KR20160089791A - System and method for recognizing gait phase - Google Patents

Abstract

According to the present invention, a system for recognizing a gait phase comprises: a gait information input means measuring gait information of a user by using a sensor mounted in the shape of the external skeleton in an external side of a leg of the user; a gait information treatment means recognizing gait information of the user measured in the sensor, and predicting a gait phase by using neural network; and a gait information output means outputting the gait phase. As stated above, the system is basically attached to a robot in the shape of the external skeleton of a leg to substitute a function of an expensive sensor installed in an environment by using an output of a sensor used to control the robot.

Description

[0001] SYSTEM AND METHOD FOR RECOGNITION GAITING [0002]

The present invention relates to a gait step recognition system and method, and more particularly, to a gait recognition system and method, and more particularly, to a gait recognition system and method using a sensor of a foot exoskeletal robot to acquire ground reaction force, joint angular velocity, The present invention relates to a technique of recognizing gait information during a gait operation of a robot in real time by learning and fusing information through a neural network and outputting the gait information in the form of 1 (angular period, Stance), 0 (stance, Swing).

In general, human walking can be described by temporal parameters and spatial parameters. Among the temporal parameters, the walking phase can be divided into a stance, in which one leg supports the body, and a swing, in which one leg moves forward. This information describes the walking of the human being, It can be thought of as important information for judging the presence or absence. In addition, the information about the walking phase is also necessary for controlling the robot system that simulates or assists human walking.

In order to accurately acquire information about the present walking step, expensive optical motion capture equipment installed in a space such as a walking analysis room, and force measuring platforms (Force Platforms) are used.

Referring to FIG. 1, the most widely used method for detecting a walking step is a three-dimensional motion analysis system. The infrared camera 12 is used in synchronization with a Force Platform 14, It uses kinematic data such as trajectory.

However, the method using the 3D motion analysis system is advantageous in that accurate gait detection and various gait analysis can be performed. However, the apparatus is very expensive and can be used only in a laboratory equipped with equipment. And it is possible to extract only data of one or two walking cycles within the effective space of the camera 12.

In addition, when the force measurement plate 14 used in synchronism with the three-dimensional motion analysis system is used, there is a limitation in the measurable walking period according to the expensive price and the number of sensors. Therefore, it is difficult to apply this method to hemiplegic patients with relatively short step length compared with normal subjects.

Korean Patent Publication No. 10-2005-0021288

Therefore, the conventional walking step recognizing methods have a problem that expensive sensors are required, only walking can be recognized in an environment equipped with a specific sensor, or the recognition rate is significantly lowered. The object of the present invention is to solve the problems of the prior art, and it is an object of the present invention to propose a method for acquiring general sensor information attached to a robot having an exoskeletal lower limb, The present invention provides a gait step recognition system and method in which a recognizer using a network is learned to develop a recognizer, and robot sensor information during gait is input to the developed recognizer to output a current gait step in real time.

According to an aspect of the present invention, a walking step recognition system includes walking information input means for measuring walking information of a user using a sensor mounted in an exoskeleton form outside the user's lower back, Walking information processing means for recognizing the walking information of the user measured from the sensor and predicting the walking step using the neural network, and walking information output means for outputting the walking step.

According to another aspect of the present invention, a gait step recognition method of the present invention includes steps of measuring a ground reaction force, estimating whether a ground contact is made, estimating angles of a hip joint and a knee joint, estimating angular velocities of the hip joint and the knee joint Measuring the angular velocity and acceleration of the body part, estimating the body part posture, and driving the recognizer using the ground contact information, the angular velocity information, and the body part attitude information as input parameters, And outputting the target to one of a stance or a swinging phase.

As described above, according to the configuration of the present invention, not only the function of an expensive sensor installed in the environment is replaced by using only the output of a sensor that is basically attached to a robot having an exoskeletal shape and used for controlling the robot, This information can be used to control the robot accurately.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a diagram showing an optical motion capturing apparatus according to the prior art and a walking step recognizing method using a force measuring plate. FIG.
2 is a system block diagram showing a configuration for recognizing a walking step using a neural network according to the present invention.
Fig. 3 is a walking step table showing the target output according to the present invention. Fig.
FIG. 4 is a flow chart for recognizing a walking step through a sensor output value and an acquisition process according to the present invention; FIG.

Brief Description of the Drawings The advantages and features of the present invention, and how to achieve them, will become apparent with reference to the embodiments described in detail below with reference to the accompanying drawings. The present invention may, however, be embodied in many different forms and should not be construed as being limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. Is provided to fully convey the scope of the invention to those skilled in the art, and the invention is only defined by the scope of the claims. The dimensions and relative sizes of layers and regions in the figures may be exaggerated for clarity of illustration. Like reference numerals refer to like elements throughout the specification.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, preferred embodiments of a walking stage recognition system according to the present invention will be described in detail with reference to the accompanying drawings.

By measuring the interaction between the user and the lower extremity skeleton robot using the walking phase recognition system of the present invention, the user can recognize the walking phase, and can directly control the lower extremity skeleton robot based on the walking information, And the walking can be strengthened or suppressed.

2, the walking step recognition system 100 includes walking information input means 110 for measuring the user's walking information using a sensor mounted in the form of an exoskeleton outside the user's lower extremity, A walking information processing unit 120 for recognizing the user's walking information and predicting the walking step using the neural network, and a walking information output unit 130 for outputting the walking step.

Each sensor used in the present invention can be mounted on the pelvis or leg of a user (human, animal, or robot) in the form of an exoskeleton robot. Each sensor can be attached to the outside of the user's lower limb in the form of an exoskeleton, and can have a similar shape to that of the lower limbs.

The sensors used in the present invention include a pair of left and right ground reaction force measurement sensors 112 mounted on the foot of the user, i.e., a portion contacting the ground, for measuring the foot pressure, and mounted on joints (hip and knee joints) A pair of hinge / knee angle measurement sensors 114 and 116, and a pair of body inertia measurement sensors 118 attached to the body part of the user.

The body inertia measurement sensor 118 can measure a body frame of three axes and a global frame of three axes and measures the inertia of the hip joint and the knee roll ), And pitch information. Therefore, the gyroscope of the interbody inertia measurement sensor 118 and the accelerometer output can be fused with other sensors 114 and 116 to estimate the roll and pitch information in the three-dimensional attitude of the trunk portion.

For example, the floor reaction force measuring sensor 112 may be used to measure the foot floor reaction force to estimate whether or not the foot floor is in contact with the ground. Further, the acceleration or angular velocity of the trunk portion is measured using the body inertia measurement sensor 118, and the angular velocity of the hip / knee joint is measured using the hip joint / knee joint angle sensors 114 and 116, Can be estimated.

Also, the acceleration and angular velocity of the trunk measured by the trunk part inertia measurement sensor 118 are fused with the angle of the hip / knee joint measured by the hip joint / knee joint angle measuring sensors 114 and 116 to calculate the hip joint pitch information, The pitch information can be obtained, and as a result, the pelvic posture of the trunk portion can be estimated.

As described above, among the three-dimensional body posture of the lower extremity skeleton robot, the pitch information is measured by being fused with the joint angle sensors 114 and 116 mounted on the joints of the lower extremity skeleton robot, and the pitch (Pitch) information.

The walking information processing means 120 uses a neural network. The neural network used to construct the recognizer in the present invention consists of an input layer having five input nodes, a hidden layer having at least 20 intermediate nodes, and an output layer having at least one output node, The values of the sensors 112 to 118 are inputted, and the walking step information at this time is learned as a desired output.

For example, the five input nodes are:

First, the foot surface contact pressure parameter X1 of the foot exoskeletal robot can be specified and input using the ground reaction force measurement sensor 112. [

Second, the hip joint angular velocity parameter X2 can be specified and input using the hip joint angle sensor 114. [

Third, the knee angular velocity parameter X3 can be specified and input using the knee angle measurement sensor 116. [

Fourth, the hip joint posture, that is, the hip joint pitch information parameter X4, can be specified and input by fusing the hip joint angular velocity information and the body part pelvic posture information.

Fifth, the knee joint posture, that is, the knee pitch information parameter X5, can be specified and input by fusing the knee angular velocity information and the trunk pelvis posture information.

3, the output values of the sensors used in the present invention are input to the recognizer, and the walking information output means 130 outputs the current walking step information as 1 (stance phase) , Or 0 (swing phase, Swing Phase). The walking phase of the present invention is largely classified into a stance phase and a swing phase depending on the presence or absence of a ground contact. The stance frame includes an initial contact, a loading response, a mid stance, a terminal stance, and a pre-swing. The swinging period includes an initial swing, a mid swing, and a terminal swing.

Or current gait information can be classified into weight acceptance, single limb support, and lim advaccement. In this case, the initial contact and the loading response correspond to weight acceptance, and the mid stance, the terminal stance, and the pre- swing is a single limb support with only one foot and includes pre-swing, initial swing, mid swing, and end limb terminal swing corresponds to limb advaccement.

1. Heel contact (initial contact)

In the initial stance phase, with the heel touching the ground, the hip flexes, the knee flexes, and the heel touches the ground. At this time, the other foot is in the late phase.

2. Folding soles (loading response)

As weight shifts forward, the hip does not change, but the knee flexes slightly, and the heel touches the ground and then begins to hit the soles. At this time, the other foot is in the stage where the foot ending motion is performed.

3. The mid-stance phase (mid stance)

The hip is changed from a bent state to an extended state, the knee joint is unchanged, and the center of gravity of the body passes through the center of the sole. At this time, the other foot is in the middle period.

4. The end of the stance phase (terminal stance)

The hip and knee joints gradually stay open and the heel falls. At this time, the other foot is in preparation to reach the heel.

5. The entire swing phase (pre-swing)

With the hip and knee gradually bent, the tip of the toe falls from the ground. At this time, the other foot is in the step of reaching the soles.

6. Initial swing

The hip and knee joints are bent, and the feet are swinging in the air. At this time, the other foot is in the early stage of the middle stance phase.

7. respective medium-term oil (mid swing)

The hip and knee joints gradually change from being bent over the centerline of the body to the spreading motion, and the foot is still in the air. At this time, the other foot is at the last stage of the middle stance phase.

8. The late swing phase (terminal swing)

The hip remains slightly bent, but the knee remains fully open and the foot is in contact with the ground. At this time, the other foot is in the late phase.

Hereinafter, a walking step recognizing method according to the present invention will be described with reference to FIG.

The ground repulsion force is measured (S10), and whether the ground contact is estimated (S20).

The angle of the hip joint, and the angle of the knee joint are respectively measured (S30), and the angular velocity of the hip joint and the knee joint are respectively estimated (S40).

The angular velocity and acceleration of the body part are measured (S50), and the body part posture is estimated (S60).

The hip joint, and the knee joint angular velocity information and the body posture information to estimate the hip joint pitch information and the knee pitch information (S70)

(Step S80). In step S80, the controller inputs the ground contact information, the angular velocity information, the body part attitude information, the hip joint pitch information, and the knee joint pitch information as input parameters.

As described above, the present invention combines values measured by a sensor to extract upper-level information, uses the extracted upper-level information as an input value of a neuron network, and then outputs a value output through a neural network And a step of recognizing a more accurate gait step by using the present invention. Many other modifications will be possible to those skilled in the art, within the scope of the basic technical idea of the present invention.

100: walking step recognition system 110: walking information input means
112: ground reaction force measuring sensor 114: hip joint angle measuring sensor
116: Knee joint angle measurement sensor 118: Sieve inertia measurement sensor
120: walking information processing means 130: walking information output means

Claims (10)

Walking information input means for measuring the walking information of the user by using a sensor mounted in the form of an exoskeleton outside the user's lower back;
Walking information processing means for recognizing the walking information of the user measured from the sensor and predicting a walking step using a neural network; And
And a walking information outputting means for outputting the walking step. The method according to claim 1,
The user may include a person, an animal, or a robot,
Wherein the sensor is mounted on the person, the animal, or the pelvis or the leg of the robot in the form of a base exoskeleton robot. 3. The method of claim 2,
The sensor includes:
A ground reaction force measuring sensor mounted on the foot of the user;
A hip / knee joint angle sensor mounted on the pelvis or the leg of the user; And
And a body trunk inertia measurement sensor mounted on the body part of the user. The method of claim 3,
Wherein the ground surface reaction force measuring sensor measures the ground contact pressure of the foot and the body inertia measuring sensor measures an acceleration or an angular velocity of the body part and the hip joint / Respectively,
And acquiring the hip joint / knee pitch information by fusing acceleration or angular velocity of the body part measured by the body inertia measurement sensor with the angle of the hip joint / knee joint measured by the hip joint / knee joint angle sensor. Walking step recognition system. 5. The method of claim 4,
The neural network includes:
An input layer having five input nodes;
A concealment layer having at least 20 intermediate nodes; And
And an output layer having at least one output node. 6. The method of claim 5,
Wherein the five input nodes comprise:
The ground contact pressure parameter X1,
The hip joint angular velocity parameter X2,
The knee angular velocity parameter X3,
The hip joint pitch information parameter X4, and
And the knee pitch information parameter (X5). The method according to claim 6,
Wherein the one output node is output in the form of one of a stance unit and a swing unit. 8. The method of claim 7,
The one output node may include at least one of an initial contact, a loading response, a mid stance, a terminal stance, a pre-swing, wherein the output is in the form of one of an initial swing, a mid swing, and a terminal swing. Measuring the ground surface reaction force and estimating whether or not the ground contact is made;
Measuring an angle of the hip joint, a hip joint, and a knee joint, and estimating an angular velocity of the hip joint and the knee joint;
Measuring the angular velocity and acceleration of the body part and estimating the body part posture; And
And a step of outputting the walking step information to one of a stance unit and a swing unit by driving a recognizer having the ground contact information, the angular speed information, and the body part attitude information as input parameters. Way. 10. The method of claim 9,
And estimating the hip joint pitch information and the knee joint pitch information by fusing the hip joint, the knee joint angular velocity information, and the knee angular velocity information with the body part posture information to estimate the hip joint pitch information, And the knee pitch information is further included as an input parameter.

Images