CN112440267B - Gait phase identification method based on inertial sensor - Google Patents
Gait phase identification method based on inertial sensor Download PDFInfo
- Publication number
- CN112440267B CN112440267B CN202011367253.4A CN202011367253A CN112440267B CN 112440267 B CN112440267 B CN 112440267B CN 202011367253 A CN202011367253 A CN 202011367253A CN 112440267 B CN112440267 B CN 112440267B
- Authority
- CN
- China
- Prior art keywords
- axis
- moment
- angular velocity
- acquired
- inertial sensor
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/0006—Exoskeletons, i.e. resembling a human figure
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1694—Programme controls characterised by use of sensors other than normal servo-feedback from position, speed or acceleration sensors, perception control, multi-sensor controlled systems, sensor fusion
Landscapes
- Engineering & Computer Science (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
The invention relates to a gait phase identification method based on an inertial sensor, which is used for carrying out extreme value identification by utilizing the body side angular velocity collected by the inertial sensor bound at the ankle. A complete gait cycle can be roughly divided into two phases of pendulum dynamic state and support state, and the support state can be divided into four stages of heel landing, full sole landing, heel lift-off and sole lift-off in sequence. When in the full ball strike stage of the support state, the foot is relatively static with respect to the ground, i.e., the magnitude of the body-side angular velocity at the ankle is nearly zero. By utilizing the characteristic that the angular speed is zero when touching the ground, the calibration result of the pressure insole is compared, the whole supporting state phase can be identified by detecting two maximum values of the input angular speed, and the aims of high accuracy, low time delay, low hardware requirement and easy use can be realized.
Description
Technical Field
The invention relates to a gait recognition method based on an inertial sensor, and belongs to the technical field of gait recognition.
Background
The exoskeleton robot is a set of human-computer system, and various information needs to be interacted between human body movement and exoskeleton movement. The exoskeleton needs to identify the gait of the human body movement and the movement intention of the human body so as to drive the exoskeleton to move. Gait information needs to be analyzed by gait data collected by the sensor system. Gait data acquired by different sensor systems are different, so that gait characteristics are different, and gait recognition methods are different. The main gait data includes joint angle data, acceleration data of limbs, electromyographic signals of relevant muscle groups, brain wave signals, plantar pressure signals, video-based image data, and the like.
Scholars at home and abroad complete a great deal of research on the man-machine interaction control strategy of the lower limb exoskeleton robot. Pappas uses the angular velocity of the foot and plantar pressure signals to detect different phases of the person in motion and proposes a rule-based detection algorithm and the system can identify disturbances when not walking, such as sitting up, etc. Skelly proposes a real-time monitoring system, which includes different levels of signal processing algorithms, such as fuzzy logic algorithm. Han et al propose a non-constrained dynamic detection system, the hardware includes a triaxial acceleration transducer and a video camera; lauer uses an adaptive fuzzy neural network for gait detection; lee integrates data of various sensors such as a gyroscope, an accelerometer and the like to complete gait recognition; havasi takes the initiative of performing gait recognition by adopting video images. Wilcox et al propose a study method based on Electromyogram (EMU) signals, which utilizes electromyogram signals to analyze object motion intention recognition and control a lower limb exoskeleton robot to assist a patient with spinal cord injury to perform rehabilitation exercise training, and is easily influenced by factors such as human body surface sweating. Huo et al propose a novel active impedance control strategy applied to lower limb exoskeleton, design a human body joint torque observer to estimate human body joint torque, construct a human body exoskeleton system time-varying ideal impedance model, reduce the mechanical impedance of the human body exoskeleton system below the athletic ability level of a wearer, and provide power assistance for sitting and standing of the wearer. Longyi et al propose an exoskeleton robot control method for predicting human body movement intention based on Kalman filtering, measure human-computer interaction information by using a torque sensor, compensate intention delay by Kalman filtering, and estimate the movement locus of human body lower limb joints, and the method needs to perform complex parameter optimization on PD control law. The Tpeak and the like provide gait prediction methods based on a gray theory, ankle joints are used as research objects, the spatial position coordinates of the manic joints in a natural walking state are captured through video capture equipment (Kinect), a gray prediction system is used for prediction, the gray prediction system can be used for modeling only by needing few data, and the original data are required to basically meet the feasibility of a gray prediction model.
The prior gait recognition technology has the following problems:
(1) the requirements of gait recognition technologies such as pressure insoles, optical capture and camera equipment, electromechanical or electroencephalogram acquisition equipment and the like on hardware are high, the equipment is inconvenient for users to wear quickly, and some of the equipment are expensive.
(2) Machine learning gait phase identification algorithm problem based on inertial sensor: the machine learning algorithm needs a large amount of training data to improve the performance, so a certain time is needed for collecting a training set for the algorithm, and users with large differences cannot use the algorithm immediately; the generalization capability of the machine learning algorithm is different, and the adaptability to different terrains depends on the quality of a training set; the machine learning algorithm delay is difficult to control in millisecond level, and has hysteresis feeling in actual use. The machine learning algorithm occupies higher computational resources and has certain requirements on the power and the computational power of the processor.
(3) The threshold discrimination gait phase recognition algorithm problem based on the inertial sensor comprises the following steps: the threshold of the threshold discrimination algorithm is difficult to define and needs a large amount of data to provide a basis; the threshold discrimination algorithm needs to continuously adjust parameters aiming at different users and terrains, and the robustness is poor; the threshold discrimination algorithm is easy to have misjudgment and missed judgment aiming at special actions in a gait cycle.
Disclosure of Invention
The technical problem solved by the invention is as follows: the gait phase identification method based on the inertial sensor can achieve the aims of high accuracy, low time delay, low hardware requirement and easiness in use.
The technical scheme of the invention is as follows: a gait phase identification method based on an inertial sensor is characterized by comprising the following steps:
(1) installing inertial sensors on the outer sides of two ankles of a human body respectively, and enabling one axis in the inertial sensors to be perpendicular to the sagittal plane of the human body and defined as a Y axis; completing the installation of the inertial sensor;
(2) outputting the angular speed which is acquired by the two inertial sensors in real time and rotates around the Y axis to a processor through a CAN bus;
(3) the processor compares the angular speed of the inertial sensor, which is acquired at the current moment and rotates around the Y axis, with the angular speed of the inertial sensor, which is acquired at the previous moment and rotates around the Y axis, aiming at the angular speed of the inertial sensor, which is acquired in real time and rotates around the Y axis, and determines the state of the angular speed of the inertial sensor, which is acquired at the current moment and rotates around the Y axis;
(4) comparing the state of the angular speed which is acquired at the current moment and rotates around the Y axis and determined in the step (3) with the state of the angular speed which is acquired at the previous moment and rotates around the Y axis, and judging the wave crest or the wave trough of the angular speed which is acquired at the previous moment and rotates around the Y axis;
(5) setting a first threshold value; if the angular speed collected at the previous moment and rotating around the Y axis is a trough, judging the angular speed collected at the previous moment and rotating around the Y axis, which is judged to be the trough, and if the angular speed collected at the previous moment and rotating around the Y axis is smaller than a set first threshold, judging the angular speed collected at the previous moment and rotating around the Y axis to be a ground contact pre-judgment point; otherwise, recording the angular speed which is acquired at the last moment of the wave trough and rotates around the Y axis;
(6) setting a second threshold value; if the angular velocity of the rotation around the Y axis collected at the previous moment is a wave crest, judging the angular velocity of the rotation around the Y axis collected at the previous moment which is judged as the wave crest, and if the angular velocity of the rotation around the Y axis collected at the previous moment is larger than a set second threshold value and the previous wave trough of the wave crest is a grounding pre-judgment point, judging the angular velocity of the rotation around the Y axis collected at the previous moment is the supporting phase starting moment; otherwise, neglecting the angular velocity which is collected at the last moment of the determined wave crest and rotates around the Y axis, and realizing the determination of the starting moment of the support phase of one foot;
(7) while the step (3) is carried out, the processor compares the other angular velocity which is acquired before the inertial sensor and rotates around the Y axis with the angular velocity which is acquired at the last moment aiming at the angular velocity which is transmitted by the other inertial sensor and acquired in real time and rotates around the Y axis, and determines the state of the angular velocity which is acquired at the current moment and rotates around the Y axis;
(8) comparing the state of the angular speed which is acquired at the current moment and rotates around the Y axis and determined in the step (7) with the state of the angular speed which is acquired at the previous moment and rotates around the Y axis, and judging the wave crest or the wave trough of the angular speed which is acquired at the previous moment and rotates around the Y axis;
(9) setting a first threshold value; if the angular speed collected at the previous moment and rotating around the Y axis is a trough, judging the angular speed collected at the previous moment and rotating around the Y axis, which is judged to be the trough, and if the angular speed collected at the previous moment and rotating around the Y axis is smaller than a set first threshold, judging the angular speed collected at the previous moment and rotating around the Y axis to be a ground contact pre-judgment point; otherwise, recording the angular speed which is acquired at the last moment of the wave trough and rotates around the Y axis;
(10) setting a second threshold value; if the angular velocity of the rotation around the Y axis collected at the previous moment is a wave crest, judging the angular velocity of the rotation around the Y axis collected at the previous moment which is judged as the wave crest, and if the angular velocity of the rotation around the Y axis collected at the previous moment is larger than a set second threshold value and the previous wave trough of the wave crest is a grounding pre-judgment point, judging the angular velocity of the rotation around the Y axis collected at the previous moment is the supporting phase starting moment; otherwise, neglecting the angular velocity of the rotation around the Y axis collected at the last moment of the determined peak; the determination of the starting moment of the other foot supporting phase is realized;
(11) taking the starting moment of the supporting phase of one foot determined in the step (6) as the starting moment of the swinging phase of the other foot; determining the corresponding foot state from the starting time of one foot supporting phase to the starting time of the foot swinging phase of the foot as the foot supporting phase; determining the foot swing phase; the foot state corresponding to the time from the next moment of the starting moment of the supporting phase of the one foot to the previous moment of the starting moment of the foot swinging phase of the foot is judged as the foot supporting phase;
judging the state of the other foot corresponding to the time from the starting moment of the other foot supporting phase to the starting moment of the foot swinging phase of the other foot as the other foot supporting phase; the state of the other foot corresponding to the time from the next moment of the starting moment of the support phase of the other foot to the last moment of the starting moment of the foot swing phase of the other foot is judged as the support phase of the other foot, and the gait phase identification is realized.
Preferably, the angular velocity of the rotation around the Y axis acquired at the current time of the inertial sensor is compared with the angular velocity of the rotation around the Y axis acquired at the previous time, and the state of the angular velocity of the rotation around the Y axis acquired at the current time is determined;
the angular velocity of the rotation around the Y axis acquired at the current moment is greater than the angular velocity of the rotation around the Y axis acquired at the previous moment, and the state of the angular velocity of the rotation around the Y axis acquired at the current moment is an ascending state;
the angular velocity of the rotation around the Y axis acquired at the present time is smaller than the angular velocity of the rotation around the Y axis acquired at the previous time, and the state of the angular velocity of the rotation around the Y axis acquired at the present time is a rising state.
Preferably, the peak is defined as: if the angular velocity w of the rotation around the Y axis acquired at the present momenttIs in a descending state, and the angular velocity w of the rotation around the Y axis acquired at the previous momentt-1Is in the rising state, the angular velocity w of the rotation around the Y axis collected at the previous momentt-1Is the peak.
Preferably, the trough is defined as: if the angular velocity w of the rotation around the Y axis acquired at the present momenttIs in an up state, and the angular velocity w of the rotation around the Y axis acquired at the previous momentt-1Is in a descending state, the angular velocity w of the rotation around the Y axis collected at the previous momentt-1Is a wave trough.
Preferably, the touchdown predetermined threshold is required to be: the threshold is typically the minimum angular velocity of the human body during each gait cycle when walking at low speeds. (preferably 2km/h to 3km/h, the minimum angular velocity in each gait cycle Thp, preferably-90 °/s;)
Preferably, the touchdown determination threshold is required to be: the threshold value is preferably set to Thh 0 °/s
Preferably, the inertial sensor is capable of acquiring at least one dimension of real-time angular velocity;
preferably, the angular velocity range of the inertial sensor should be greater than +/-500 °/s;
preferably, the static drift of the inertial sensor is within 0.05 °/s;
preferably, the inertial sensor communicates at least through a CAN bus or a serial port;
preferably, the sampling frequency of the inertial sensor is above 20Hz at the lowest;
preferably, the working voltage and working current of the inertial sensor should be within the human body safety electricity utilization range.
Compared with the prior art, the invention has the advantages that:
(1) in the invention, the main data processing process of the algorithm (the algorithm is formed from the step (3) to the step (11)) is to compare the angular speed of the inertial sensor at the previous moment with the angular speed of the inertial sensor at the previous moment, and the angular speed is irrelevant to the data precision of the angular speed of the inertial sensor, so that the algorithm has low requirement on the data precision of the inertial sensor, and the requirement on the performance of the sensor is reduced. The effect of the existing algorithm depends on the data precision of the sensor.
(2) In the invention, the angular speed rotating around the Y axis is used as algorithm input in the step (3), and the angular speed rotating around the Y axis is relatively stable in the human body movement process, so that the algorithm has low requirement on the position precision of the inertial sensor, the device can be conveniently and quickly put on and taken off, and the usability of the exoskeleton is improved. The problem that the requirement on the position precision of the sensor is too high in the prior art is solved.
(3) In the invention, because the main data processing process of the algorithm is to compare the angular velocity of the inertial sensor at the previous moment with the angular velocity of the inertial sensor at the previous moment, the algorithm is not influenced by the self accumulated error of the inertial sensor, and the robustness of the algorithm is improved. The problem of angular velocity interference caused by accumulated errors generated by long-term use of the inertial sensor is avoided.
(4) The invention utilizes the characteristic that the angular velocity of each foot in the sagittal plane of the body is zero when different users move at different speeds in different terrains, so that the algorithm is applicable to different users moving at different speeds in different terrains without being respectively adjusted aiming at different users, different speeds or different terrains, thereby improving the robustness and the usability of the algorithm. The existing algorithm mostly needs to make corresponding adjustment according to different situations.
(5) The algorithm is fed back according to experimental data (walking and going upstairs and downstairs) under different conditions, the real-time accuracy rate is stabilized to be more than 95% when the algorithm is continuously used for a long time, and the requirement of high accuracy rate is met. The online accuracy of the existing real-time algorithm is mostly kept about 90%.
(6) Because the time complexity and the space complexity of the algorithm are low, the delay of the preferably used 100Hz sensor can be controlled within 15ms, the delay can be further shortened according to the improvement of the sampling frequency of the inertial sensor, and the high real-time performance is realized.
(7) Because the time complexity and the space complexity of the algorithm are low, the algorithm has low requirements on the performance of the processor (STM 32), more data space and calculation power can be saved, and the endurance and the multitask performance of the whole exoskeleton system are improved.
Drawings
FIG. 1 is a hardware composition diagram of the present invention.
Fig. 2 is a schematic view of the inertial sensor binding position of the present invention.
FIG. 3 is a flow chart of the algorithm of the present invention.
FIG. 4 is a graph showing the algorithm effect of the present invention in running at 10km/h, and the data input and algorithm effect of upstairs at 120 th/min.
Detailed Description
The invention is described in further detail below with reference to the figures and specific embodiments.
The invention relates to a gait phase identification method based on an inertial sensor, which is used for carrying out extreme value identification by utilizing the body side angular velocity collected by the inertial sensor bound at the ankle. A complete gait cycle can be roughly divided into two phases of pendulum dynamic state and support state, and the support state can be divided into four stages of heel landing, full sole landing, heel lift-off and sole lift-off in sequence. When in the full ball strike stage of the support state, the foot is relatively static with respect to the ground, i.e., the magnitude of the body-side angular velocity at the ankle is nearly zero. By utilizing the characteristic that the angular speed is zero when touching the ground, the calibration result of the pressure insole is compared, the whole supporting state phase can be identified by detecting two maximum values of the input angular speed, and the aims of high accuracy, low time delay, low hardware requirement and easy use can be realized.
The method of the invention has been applied to lower limb assistance exoskeleton systems, which aim to load torque on joints at appropriate times to achieve the effect of assisting in movement. The method can provide the gait phase recognition result of the wearer for the lower limb assistance exoskeleton system, so that the lower limb assistance exoskeleton system can determine the moment loading moment and the moment cancelling moment in time, the key problem that the lower limb assistance exoskeleton system is difficult to determine the assistance opportunity is solved, the assistance experience of the wearer is improved, and the accident that the lower limb assistance exoskeleton system causes injury to the user due to wrong moment loading or cancelling moment is avoided.
The preferred scheme of the inertial sensor in the invention is as follows:
the inertial sensor is capable of acquiring at least one dimension of real-time angular velocity.
The angular velocity range of the inertial sensor should be greater than ± 500 °/s.
The static drift of the inertial sensor is within 0.05 DEG/s.
The inertial sensor CAN communicate at least through a CAN bus or a serial port.
Because the delay of the algorithm (the algorithm is formed from the step (3) to the step (11)) is directly related to the sampling frequency of the inertial sensor, if the algorithm delay is Ta, the sampling frequency of the inertial sensor is K, and the code execution time is Tc, then
Ta ═ 1/K) + Tc, where Tc < <1 ms.
Therefore, the delay of the algorithm can be greatly reduced by increasing the sampling frequency of the inertial sensor, and the highest sampling frequency of the inertial sensor is not required; since the frequency of human motion is 15Hz to 20Hz at the fastest, the sampling frequency of the inertial sensor is required to be 20Hz or more at the lowest.
The working voltage and working current of the inertial sensor are within the range of human body safety electricity utilization.
The sensor adopted by the invention CAN measure three-dimensional angular velocity, the angular velocity range is +/-1000 degrees/s, the maximum static drift of the angular velocity is 0.04 degrees/s, the communication CAN be carried out through a CAN bus or a serial port, the sampling frequency is 100Hz, the working voltage is 3.3V-5V, and the working current is less than 25 mA.
The processor of the invention has the preferable scheme that: the processor is at least STM32L series chips and CAN communicate through a CAN bus.
The processor employed in the present invention is preferably STM32F 767.
As shown in fig. 1, the hardware system for implementing the method of the present invention mainly comprises two IMU (inertial sensor) modules, a DSP backplane and an STM32 processor, and the pressure insole in fig. 1 is used as a preferred verification device and can be selected as a system, which is not necessarily configured for the system.
The data acquired by the IMU module is sent to the DSP bottom plate through a serial port; the DSP bottom plate transmits IMU module data to an STM32 processor through a CAN bus to be used as algorithm input; and through the operation of an STM32 processor, an algorithm result is transmitted to an upper computer through a serial port for reference.
The invention relates to a gait phase identification method based on an inertial sensor, which preferably comprises the following steps:
(1) installing inertial sensors on the outer sides of two ankles of a human body respectively, and enabling any axis which can measure angular speed around the axis in the inertial sensors to be vertical to a sagittal plane of the human body and defined as a Y axis; completing the installation of the inertial sensor; as shown in fig. 2, the positive Y-axis direction of the inertial sensor attached to the left foot is directed to the left, and the positive Y-axis direction of the inertial sensor attached to the right foot is directed to the right. The installation mode can be bound by a binding band, adhesive tape and the like. The installation precision should be guaranteed as far as possible that the Y axis is perpendicular to the sagittal plane of the body, but no specific precision requirement exists. The installation strength is required to ensure that the inertial sensor is not connected with the human body and loosened due to the violent movement of the human body;
(2) outputting the angular speed which is acquired by the two inertial sensors in real time and rotates around the Y axis to a processor through a CAN bus at the transmission frequency of 100 Hz; each inertial sensor samples and sends one byte of hexadecimal angular velocity data each time, the processor receives the angular velocity data sent by the inertial sensors through interruption, and an algorithm in the processor converts the received angular velocity data of the inertial sensors into floating point numbers with symbols;
(3) the processor aims at the real-time acquired angular speed which is sent by one of the inertial sensors and rotates around the Y axis and is recorded as wtAs shown in fig. 3, the angular velocity w around the Y axis acquired at the current moment of the inertial sensor is measuredtAngular velocity w of rotation with the Y axis acquired at the previous momentt-1Comparing (in the invention, the initial time angular velocities of the inertial sensors on two feet are preset to be 0), determining the state of the angular velocity collected at the current time and rotating around the Y axis, wherein the state of the angular velocity at the current time is an attribute parameter of the angular velocity at the current time, and can be used for determining the state of the angular velocity at the current timeIs marked as StThe state includes a falling state (S)t1), rising state (S)t2) and horizontal state (S)t0); the descending state means that the angular velocity (w) of the rotation around the Y axis acquired at the current moment is smaller than the angular velocity (w) of the rotation around the Y axis acquired at the previous momentt-1-wt>0) (ii) a The ascending state means that the angular velocity (w) of the rotation around the Y axis acquired at the current moment is greater than the angular velocity (w) of the rotation around the Y axis acquired at the last momentt-1-wt<0) (ii) a The horizontal state means that the angular velocity of the rotation around the Y axis acquired at the current moment is equal to the angular velocity (w) of the rotation around the Y axis acquired at the previous momentt-1-wt0); the state of the initial angular velocity is a horizontal state;
(4) as shown in fig. 3, the state S of the angular velocity about the Y axis acquired at the current time determined in step (3) is settWith the state S of angular velocity about the Y axis acquired at the previous momentt-1The comparison is made, as shown in fig. 3, by taking the state S of the angular velocity about the Y axis acquired at the previous time after the comparison is completedtAssigning a state S to the angular velocity about the Y-axis acquired at the previous momentt-1(ii) a Judging whether the angular speed which is collected at the previous moment and rotates around the Y axis is a wave crest or a wave trough; if the angular velocity w of the rotation around the Y axis acquired at the present momenttIs in a descending state (S)t1) and the angular velocity w of the rotation about the Y axis acquired at the previous momentt-1Is in the rising state (S)t2), the angular velocity w of the rotation about the Y axis acquired at the previous momentt-1Is a wave crest; if the angular velocity w of the rotation around the Y axis acquired at the present momenttIs in the rising state (S)t2) and the angular velocity w of the rotation about the Y axis acquired at the previous momentt-1Is in a descending state (S)t1), the angular velocity w of the rotation about the Y axis acquired at the previous momentt-1Is a wave trough; if the two conditions are not the same, acquiring the state w of the angular velocity rotating around the Y axis at the previous momenttAssigning a state w to the angular velocity of rotation about the Y-axis acquired at the previous momentt-1；
(5) Setting a touchdown predetermined threshold value Thp, which is generally the minimum angular velocity of the human body during each gait cycle when walking at low speed (2km/h-3km/h)And can be determined as Thp-90 °/s by experimental data summarization; as shown in fig. 3, the angular velocity w about the Y-axis, as acquired at the previous timet-1If the angular velocity is a trough, the angular velocity w of the rotation around the Y axis collected at the last moment of the trough is determinedt-1Judging that the angular velocity w of the rotation around the Y axis collected at the previous moment is less than the preset grounding pre-judgment threshold value Thpt-1The method comprises the following steps of (1) recording a touchdown pre-determination point P for serving as one of determination conditions of a support phase starting moment; otherwise, the state w of the angular velocity of rotation around the Y axis acquired at the previous momenttAssigning a state w to the angular velocity of rotation about the Y-axis acquired at the previous momentt-1；
(6) Setting a touchdown determination threshold Thh, which is usually set to 0 °/s; as shown in fig. 3, the angular velocity w about the Y-axis, as acquired at the previous timet-1If the peak is determined, the angular velocity w of the rotation around the Y axis collected at the previous moment of the peak is determinedt-1Judging, if the judgment result is larger than a set touchdown judgment threshold value Thh and the last trough of the wave crest is a touchdown pre-judgment point P, judging the angular speed w collected at the last moment and rotating around the Y axist-1For supporting the phase start time and acquiring the state w of angular velocity around the Y axistAssigning a state w to the angular velocity of rotation about the Y-axis acquired at the previous momentt-1(ii) a Otherwise, the state w of the angular velocity of rotation around the Y axis acquired at the previous momenttAssigning a state w to the angular velocity of rotation about the Y-axis acquired at the previous momentt-1The determination of the starting moment of the supporting phase of one foot is realized;
(7) while the step (3) is carried out, the processor compares the angular speed of the other inertial sensor which is acquired in real time and rotates around the Y axis with the angular speed of the Y axis which is acquired at the previous moment, and determines the state of the angular speed of the current inertial sensor which rotates around the Y axis; the state of the angular velocity at the present time is an attribute parameter of the angular velocity at the present time, and can be recorded as Sother_tThe state includes a falling state (S)other_t＝1)Rising state (S)other_t2) and horizontal state (S)other_t0); the descending state means that the angular velocity (w) of the rotation around the Y axis acquired at the current moment is smaller than the angular velocity (w) of the rotation around the Y axis acquired at the previous momentother_t-1-wother_t>0) (ii) a The ascending state means that the angular velocity (w) of the rotation around the Y axis acquired at the current moment is greater than the angular velocity (w) of the rotation around the Y axis acquired at the last momentother_t-1-wother_t<0) (ii) a The horizontal state means that the angular velocity of the rotation around the Y axis acquired at the current moment is equal to the angular velocity (w) of the rotation around the Y axis acquired at the previous momentother_t-1-wother_t0); the state of the initial angular velocity is a horizontal state;
(8) the state S of the angular speed which is collected at the current moment and rotates around the Y axis and is determined in the step (7)other_tWith the state S of angular velocity about the Y axis acquired at the previous momentother_t-1Comparing the states S of the angular speeds of the rotation around the Y axis collected at the previous momentother_tAssigning a state S to the angular velocity about the Y-axis acquired at the previous momentother_t-1(ii) a Determining the angular velocity w of the rotation about the Y-axis acquired at the previous momentother_t-1Whether it is a peak or a trough; if the angular velocity w of the rotation around the Y axis acquired at the present momenttIs in a descending state (S)other_t1) and the angular velocity w of the rotation about the Y axis acquired at the previous momentother_t-1Is in the rising state (S)other_t2), the angular velocity w of the rotation about the Y axis acquired at the previous momentother_t-1Is a wave crest; if the angular velocity w of the rotation around the Y axis acquired at the present momentother_tIs in the rising state (S)other_t2) and the angular velocity w of the rotation about the Y axis acquired at the previous momentother_t-1Is in a descending state (S)other_t1), the angular velocity w of the rotation about the Y axis acquired at the previous momentother_t-1Is a wave trough; if the two conditions are not the same, acquiring the state w of the angular velocity rotating around the Y axis at the previous momentother_tAssigning a state w to the angular velocity of rotation about the Y-axis acquired at the previous momentother_t-1；
(9) Setting a predetermined touchdown threshold value ThpotherThe threshold value is onThe minimum angular velocity in each gait cycle is often determined as Thp through experimental data summarization when the human body walks at low speed (2km/h-3km/h)other-90 °/s; angular velocity w of rotation about the Y-axis, as acquired at the previous momentother_t-1If the angular velocity is a trough, the angular velocity w of the rotation around the Y axis collected at the last moment of the trough is determinedother_t-1Making a judgment, and if the judgment is less than the set grounding judgment threshold value ThpotherThen the angular speed w of the rotation around the Y axis collected at the last moment is determinedother_t-1Prejudging point P for touchdownotherAnd recording the touchdown predetermination point PotherTouchdown predetermined decision point PotherIs used as one of the judgment conditions of the starting time of the support phase; otherwise, the state w of the angular velocity of rotation around the Y axis acquired at the previous momentother_tAssigning a state w to the angular velocity of rotation about the Y-axis acquired at the previous momentother_t-1；
(10) Setting a touchdown determination threshold ThhotherThe threshold is usually set to Thh other0 °/s; angular velocity w of rotation about the Y-axis, as acquired at the previous momentother_t-1If the peak is determined, the angular velocity w of the rotation around the Y axis collected at the previous moment of the peak is determinedother_t-1Judging if the contact pressure is higher than a set contact pressure judgment threshold value ThhotherAnd the last wave trough of the wave crest is a grounding pre-determination point PotherThen the angular speed w of the rotation around the Y axis collected at the last moment is determinedother_t-1For supporting the phase start time and acquiring the state w of angular velocity around the Y axisother_tAssigning a state w to the angular velocity of rotation about the Y-axis acquired at the previous momentother_t-1(ii) a Otherwise, the state w of the angular velocity of rotation around the Y axis acquired at the previous momentother_tAssigning a state w to the angular velocity of rotation about the Y-axis acquired at the previous momentother_t-1(ii) a The determination of the starting moment of the other foot supporting phase is realized;
(11) taking the starting moment of the supporting phase of one foot determined in the step (6) as the starting moment of the swinging phase of the other foot; determining the corresponding foot state from the starting time of one foot supporting phase to the starting time of the foot swinging phase of the foot as the foot supporting phase; determining the foot swing phase; the foot state corresponding to the time from the next moment of the starting moment of the supporting phase of the one foot to the previous moment of the starting moment of the foot swinging phase of the foot is judged as the foot supporting phase;
judging the state of the other foot corresponding to the time from the starting moment of the other foot supporting phase to the starting moment of the foot swinging phase of the other foot as the other foot supporting phase; the state of the other foot corresponding to the time from the next moment of the starting moment of the support phase of the other foot to the last moment of the starting moment of the foot swing phase of the other foot is judged as the support phase of the other foot, and the gait phase identification is realized.
The invention realizes a further optimized scheme of algorithm delay reduction, which is as follows:
if the calculation method delay is Ta, the sampling frequency of the inertial sensor is K, and the code execution time length is Tc, then Ta is (1/K) + Tc,
according to the formula, the main parameters influencing the algorithm delay Ta are that the sampling frequency of the inertial sensor is K and the code execution time is Tc. Because the complexity of the algorithm is low, the code execution time Tc is usually much less than 1ms, while the 1/K of the invention is 10ms, so that increasing the sampling frequency of the inertial sensor to K can further reduce the algorithm delay.
The invention realizes a further preferable scheme for improving the identification accuracy, which comprises the following steps:
if the angular speed static drift amount of the inertial sensor is E and the working time of the inertial sensor is Tw, the optimal constraint conditions are as follows: when Tw is 60, | E2*Tw|<1, satisfying the optimal condition, the recognition accuracy can be further improved.
The invention has been tested, and the test is divided into two parts of walking test and stair test. The subject was a 30-year-old healthy male 170cm in height. To test the effect of the present invention in fast exercise (flat running at 8-10km/h, 100-.
The test effect of the invention is shown in fig. 4, the recognition accuracy rate is 99.8% within 3min of continuous running at the speed of 10km/h, and the algorithm delay is stable between 11ms and 15 ms. The identification accuracy rate is 99.1% within 3min when the user climbs the stairs at the speed of 120/min, and the algorithm delay is stable between 11ms and 15 ms.
In the invention, the main data processing process of the algorithm is to compare the angular velocity of the inertial sensor at the previous moment with the angular velocity of the inertial sensor at the previous moment, and the angular velocity is irrelevant to the data precision of the angular velocity of the inertial sensor, so the algorithm has low requirement on the data precision of the inertial sensor, and the requirement on the performance of the sensor is reduced. The effect of the existing algorithm mostly depends on the data precision of the sensor; in the invention, the angular speed rotating around the Y axis is used as algorithm input in the step (3), and the angular speed rotating around the Y axis is relatively stable in the human body movement process, so that the algorithm has low requirement on the position precision of the inertial sensor, the device can be conveniently and quickly put on and taken off, and the usability of the exoskeleton is improved. The problem that the requirement on the position precision of the sensor is too high in the prior art is solved; in the invention, because the main data processing process of the algorithm is to compare the angular velocity of the inertial sensor at the previous moment with the angular velocity of the inertial sensor at the previous moment, the algorithm is not influenced by the self accumulated error of the inertial sensor, and the robustness of the algorithm is improved. The problem of angular velocity interference caused by accumulated errors generated by long-term use of the inertial sensor is avoided.
The invention utilizes the characteristic that the angular velocity of each foot in the sagittal plane of the body is zero when different users move at different speeds in different terrains, so that the algorithm is applicable to different users moving at different speeds in different terrains without being respectively adjusted aiming at different users, different speeds or different terrains, thereby improving the robustness and the usability of the algorithm. The existing algorithm mostly needs to make corresponding adjustment aiming at different situations; the algorithm is fed back according to experimental data (walking and going upstairs and downstairs) under different conditions, the real-time accuracy rate is stabilized to be more than 95% when the algorithm is continuously used for a long time, and the requirement of high accuracy rate is met. The online accuracy of the existing real-time algorithm is mostly kept about 90%.
Because the time complexity and the space complexity of the algorithm are low, the delay of the preferably used 100Hz sensor can be controlled within 15ms, the delay can be further shortened according to the improvement of the sampling frequency of the inertial sensor, and the high real-time performance is realized; because the time complexity and the space complexity of the algorithm are low, the algorithm has low requirements on the performance of the processor (STM 32), more data space and calculation power can be saved, and the endurance and the multitask performance of the whole exoskeleton system are improved.
Claims (10)
1. A gait phase identification method based on an inertial sensor is characterized by comprising the following steps:
(1) installing inertial sensors on the outer sides of two ankles of a human body respectively, and enabling one axis in the inertial sensors to be perpendicular to the sagittal plane of the human body and defined as a Y axis; completing the installation of the inertial sensor;
(2) outputting the angular speed which is acquired by the two inertial sensors in real time and rotates around the Y axis to a processor through a CAN bus;
(3) the processor compares the angular speed of the inertial sensor, which is acquired at the current moment and rotates around the Y axis, with the angular speed of the inertial sensor, which is acquired at the previous moment and rotates around the Y axis, aiming at the angular speed of the inertial sensor, which is acquired in real time and rotates around the Y axis, and determines the state of the angular speed of the inertial sensor, which is acquired at the current moment and rotates around the Y axis;
(4) comparing the state of the angular speed which is acquired at the current moment and rotates around the Y axis and determined in the step (3) with the state of the angular speed which is acquired at the previous moment and rotates around the Y axis, and judging the wave crest or the wave trough of the angular speed which is acquired at the previous moment and rotates around the Y axis;
(5) setting a first threshold value; if the angular speed collected at the previous moment and rotating around the Y axis is a trough, judging the angular speed collected at the previous moment and rotating around the Y axis, which is judged to be the trough, and if the angular speed collected at the previous moment and rotating around the Y axis is smaller than a set first threshold, judging the angular speed collected at the previous moment and rotating around the Y axis to be a ground contact pre-judgment point; otherwise, recording the angular speed which is acquired at the last moment of the wave trough and rotates around the Y axis;
(6) setting a second threshold value; if the angular velocity of the rotation around the Y axis collected at the previous moment is a wave crest, judging the angular velocity of the rotation around the Y axis collected at the previous moment which is judged as the wave crest, and if the angular velocity of the rotation around the Y axis collected at the previous moment is larger than a set second threshold value and the previous wave trough of the wave crest is a grounding pre-judgment point, judging the angular velocity of the rotation around the Y axis collected at the previous moment is the supporting phase starting moment; otherwise, neglecting the angular velocity which is collected at the last moment of the determined wave crest and rotates around the Y axis, and realizing the determination of the starting moment of the support phase of one foot;
(7) while the step (3) is carried out, the processor compares the other angular velocity which is acquired before the inertial sensor and rotates around the Y axis with the angular velocity which is acquired at the last moment aiming at the angular velocity which is transmitted by the other inertial sensor and acquired in real time and rotates around the Y axis, and determines the state of the angular velocity which is acquired at the current moment and rotates around the Y axis;
(8) comparing the state of the angular speed which is acquired at the current moment and rotates around the Y axis and determined in the step (7) with the state of the angular speed which is acquired at the previous moment and rotates around the Y axis, and judging the wave crest or the wave trough of the angular speed which is acquired at the previous moment and rotates around the Y axis;
(9) setting a first threshold value; if the angular speed collected at the previous moment and rotating around the Y axis is a trough, judging the angular speed collected at the previous moment and rotating around the Y axis, which is judged to be the trough, and if the angular speed collected at the previous moment and rotating around the Y axis is smaller than a set first threshold, judging the angular speed collected at the previous moment and rotating around the Y axis to be a ground contact pre-judgment point; otherwise, recording the angular speed which is acquired at the last moment of the wave trough and rotates around the Y axis;
(10) setting a second threshold value; if the angular velocity of the rotation around the Y axis collected at the previous moment is a wave crest, judging the angular velocity of the rotation around the Y axis collected at the previous moment which is judged as the wave crest, and if the angular velocity of the rotation around the Y axis collected at the previous moment is larger than a set second threshold value and the previous wave trough of the wave crest is a grounding pre-judgment point, judging the angular velocity of the rotation around the Y axis collected at the previous moment is the supporting phase starting moment; otherwise, neglecting the angular velocity of the rotation around the Y axis collected at the last moment of the determined peak; the determination of the starting moment of the other foot supporting phase is realized;
(11) taking the starting moment of the supporting phase of one foot determined in the step (6) as the starting moment of the swinging phase of the other foot; determining the corresponding foot state from the starting time of one foot supporting phase to the starting time of the foot swinging phase of the foot as the foot supporting phase; determining the foot swing phase; the foot state corresponding to the time from the next moment of the starting moment of the supporting phase of the one foot to the previous moment of the starting moment of the foot swinging phase of the foot is judged as the foot supporting phase;
judging the state of the other foot corresponding to the time from the starting moment of the other foot supporting phase to the starting moment of the foot swinging phase of the other foot as the other foot supporting phase; the state of the other foot corresponding to the time from the next moment of the starting moment of the support phase of the other foot to the last moment of the starting moment of the foot swing phase of the other foot is judged as the support phase of the other foot, and the gait phase identification is realized.
2. The gait phase recognition method based on the inertial sensor as claimed in claim 1, characterized in that: comparing the angular speed of the inertial sensor, which is collected at the current moment and rotates around the Y axis, with the angular speed of the inertial sensor, which is collected at the previous moment and rotates around the Y axis, and determining the state of the angular speed of the inertial sensor, which is collected at the current moment and rotates around the Y axis;
the angular velocity of the rotation around the Y axis acquired at the current moment is greater than the angular velocity of the rotation around the Y axis acquired at the previous moment, and the state of the angular velocity of the rotation around the Y axis acquired at the current moment is an ascending state;
the angular velocity of the rotation around the Y axis acquired at the present time is smaller than the angular velocity of the rotation around the Y axis acquired at the previous time, and the state of the angular velocity of the rotation around the Y axis acquired at the present time is a rising state.
3. The gait phase recognition method based on the inertial sensor as claimed in claim 1, characterized in that: the peak, defined as: if the angular velocity w of the rotation around the Y axis acquired at the present momenttIs in a descending state, and the angular velocity w of the rotation around the Y axis acquired at the previous momentt-1Is in the rising state, the angular velocity w of the rotation around the Y axis collected at the previous momentt-1Is the peak.
4. The gait phase recognition method based on the inertial sensor as claimed in claim 1, characterized in that: a trough, defined as: if the angular velocity w of the rotation around the Y axis acquired at the present momenttIs in an up state, and the angular velocity w of the rotation around the Y axis acquired at the previous momentt-1Is in a descending state, the angular velocity w of the rotation around the Y axis collected at the previous momentt-1Is a wave trough.
5. The gait phase recognition method based on the inertial sensor as claimed in claim 1, characterized in that: the touchdown pre-determined threshold is required to be: the threshold is typically the minimum angular velocity of the human body during each gait cycle when walking at low speeds.
6. The gait phase recognition method based on the inertial sensor as claimed in claim 1, characterized in that: touchdown determination threshold, the requirements are: the threshold value is preferably set to Thh ═ 0 °/s.
7. The gait phase recognition method based on the inertial sensor as claimed in claim 1, characterized in that: the inertial sensor is capable of acquiring at least one dimension of real-time angular velocity.
8. The gait phase recognition method based on the inertial sensor as claimed in claim 1, characterized in that: the angular speed range of the inertial sensor is more than +/-500 degrees/s; the static drift of the inertial sensor is within 0.05 DEG/s.
9. The gait phase recognition method based on the inertial sensor as claimed in claim 1, characterized in that: the inertial sensor communicates at least through a CAN bus or a serial port.
10. The gait phase recognition method based on the inertial sensor as claimed in claim 1, characterized in that: the lowest sampling frequency of the inertial sensor is above 20 Hz; the working voltage and working current of the inertial sensor are within the range of human body safety electricity utilization.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011367253.4A CN112440267B (en) | 2020-11-27 | 2020-11-27 | Gait phase identification method based on inertial sensor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011367253.4A CN112440267B (en) | 2020-11-27 | 2020-11-27 | Gait phase identification method based on inertial sensor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112440267A CN112440267A (en) | 2021-03-05 |
| CN112440267B true CN112440267B (en) | 2022-04-22 |
Family
ID=74738221
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011367253.4A Active CN112440267B (en) | 2020-11-27 | 2020-11-27 | Gait phase identification method based on inertial sensor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112440267B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104729507A (en) * | 2015-04-13 | 2015-06-24 | 大连理工大学 | Gait recognition method based on inertial sensor |
| CN106618979A (en) * | 2017-03-03 | 2017-05-10 | 杭州福祉医疗器械有限公司 | Gait rehabilitation robot and method for controlling gait rehabilitation robot |
| CN109540133A (en) * | 2018-09-29 | 2019-03-29 | 中国科学院自动化研究所 | Adaptive gait division methods, system based on micro- inertial technology |
| CN109998551A (en) * | 2019-04-11 | 2019-07-12 | 北京航空航天大学 | A kind of gait phase analysis method of segmented local peak detection |
| JP2020110332A (en) * | 2019-01-11 | 2020-07-27 | 株式会社Ａｔｏｕｎ | Movement support apparatus and movement support method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102098986B (en) * | 2008-07-23 | 2015-09-09 | 伯克利仿生技术公司 | The method of ectoskeleton and this ectoskeletal lower limb of taking a step of control |
-
2020
- 2020-11-27 CN CN202011367253.4A patent/CN112440267B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104729507A (en) * | 2015-04-13 | 2015-06-24 | 大连理工大学 | Gait recognition method based on inertial sensor |
| CN106618979A (en) * | 2017-03-03 | 2017-05-10 | 杭州福祉医疗器械有限公司 | Gait rehabilitation robot and method for controlling gait rehabilitation robot |
| CN109540133A (en) * | 2018-09-29 | 2019-03-29 | 中国科学院自动化研究所 | Adaptive gait division methods, system based on micro- inertial technology |
| JP2020110332A (en) * | 2019-01-11 | 2020-07-27 | 株式会社Ａｔｏｕｎ | Movement support apparatus and movement support method |
| CN109998551A (en) * | 2019-04-11 | 2019-07-12 | 北京航空航天大学 | A kind of gait phase analysis method of segmented local peak detection |
Non-Patent Citations (1)
| Title |
|---|
| 基于惯性测量器件的无线步态分析平台;姜鸣等;《大连理工大学学报》;20161231(第05期);第81-87页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112440267A (en) | 2021-03-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xu et al. | Real-time on-board recognition of continuous locomotion modes for amputees with robotic transtibial prostheses | |
| Zheng et al. | Gait phase estimation based on noncontact capacitive sensing and adaptive oscillators | |
| Stolyarov et al. | Translational motion tracking of leg joints for enhanced prediction of walking tasks | |
| CN112754468B (en) | Human body lower limb movement detection and identification method based on multi-source signals | |
| CN108836757A (en) | A kind of assisted walk exoskeleton robot system with self-regulation | |
| CN112842824B (en) | Training method for rehabilitation and recovery of lower limbs | |
| CN112336590A (en) | Power-assisted exoskeleton movement intention and gait planning method based on multi-sensing information | |
| CN111506189B (en) | Motion mode prediction and switching control method for complex motion of human body | |
| Song et al. | Adaptive neural fuzzy reasoning method for recognizing human movement gait phase | |
| CN102885661A (en) | Dynamic prosthetic knee joint | |
| CN109938892A (en) | A kind of intelligent lap artificial limb wearer riding condition identifying system | |
| CN113143256A (en) | Gait feature extraction method, lower limb assessment method, lower limb control method, device and medium | |
| Ye et al. | An adaptive method for gait event detection of gait rehabilitation robots | |
| CN111568700A (en) | Gait control method, device and equipment for lower limb wearable robot | |
| CN112440267B (en) | Gait phase identification method based on inertial sensor | |
| CN116018119A (en) | System and process for gait phase estimation of a wearable robot | |
| CN111469117B (en) | Human motion mode detection method of rigid-flexible coupling active exoskeleton | |
| CN114224326B (en) | Wearable gait phase and motion recognition device and method | |
| CN110403737A (en) | A kind of knee joint artificial limb and its speed change ambulation control method and system | |
| CN114948579B (en) | Ankle exoskeleton and power-assisted control method and device thereof, and readable storage medium | |
| CN110558990B (en) | Gait analysis method and device | |
| Zhang et al. | Real-Time Assistive Control via IMU Locomotion Mode Detection in a Soft Exosuit: An Effective Approach to Enhance Walking Metabolic Efficiency | |
| Zheng et al. | An automatic labeling strategy for locomotion mode recognition with robotic transtibial prosthesis | |
| Ling et al. | A domain adaptive convolutional neural network for sEMG-based gait phase recognition against to speed changes | |
| CN110555429A (en) | Human motion mode identification method for hip joint exoskeleton |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |