the two integral calculations in step 300 will generate and amplify the integral error, in the middle of the stance phase of each step in the walking process, if the zero-velocity update algorithm condition is satisfied, the velocity and acceleration at the moment when the foot is completely attached to the ground are equal to The specific domain assumption of zero eliminates the accumulation of sensor errors, making the human gait parameters more accurate, and the error is limited within each small step of walking motion, minimizing the error.

When the moment when the calf is vertical to the ground in the middle stage of the stance phase is detected, the center of gravity of the subject’s body moves forward with the leg on the ground as the swing axis, and a first-order inverted pendulum model is established for the entire human body, and the zero-velocity update algorithm is executed to eliminate the error and update to obtain human gait parameters. The specific process is: in the process of walking, the human body supports the ground with one foot, and the center of gravity of the subject's body moves forward with the leg on the ground as the swing axis. A first-level inverted pendulum model can be established for the entire human body. According to the gravitational potential energy of the inverted pendulum model The principle of reaching the maximum value at the highest point and the minimum value of the speed, combined with the acceleration signal of the sensor placed on the lower leg to find the minimum leg speed moment in the standing phase, this moment is the starting moment for the zero-speed update algorithm. Use the fixed sensor at the calf to detect the minimum speed moment, and determine that this is the moment when the calf is perpendicular to the ground, that is, the moment when the zero-speed update algorithm starts to execute. Executing the zero-velocity update algorithm eliminates errors and avoids introducing integral errors into the next gait cycle, which can improve the effectiveness of the zero-velocity update algorithm.

When the full stance phase is detected, the foot is fully attached to the ground, and the foot, calf, and thigh can be considered as rigid bodies connected by hinge joints, and the lower body kinematics are established using the Denavit-Hartenberg method model, and fuse the motion signal of the leg and the motion signal of the foot to eliminate the error and update the obtained human gait parameters. The specific process is as follows: Referring to Figure 2, combined with the physiological constraints and linkage relationship of the thigh, calf and foot, for example, the knee joint has only one degree of freedom of movement, and the ankle joint has three degrees of freedom of movement. For legs that cannot directly use the zero-velocity update algorithm The lower limb kinematics model is established using the Dinavitt-Hartenberg method, and the two constraints established by the lower limb kinematics model are used to reduce the position and azimuth estimation errors of the leg sensor. The two constraints of the lower limb kinematics model are:

in, represents the knee joint displacement vector based on the calf motion, represents the ankle position vector calculated by the foot-based sensors, represents the hip displacement vector calculated by the sensors placed in the thigh,

Among them, α, β, and γ are the Euler angles around the X, Y, and Z axes respectively, and L ₁ is the length of the calf measured by the anthropometer.

Step 500, standardize the acquired human gait parameters, and then establish a human gait database.

Since everyone instinctively chooses the pace, length, and frequency that best suit their athletic ability, a significant decrease in pace is a clear indicator of pathology. Therefore, it is necessary to consider the influence of height on the detection results when standardizing a large number of paces, step lengths, and stride frequencies obtained in experiments. It is obvious that the pace is closely positively correlated with the length of the lower limbs l ₀ , l ₀ is difficult to measure accurately, so the height l ₁ is used instead, and the average height of the subjects is recorded as L _m . The process of standardizing the obtained human gait parameters is as follows: through the formula To normalize the pace, through the formula To normalize the step size, through the formula Standardize the stride frequency, where l ₁ is the height of the testee, l _m is the average height of the age group of the testee; V is the pace of the testee obtained by integrating the sensor signal, V _rel is the standard step Speed, the unit is m/s, L is the step length obtained by integrating the step speed, L _rel is the standard step length, the unit is m/step, C is the step that the subject walks per unit time calculated through the walking cycle C _rel is the standard stride frequency, and the unit is steps per second. A gait index that deviates significantly from the standard range means an unstable gait. Normalizing the calculated parameters of stride speed, stride length and stride frequency enables lateral comparability of the calculated gait parameters.

By establishing a database of human gait parameters such as pace speed, step length, stride frequency, walking cycle, and walking trajectory, it is possible to perform quantile regression analysis on massive gait data for subjects of different genders and ages, and obtain Trends in human gait parameters. And the calculated gait parameters such as pace, step length, stride frequency, and walking trajectory can be intuitively displayed on the database interface in the form of pie charts, histograms, and radar charts.

The human body gait analysis method based on multi-sensor fusion provided by the present invention is not only suitable for walking on flat ground, but also suitable for moving up and down stairs. The scheme provided by this embodiment is described below in the form of an example:

Example 1, in the embodiment of indoor walking on flat ground, the subject performs a "stomp" action after binding the gait analysis system, so that the sensor worn by the subject senses an obvious initial signal, which is used to Complete the data synchronization between the video tracking system VICON and the sensor. The following two groups of experiments were: the subject walked two circles along the rectangular route in the VICON measurement area of the optical tracking system; the subject walked ten steps along a straight line at his own comfortable pace. The upper computer is the data calculation and output terminal, and the data collected by the lower computer is transmitted to the gait analysis upper computer software through the mobile storage medium, and various gait parameters are calculated. The flow chart of data analysis and processing is shown in Figure 1. The gait data collected by the lower computer is uploaded to the upper computer gait analysis software, and firstly processed by high-pass and low-pass digital filters to filter out high-frequency and low-frequency signals irrelevant to the application of gait analysis. Since the subject must stand still at the beginning of the measurement, the system initialization conditions are met. At this time, the accelerometer only feels the acceleration of gravity, and the initial pitch angle and roll angle are calculated using the accelerometer readings. Use the electronic compass to detect the geomagnetic component and calculate the initial yaw angle. Errors were eliminated using a zero-velocity update algorithm in the middle of the stance phase after the onset of walking. Fig. 5 is the three-dimensional direction angle information obtained by the subject calculated in Example 1 in the embodiment of the present invention and walks two circles along the rectangular route. The calculation result in Fig. 5 is compared with the numerical value calculated by the video tracking system, and the angle error is less than 2°, indicating that This method can accurately calculate gait information for a long time. In order to obtain the details of the foot angle and exclude the influence of changes in the walking direction, the subject walked ten steps in a straight line according to his own comfortable pace, and the sensor placed on the foot calculated the change of the foot angle during walking, and the internal rotation of the foot could be evaluated Or the degree of foot eversion, to help the subject choose the right shoes. FIG. 6 is a schematic diagram of the three-dimensional angle information of the foot of the subject in Example 1 walking straight in the embodiment of the present invention. Fig. 7 is a schematic diagram of the relationship between lower limb pain and lower limb joint rotation angle.

Example 2, in this embodiment, the subject completes the movement of climbing two floors of stairs, and the collected data is first stored in the SD card. After the experiment is over, the card reader is passed into the gait analysis host computer software to calculate various parameters . The endurance of stair exercise can be evaluated by performing the stair exercise experiment on the same subject several times per unit time. Stair exercise can improve cardiovascular function, strengthen myocardial contractility, improve lung function, increase lung capacity, develop lower limb muscle strength, and improve knee joint toughness. The stair exercise capacity indirectly reflects the subjects' cardiovascular and pulmonary function indicators, as well as the strength and toughness of the joints of the lower extremities. FIG. 8 is a schematic diagram of the walking track in the three-dimensional coordinate system when the subject climbs two floors of stairs calculated in Example 2 of the embodiment of the present invention.

The human gait analysis method based on multi-sensor fusion provided by this embodiment can solve the problem of the validity of the zero-velocity update algorithm used in the prior art, and can accurately find the suitable zero-velocity update algorithm by fusing the sensor data of feet and legs. At the initial moment of the algorithm, based on the specific domain assumption that the speed and acceleration are zero when the foot is completely attached to the ground, the error or error accumulation is reduced, and the error is limited to each small step in the walking process to avoid error accumulation; In the case where the zero-velocity update algorithm is not applicable to the calculation of the leg position, the constraint conditions established by the topological structure model of the lower limbs can be used to reduce the position calculation error. Aiming at the influence of the height factor on the gait parameters, the present invention standardizes the parameters of the pace, step length and stride frequency, which can eliminate the influence of the height factor on the gait parameters and achieve the effect that the gait parameters of different people are comparable. The gait database is established by summarizing standardized gait parameters, and the variation trend of human gait parameters with gender and age can be obtained through quantile regression analysis. The method provided in this embodiment can accurately find the initial moment for which the zero-speed update algorithm is applicable, and ensure the validity of the zero-speed update algorithm, thereby reducing integral errors and improving the reliability of gait parameter calculation. The method provided by the invention can also be used for the measurement of motion information of other body parts.

Embodiment two

FIG. 9 is a schematic structural diagram of a human gait analysis system based on multi-sensor fusion provided by an embodiment of the present invention. As shown in Figure 9, the human gait analysis system based on multi-sensor fusion provided by the embodiment of the present invention includes: a data acquisition device and a data analysis and processing device, the data acquisition device is used to use sensors to collect movement signals and The three-dimensional geomagnetic field component signal, the motion signal includes three-dimensional acceleration and three-dimensional angular velocity; the data analysis and processing device includes an initial posture analysis unit, a gait parameter calculation unit and an error correction unit;

The initial posture analysis unit is used to obtain the initial posture of the human body according to the collected motion signal and the three-dimensional geomagnetic field component signal, the motion signal includes the three-dimensional acceleration signal and the three-dimensional angular velocity signal, and obtains the sensor coordinate system and the ground coordinate according to the initial posture of the human body System deviation, using the deviation to modify the rotation matrix transformed from the sensor coordinate system to the ground coordinate system, to compensate the pitch angle, roll angle and yaw angle, and obtain the compensated initial orientation information of the human body;

The gait parameter calculation unit is used to obtain the start moment when the human body switches from the standing state to the walking state according to the collected motion signal and the corrected rotation matrix, and use the extended Kalman filter to perform sensor Data fusion, update the orientation information of the human body, and detect the gait phase in the process of human walking according to the collected motion signals, and then obtain the gait parameters of the human body, wherein the gait phase includes the support phase and the swing phase, and the support phase The phases are divided into the heel-strike phase, the middle phase of the stance phase, the complete standing phase and the heel-off phase, and the swing phase is divided into the acceleration phase, the swing phase and the deceleration phase; , step length, stride frequency, walking cycle and walking trajectory;

The error correction unit is used to eliminate the accumulation of sensor errors and update the gait parameters of the human body, including: when the moment when the calf is perpendicular to the ground in the middle stage of the stance phase is detected, the center of gravity of the subject's body moves forward with the leg on the ground as the swing axis, and the whole The human body establishes a first-level inverted pendulum model, executes a zero-velocity update algorithm to eliminate errors, and updates the obtained human gait parameters; when the full stance phase is detected, the foot is completely attached to the ground, and the calf and thigh of the foot are connected through a hinge joint The rigid body of the lower limbs is established using the Dinavitt-Hartenberg method, and the motion signals of the legs and feet are fused to eliminate errors and update the obtained human gait parameters.

In the above solution, the data acquisition device includes an inertial sensor calibration unit, a sensor signal acquisition unit, a data filtering unit, and a self-contained data storage unit; the inertial sensor calibration unit is used to calibrate the sensor through a three-dimensional turntable and a three-dimensional guide rail; The sensor signal acquisition unit is used to collect the motion signal of the lower limbs and the three-dimensional geomagnetic field component signal during human walking; the data filtering unit is used to denoise the collected motion signal and the three-dimensional geomagnetic field component signal; the self-contained data The storage unit is used to store the collected motion signals and three-dimensional geomagnetic field component signals in a storage device. Specifically, the denoising processing of the data preprocessing unit includes high-pass (0.001 Hz) filtering, low-pass (5 Hz) filtering and notch (50 Hz )deal with. Among them, the self-contained data storage unit is to store the collected motion signals of the lower limbs and the three-dimensional geomagnetic field component signals in the mobile memory card (SD card), and then transmit them to the data analysis and processing through the card reader. device.

Wherein, the data acquisition device includes a three-axis accelerometer, a three-axis gyroscope and a three-axis electronic compass; the three-axis accelerometer and the three-axis gyroscope collect the movement signals of the lower limbs during human walking; the three-axis electronic compass collects the human walking process The three-dimensional geomagnetic field component signal; the three-axis accelerometer, the three-axis gyroscope and the three-axis electronic compass are installed on the mid-thigh, mid-calf and instep of the subject.

Further, the human gait analysis system based on multi-sensor fusion also includes:

The gait database unit is used to standardize the obtained human gait parameters, and then establish a human gait parameter database, which is specifically used for; through the formula To normalize the pace, through the formula To normalize the step size, through the formula Standardize the stride frequency, where l ₁ is the height of the testee, l _m is the average height of the age group of the testee; V is the pace of the testee obtained by integrating the sensor signal, V _rel is the standard step speed, the unit is m/s; L is the step length obtained by integrating the step speed; L _rel is the standard step length, the unit is m/step; Frequency, C _rel is the standard stride frequency, the unit is step/second.

In the error correction unit, the two constraints established by the lower limb kinematics model are:

in, is the knee joint displacement vector based on calf motion, is the ankle position vector calculated by the foot-based sensors, is the hip displacement vector calculated by the sensors placed in the thigh.

The human gait analysis system based on multi-sensor fusion provided by this embodiment collects the movement signals of the lower limbs and the three-dimensional geomagnetic field component signals during human walking through the data acquisition device, obtains the initial orientation information of the human body through the initial posture analysis unit, and obtains the initial orientation information of the human body through the gait The parameter calculation unit obtains the gait parameters of the human body, and eliminates the accumulation of sensor errors through the error correction unit to update the gait parameters of the human body. The system provided in this embodiment can accurately find the initial moment when the zero-speed update algorithm is applicable, ensuring the accuracy of the zero-speed update algorithm. Effectiveness, thereby reducing the integral error and improving the reliability of gait parameter calculation.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: Modifications to the technical solutions described in the foregoing embodiments, or equivalent replacement of some or all of the technical features thereof, do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the various embodiments of the present invention.

Claims (8)

1. A human gait analysis method based on multi-sensor fusion is characterized by comprising the following steps:

step 100, acquiring motion signals of lower limbs and three-dimensional geomagnetic field component signals in the walking process of a human body by using a sensor, wherein the motion signals comprise three-dimensional acceleration signals and three-dimensional angular velocity signals;

200, acquiring an initial posture of a human body according to the acquired motion signal and the three-dimensional geomagnetic field component signal, wherein the initial posture comprises a pitch angle, a roll angle and a yaw angle of the human body in an initial static standing state, acquiring a deviation amount of a sensor coordinate system and a ground coordinate system according to the initial posture of the human body, correcting a rotation matrix transformed from the sensor coordinate system to the ground coordinate system by using the deviation amount, compensating the pitch angle, the roll angle and the yaw angle, and acquiring compensated initial orientation information of the human body;

step 300, according to the collected motion signals and the corrected rotation matrix, obtaining the starting time of switching the human body from a static standing state to a walking state, performing sensor data fusion by using an expanded Kalman filter from the starting time, updating human body orientation information, and detecting a gait time phase in the walking process of the human body according to the collected motion signals so as to obtain human body gait parameters, wherein the gait time phase comprises a support time phase and a swing time phase, the support time phase comprises a heel strike period, a standing phase middle period, a full standing period and a heel off period, and the swing time phase comprises an acceleration period, a swing period and a deceleration period; the human body gait parameters comprise the pace, the step length, the step frequency, the walking cycle and the walking track in the human body walking process;

step 400, eliminating sensor error accumulation and updating human gait parameters, comprising:

when the time that the crus are vertical to the ground in the standing phase is detected, the center of gravity of the tested person body moves forwards by taking the grounded legs as swing axes, a first-stage inverted pendulum model is built for the whole person, a zero-speed updating algorithm is executed to eliminate errors, and the obtained gait parameters of the person are updated;

when a full standing period is detected, the foot is completely attached to the ground, the crus and the thighs of the foot form a rigid body connected through hinge joints, a lower limb kinematics model is established by using a Dinavier-Hartmanberg method, and a motion signal of the legs and a motion signal of the foot are fused to eliminate errors and update the obtained human gait parameters.

2. The human gait analysis method based on multi-sensor fusion of claim 1, wherein the collecting the motion signals of the lower limbs and the three-dimensional geomagnetic field component signals during the human walking process comprises:

calibrating the sensor through the three-dimensional turntable and the three-dimensional guide rail;

collecting motion signals of lower limbs and three-dimensional geomagnetic field component signals in the walking process of a human body;

denoising the acquired motion signals and the three-dimensional geomagnetic field component signals;

and storing the acquired motion signal and the three-dimensional geomagnetic field component signal into a storage device.

3. The multi-sensor fusion based human gait analysis method according to claim 1, characterized in that after step 400, it further comprises:

step 500, performing standardization processing on the obtained human gait parameters, and further establishing a human gait database, which comprises the following steps:

by the formulaNormalizing the pace by formulaNormalizing the step size by formulaStandardizing the step frequency to obtain the human gait parameters after the standardized treatment, wherein l₁For the height of the subject,. l_mThe average height of the population in the age group of the tested person; v is the pace of the person to be measured obtained by integrating the sensor signal, V_relIs standard pace in meter/second, L is step length obtained by integrating pace_relIs standard step length in meter/step, C is walking frequency of the measured person in unit time calculated by walking period, C_relIs a standard step frequency in steps/second.

4. The method for analyzing human gait based on multi-sensor fusion according to claim 1, characterized in that the two constraints established by the lower limb kinematics model are:

$\∫{\∫}_0^t{a}_k^{G}dt={\[x_k^G,y_k^G,z_k^G\]}^T+{\[x_h^G,y_h^G,z_h^G\]}^T;$

$\∫{\∫}_0^t{a}_a^{G}dt=\∫{\∫}_0^t{a}_k^{G}dt+{\[x_a^G,y_a^G,z_a^G\]}^T;$

wherein,is a knee joint displacement vector based on the calf motion,an ankle position vector calculated for the foot-based sensor,is the hip joint displacement vector calculated by a sensor placed on the thigh part.

5. A multi-sensor fusion based human gait analysis system, characterized in that the multi-sensor fusion based human gait analysis system comprises: the data acquisition device is used for acquiring motion signals of lower limbs and three-dimensional geomagnetic field component signals in the walking process of a human body by using a sensor, and the motion signals comprise three-dimensional acceleration and three-dimensional angular velocity; the data analysis processing device comprises an initial posture analysis unit, a gait parameter calculation unit and an error correction unit;

the system comprises an initial attitude analysis unit, a ground coordinate system and a sensor coordinate system, wherein the initial attitude analysis unit is used for acquiring an initial attitude of a human body according to acquired motion signals and three-dimensional geomagnetic field component signals, acquiring a deviation amount of the sensor coordinate system and the ground coordinate system according to the initial attitude of the human body, correcting a rotation matrix transformed from the sensor coordinate system to the ground coordinate system by using the deviation amount, compensating a pitch angle, a roll angle and a yaw angle and acquiring compensated initial orientation information of the human body;

the gait parameter calculating unit is used for obtaining the starting time of switching the human body from a static standing state to a walking state according to the collected motion signals and the corrected rotation matrix, performing sensor data fusion by using an expanded Kalman filter from the starting time, updating human body position information, and detecting a gait time phase in the walking process of the human body according to the collected motion signals so as to obtain human body gait parameters, wherein the gait time phase comprises a support time phase and a swing time phase, the support time phase comprises a heel hitting period, a standing phase middle period, a full standing period and a heel off period, and the swing time phase comprises an acceleration period, a swing period and a deceleration period; the human body gait parameters comprise the pace, the step length, the step frequency, the walking cycle and the walking track in the human body walking process;

the error correction unit is used for eliminating the error accumulation of the sensor and updating the gait parameters of the human body, and comprises: when the time that the crus are vertical to the ground in the standing phase is detected, the center of gravity of the tested person body moves forwards by taking the grounded legs as swing axes, a first-stage inverted pendulum model is built for the whole person, a zero-speed updating algorithm is executed to eliminate errors, and the obtained gait parameters of the person are updated; when a full standing period is detected, the foot is completely attached to the ground, the crus and the thighs of the foot form a rigid body connected through hinge joints, a lower limb kinematics model is established by using a Dinavier-Hartmanberg method, and a motion signal of the legs and a motion signal of the foot are fused to eliminate errors and update the obtained human gait parameters.

6. The human gait analysis system based on multi-sensor fusion of claim 5, characterized in that the data acquisition device comprises an inertial sensor calibration unit, a sensor signal acquisition unit, a data filtering unit and a self-contained data storage unit;

the inertial sensor calibration unit is used for calibrating the sensor through the three-dimensional turntable and the three-dimensional guide rail;

the sensor signal acquisition unit is used for acquiring a motion signal of the lower limbs and a three-dimensional geomagnetic field component signal in the walking process of the human body;

the data filtering unit is used for denoising the acquired motion signals and the three-dimensional geomagnetic field component signals;

and the self-contained data storage unit is used for storing the acquired motion signals and the three-dimensional geomagnetic field component signals into storage equipment.

7. The multi-sensor fusion based human gait analysis system according to claim 5, characterized in that it further comprises:

the gait database unit is used for carrying out standardized processing on the obtained human gait parameters so as to establish a human gait parameter database, and is particularly used for; by the formulaNormalizing the pace by formulaNormalizing the step size by formulaNormalizing the step frequency, wherein₁For the height of the subject,. l_mThe average height of the population in the age group of the tested person; v is the pace of the person to be measured obtained by integrating the sensor signal, V_relIs a standard pace, in meters per second; l is the step length obtained by integrating the step speed, L_relIs standard step length, and the unit is meter per step; c is the unit time travel of the tested person calculated by the walking cycleStep frequency of walking, C_relIs a standard step frequency in steps/second.

8. The system for analyzing human gait based on multi-sensor fusion according to claim 5, characterized in that in the error correction unit, the two constraints established by the lower limb kinematics model are:

$\∫{\∫}_0^t{a}_k^{G}dt={\[x_k^G,y_k^G,z_k^G\]}^T+{\[x_h^G,y_h^G,z_h^G\]}^T;$

$\∫{\∫}_0^t{a}_a^{G}dt=\∫{\∫}_0^t{a}_k^{G}dt+{\[x_a^G,y_a^G,z_a^G\]}^T;$

wherein,is a knee joint displacement vector based on the calf motion,an ankle position vector calculated for the foot-based sensor,is the hip joint displacement vector calculated by a sensor placed on the thigh part.