CN109940584A - The detection method that a kind of exoskeleton robot and its detection human motion are intended to - Google Patents
The detection method that a kind of exoskeleton robot and its detection human motion are intended to Download PDFInfo
- Publication number
- CN109940584A CN109940584A CN201910227324.1A CN201910227324A CN109940584A CN 109940584 A CN109940584 A CN 109940584A CN 201910227324 A CN201910227324 A CN 201910227324A CN 109940584 A CN109940584 A CN 109940584A
- Authority
- CN
- China
- Prior art keywords
- matrix
- data
- joint
- hip
- rod piece
- 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.)
- Pending
Links
Abstract
The invention discloses a kind of exoskeleton robots, including exoskeleton robot ontology and sensory perceptual system, control system and drive system；Exoskeleton robot ontology includes upper support structure, hip rod piece, thigh rod piece, shank rod piece, foot member, hip joint, knee joint, ankle-joint；Sensory perceptual system includes the thin film strain-gauge sensor for detecting plantar pressure, the encoder for detecting joint angles, detection joint power/torque force snesor, detection movement velocity/acceleration attitude transducer, the capacitance type sensor for detecting muscle tone, it is additionally provided with connection bandage, the fixed and non-contacting capacitance type sensor of human skin on bandage is connected, capacitance type sensor includes at least one electrode patch；Control system includes the full connection network operations module formed based on CNN machine learning algorithm, control system connection sensory perceptual system and the data that can receive to perceive system acquisition, the drive system that control system connection can drive exoskeleton robot ontology to move.
Description
Technical field
The present invention relates to exoskeleton robot technical fields, and in particular to a kind of exoskeleton robot and its detection human body fortune
The dynamic detection method being intended to.
Background technique
By the development of nearly many decades, rapid progress is achieved to the research of ectoskeleton robotic technology field, and obtain
Some important achievements were obtained, provide reference for the industry application of ectoskeleton.Current many exoskeleton systems, for example dress
Formula exoskeleton robot much fails the use demand for reaching human body although realizing simple functional movement.
On the one hand, since human motion has very strong independence, freedom degree is high, and information is complicated, movement multiplicity, and in order to
It realizes good identification and prediction, then needs the sensory perceptual system with sensor that can largely reflect system mode, recognize
It is delayed small, precision of prediction is high, and computation complexity is low.But at present in terms of human motion prediction, letter is coupled with ectoskeleton to human body
The research of breath is more deficient, especially between human body and ectoskeleton acquisition of information and information processing, and be based on coupling information
The problems such as estimation is with prediction human motion is more prominent, for example, the current bio signal detection of wearable exoskeleton robot stops
The bio signal detection to specific, the single mode such as neuron signal, electromyography signal, joint motions information, pressure is stayed in, especially
It is neuron signal and electromyography signal individual difference it is very big, this outer sensor itself is also very expensive.In addition, being believed based on brain
Number obtain motion intention, although its strong synchronous signal of overall importance is without lag, for current scientific and technological level, brain signal feature
It extracts and decoding technique is still immature, in this way, which its accuracy for obtaining intention is lower；Movement meaning is obtained based on electromyography signal
Figure, although its signal, without lag, the sensor due to measuring myoelectric information is transmitted according to skin surface when muscle activity
The soft or hard degree of low current signal or muscle infers the consciousness of behavior of people, sensor surface electrode placement positions, human body sweat
The uncertain factors such as liquid and Temperature changing will affect the stability and accuracy of acquired information.
On the other hand, ectoskeleton will realize the effect of auxiliary human body, and movement cannot be brought to hinder to human body, it should true
The paleocinetic completion of guarantor's body, to play the role of real power-assisted.Due to ectoskeleton be it is typical it is man-machine be highly coupled be
System, the characteristic with " people is in inner ring ", the control methods such as predefined track, predefined gait pattern are not suitable for ectoskeleton, this
Class method can sacrifice the independence of people.Man-machine harmony research is broadly divided at present and is not based on model and is based on two kinds of sides of model
Method, the control being not based in model depends on prior scheduled all more rules, is not easy to realize, and computation complexity is high, and right
Individual adaptability is not strong.System design based on model depends on the kinetic model of ectoskeleton mostly, does not consider human-computer interaction mould
Type, the interference of uncertainty and objective reality in kinetic model, rarely has consideration, will restrict the comprehensive of such control strategy
Energy.The characteristics of considering coupled needs the real-time motivation of adjustment model of state and human-computer interaction model according to system, and estimates
Meter systems interference, eliminates the influence of interference, to further improve man-machine harmony, realizes submissive motion control, assists people
Body is more natural, easily carries out relevant action.
Currently, some units both domestic and external or enterprise put into great effort to study the relevant skill of exoskeleton robot
Art, for example, " the wearable ectoskeleton power-assisted machine that northwest electromechanical engineering research institute, China North Industries Group Corporation (202 institute) develops
People " system self weight 24kg, 4-4.8 kilometers per hour of travel speed, one is about 2 minutes when dressing, and other people assist wearing 40
Second, unloading 30 seconds and foldable transportation adapt to the height ranges of 1.6m -1.9m, specified weight bearing 50kg, can continuous work 4 it is small
When.Compared to first generation exoskeleton system weight reduction 30%, backrest system realizes integrated design；Control system servo
Tracking delay time shortens 50%；Power battery has SMBus management function；Entire sensing network is optimized, and posture is surveyed
Dimension is measured from one-dimensional promotion to three-dimensional, the inverse dynamics calculation accuracy of system is made to improve 25%.However its motion delay and inspection
The effect for surveying the time is still poor.For another example, the AIDER exoskeleton robot of University of Electronic Science and Technology is based on online SVM (support arrow
Amount machine) intention assessment, every 20ms can update a sample data, and the sample data is by Inertial Measurement Unit (IMU) and pressure
The data composition of power shoes 17 is tieed up totally, judges time >=0.02s, effect is also not so good；For another example, Japanese HAL ectoskeleton use can be examined
The biosensors such as EEG signals and electromyography signal are surveyed to obtain the motion intention of wearer, however this kind of signal is usually more micro-
It is weak, it is very sensitive to environmental change, therefore acquisition difficulty is larger, acquisition quality is not high, this is limited outside HAL to a certain extent
Ability to work of the bone under complex environment.
Therefore, how to provide one kind being capable of more accurate recognition and prediction human motion and more preferably auxiliary human body is true to play
The exoskeleton robot of positive power-assisted becomes for those skilled in the art's technical problem urgently to be solved.
Summary of the invention
The present invention provides a kind of exoskeleton robot, including exoskeleton robot ontology and is installed on the ectoskeleton machine
Sensory perceptual system, control system and drive system on human body；
The exoskeleton robot ontology includes upper support structure and lower limb structure, and the lower limb structure is symmetrically set in institute
The left and right sides below upper support structure is stated, the upper support structure includes backrest, and the lower limb structure includes hip bar
Part, thigh rod piece, shank rod piece, foot member, the hip rod piece connect the backrest, further include connecting the hip rod piece
With the hip joint of the thigh rod piece, connect the thigh rod piece and the knee joint of the shank rod piece, connect the shank bar
The ankle-joint of part and the foot member；
The sensory perceptual system includes the thin film strain-gauge sensor for detecting plantar pressure, the thin film strain-gauge sensing
Device is installed in the shoes of the foot member, further includes the encoder for detecting joint angles at the hip and knee of left and right, described
The encoder is installed respectively at the hip and knee of left and right, further includes being sensed for detecting power/torque power at the hip and knee of left and right
Device installs the force snesor at the left and right hip and knee respectively, further includes for detecting movement velocity/acceleration posture
Sensor, the left and right shank rod piece, the left and right thigh rod piece and the backrest are installed the attitude transducer respectively, are also wrapped
Include the capacitance type sensor for detecting muscle tone at left and right thigh rod piece and left and right shank rod piece, the left and right thigh rod piece
With connection bandage is respectively equipped on the left and right shank rod piece, be fixed on each connection bandage non-contact with human skin
The capacitance type sensor, each capacitance type sensor includes at least one electrode patch；
The control system includes the full connection network operations module formed based on CNN machine learning algorithm, the control
System connects the sensory perceptual system and can receive the data of the sensory perceptual system acquisition, and the control system connection can drive
The drive system of the exoskeleton robot ontology movement.
Preferably, the electrode patch is fixedly installed on the inner surface of the connection bandage, the electrode patch back
Side from the inner surface is equipped with one layer of silicon rubber；
Alternatively, the electrode patch is fixedly installed between the inner surface and the outer surface of the connection bandage.
Preferably, each capacitance type sensor includes six electrode patch, the connection bandage is surrounded
When round, six electrode patch are evenly distributed along the circular, circumferential direction.
Preferably, the drive system includes knee joint motor driver and hip joint motor driver, the knee is closed
Section motor driver is connected and installed in the motor at the knee joint, and the hip joint motor driver is connected and installed in the hip
The motor of joint.
Preferably, further include filtering processing module and CNN machine learning module, the filtering processing module with it is described
CNN machine learning module is connected, and the filtering processing module connects the sensory perceptual system.
Preferably, the quantity of the thin film strain-gauge sensor in each shoes is seven, at corresponding foot thumb
The shoes in be equipped with a thin film strain-gauge sensor, set in the shoes of corresponding heel described there are two side by side
Thin film strain-gauge sensor corresponds to and is equipped with a thin film strain-gauge sensor at the front of heel, the rear of corresponding foot thumb
Place sets the thin film strain-gauge sensor there are three side by side.
Preferably, being set in the backrest there are five the attitude transducer, it is equipped with described in two side-by-side at corresponding waist
Attitude transducer corresponds to and is equipped with three attitude transducers triangular in shape at chest.
A kind of exoskeleton robot of the invention, has the following technical effect that
The sensory perceptual system of provided exoskeleton robot is made of multiple sensors, is mainly used to detect human-computer interaction work
With with feedback system operating status comprising for reflecting the force feedback information and position feedback information of motion intention, such as can
The data such as plantar pressure, joint power/torque, speed/acceleration are acquired, the motion intention identification of capacitance sensing is particularly based on,
Muscle tension sensor is changed to flexible wearing, contactless in muscle contact, and according to installation site positioning and human parameters
Detection, can detect and adjust automatically contact position and intensity, increase substantially adaptability, compare grinding based on electromyography signal
Study carefully, not only can achieve same accuracy of identification, but also overcome the shortcomings that surface electromyogram signal measurement must reach over to skin, is
The research of muscle deformation detection field provides a completely new method, can more accurately recognize and prediction human motion；
Further, multi-modal bio signal detection is carried out based on the sensory perceptual system, it is higher and general by movement, force feedback equally accurate
Property strong sensor, it is established that the physiological parameter model of human motion, and by CNN deep learning algoritic module makes the control be
System obtains the motion intention of exoskeleton robot wearer, and then establishes accurate human-computer interaction mechanism.The exoskeleton robot,
On the one hand will be significantly reduced to human nerve member and electromyography signal real-time capture demand, on the other hand based on precision it is higher and
The learning-oriented interaction mechanism of versatile signal feedback mechanism and manual intelligent can be such that ectoskeleton machine copes with well
The site environment of strange complexity.
Preferably, the arrangement of electrode patch makes it be not directly contacted with the skin of human body, muscle can be accurately measured
Shape deformation amount, muscle capacitor, muscle density measurement, EMG (surface electromyogram signal) obtain muscle tone.
It, can be more on complete detection thigh shank preferably, six electrode patch are evenly distributed along circular, circumferential direction
Muscle tone everywhere.
Preferably, the quantity of the thin film strain-gauge sensor in each shoes is seven, and is arranged in different location, can examine
More accurate plantar pressure is measured, the plantar pressure data is based on, can definitely distinguish different motion mode.
Preferably, being set in backrest there are five attitude transducer, two side-by-side attitude transducer is equipped at corresponding waist, it is right
Answer equally can be detected more accurate speed/acceleration, is based at chest equipped with three attitude transducers triangular in shape
The data can definitely distinguish different motion mode.
The present invention also provides the detection methods that a kind of detection human motion of exoskeleton robot is intended to, and the method includes such as
Lower step:
Step 1: several groups data of the exoskeleton robot under different motion mode are chosen and construct database, every group of number
According to type include plantar pressure, joint angles, joint power/torque, speed/acceleration, muscle tone, the plantar pressure
For the pressure in left and right vola, the joint angles are the angle in four joints of left and right knee hip, and the joint power/torque is left and right
Power/the torque in four joints of knee hip, the speed/acceleration is left and right shank, five key points of left and right thigh and upper trunk are total
The speed/acceleration at nine is counted, the muscle tone is left and right thigh and the total muscle tone everywhere of left and right shank, vola pressure
The data of power constitute a small data matrix, and the data of four joint angles constitute a four row small data matrix of a column, at nine
The data of speed/acceleration constitute a nine row small data matrix of a column, and power/torque data in four joints constitute one one
Four row small data matrix of column, the data of muscle tone constitute a four row small data matrix of a column everywhere, to each of detecting
The small data matrix is filtered, and later, five small data matrixes form one according to diagonal line Heterogeneous Permutation
Big matrix；
Step 2: the database of several big matrix buildings is inputted into CNN machine learning module, to the CNN
Machine learning module is trained, and the CNN machine learning module after training can form the full connection of a matrix type formula
The line number of network, the matrix type formula of the full connection network of formation is consistent with the line number of the big matrix；
Step 3: making according to pre-launch and detect one group of new data, the type of the data of one group of new data and institute
It is identical to state a kind of type of every group of data of step, one group of new data is input to the trained CNN machine learning mould
Block is to detect the motor pattern of human body next step.
Preferably, the several groups data under different motion mode are inputted the CNN machine in the step 2
When device study module, the different motion mode of the exoskeleton robot delimited label, as follows: walking mode label is 1, on
Building mode tag is 2, and downstairs mode tag is 3, and mode tag of squatting down is 4, and mode tag of sitting down is 5, and mode tag of standing up is
6,
In the step 3, one group of new data is input in the full connection network of the matrix type formula, can be calculated
It obtains the numerical value in above-mentioned label and then determines the motor pattern of human body next step.
Preferably, the plantar pressure is the pressure of seven key points on each vola, Gong Jishi in the step 1
Plantar pressure everywhere, the joint angles be four joints of left and right knee hip angle, the speed/acceleration be left and right shank,
Speed/acceleration at five key points of left and right thigh and upper trunk total nine, the joint power/torque are left and right knee hip four
The one dimension force in a joint, the muscle tone are left and right thigh and the total muscle tone everywhere of left and right shank, 14 volas
One column Ariadne small data matrix of stress structure, four joint angles constitute a four row small data matrixes of a column, at nine
Speed/acceleration constitutes a nine row small data matrix of a column, and four joint power/torque constitutes a four row small data square of a column
Battle array, muscle tone constitutes a four row small data matrix of a column everywhere, to each of detecting that the small data matrix filters
Wave processing, later, five small data matrixes form one according to diagonal line Heterogeneous Permutation, and there are 30 five-element 35 to arrange
Big matrix,
In the step 2, the line number of the matrix type formula of the full connection network of formation is 30 five-element,
In the step 3, the one group of new data matching for arranging big matrix with 30 five-element 35 has 35
The full connection network of the capable matrix type formula, each data in the big matrix be intended to and in the full connection network it is public
Formula calculates, and the big matrix dimensionality reduction row is at the matrix of two column, 30 five-element after the completion of calculating, and wherein first row is shown as number of tags
Value or zero, when being shown as zero, does not execute any movement, when being shown as label numerical value, it is corresponding to execute the label numerical value
Motor pattern, secondary series show to be the parameter moved in corresponding sports mode, and the kinematic parameter is transferred to control system, institute
Control system is stated to obtain executing order accordingly and then controlling drive system that the robot is driven to do corresponding movement.
Preferably, institute can be used in the detection method of the several groups data under different motion mode in the step 1
The method detected in step 3 is stated, point repeated detection of the exoskeleton robot under different motion mode.
Of the invention carries out the detection method that detection human motion is intended to based on above-mentioned exoskeleton robot, except above-mentioned identical
Technical effect outside, also have the following beneficial effects:
CNN machine learning module is trained by a large amount of data, a large amount of data are exoskeleton robots in difference
Data under motor pattern using CNN machine learning algorithm, are trained the CNN machine learning mould based on these data
Block forms it into the full connection network of a matrix type formula, in turn, is made according to pre-launch and detects one group of new data, will
One group of new data is input in trained CNN machine learning module, can detect that the motor pattern of human body next step.Meanwhile
It before input data, is filtered, by filtering processing, the error generated in acquisition data procedures can be reduced.The inspection
Survey method can produce more reliable, more acurrate, more complete information by operation, and judged according to these information, estimate and
Decision, so as to the motion intention of timely and accurate detection to human body.
It include label numerical value by the numerical value obtained after operation preferably, different motion mode is carried out delimitation label,
The motor pattern of human body next step corresponding to the label numerical value can be determined according to label numerical value.
Preferably, the pressure of key point at seven is detected in each vola, at the same the speed of five key points of upper trunk/plus
Speed aloows the numerical value detected more accurately to correspond to certain motor pattern, so that training CNN machine learning module
Effect it is more preferable.
Preferably, the data in database can also be obtained using set several sensors.
Detailed description of the invention
Fig. 1 be a kind of structural schematic diagram of the exoskeleton robot ontology of exoskeleton robot provided by the present invention (not
Connection bandage at thigh is shown)；
Fig. 2 is between a kind of sensory perceptual system of exoskeleton robot provided by the present invention, control system and drive system
Structural schematic diagram；
Fig. 3 is a kind of sensory perceptual system of exoskeleton robot provided by the present invention and machine filter wave processing module, CNN machine
Structural schematic diagram between device study module；
Fig. 4 is the distribution schematic diagram of thin film strain-gauge sensor；
Fig. 5 is the distribution schematic diagram of attitude transducer；
Fig. 6 is the process that a kind of exoskeleton robot provided by the present invention detects the detection method that human motion is intended to
Figure.
Appended drawing reference is as follows in Fig. 1-6:
1 upper support structure, 2 hip rod pieces, 3 thigh rod pieces, 4 shank rod pieces, 5 foot members, 6 hip joints, 7 knees close
Section, 8 ankle-joints, 9 connection bandages, 10 thin film strain-gauge sensors, 11 encoders, 12 force snesors, 13 attitude transducers, 14
Capacitance type sensor, 15 control systems, 16 drive systems, 17 filtering processing modules, 18CNN machine learning module.
Specific embodiment
It is with reference to the accompanying drawing and specific real in order to make those skilled in the art more fully understand technical solution of the present invention
Applying mode, the present invention is described in further detail.
As shown in figures 1 to 6, Fig. 1 is a kind of knot of the exoskeleton robot ontology of exoskeleton robot provided by the present invention
(the connection bandage at thigh is not shown) in structure schematic diagram；
Fig. 2 is between a kind of sensory perceptual system of exoskeleton robot provided by the present invention, control system and drive system
Structural schematic diagram；
Fig. 3 is a kind of sensory perceptual system of exoskeleton robot provided by the present invention and machine filter wave processing module, CNN machine
Structural schematic diagram between device study module；
Fig. 4 is the distribution schematic diagram of thin film strain-gauge sensor；
Fig. 5 is the distribution schematic diagram of attitude transducer；
Fig. 6 is the process that a kind of exoskeleton robot provided by the present invention detects the detection method that human motion is intended to
Figure.
In conjunction with Fig. 1-6, a kind of a kind of exoskeleton robot provided by the invention, in specific embodiment comprising dermoskeleton
Bone robot body and the sensory perceptual system being installed on exoskeleton robot ontology, control system 15 and drive system 16；
Wherein, as shown in Figure 1, exoskeleton robot ontology includes upper support structure 1 and lower limb structure, lower limb structure pair
Claim the left and right sides for being set to 1 lower section of upper support structure, upper support structure includes backrest, and lower limb structure includes hip bar
Part 2, thigh rod piece 3, shank rod piece 4, foot member 5, hip rod piece 2 connect backrest, further include connection hip rod piece 2 and thigh
Hip joint 6 that bar is 3, connect shank rod piece 4 and foot member 5 at connection thigh rod piece 3 and the knee joint 7 of shank rod piece 4
Ankle-joint 8；
Sensory perceptual system includes the thin film strain-gauge sensor 10 for detecting plantar pressure, and thin film strain-gauge sensor 10 is pacified
It further include the encoder 11 for detecting joint angles at the hip and knee of left and right in shoes loaded on foot member 5, left and right hip knee closes
The encoder 11 is installed respectively at section, further includes the left and right for detecting power/torque force snesor 12 at the hip and knee of left and right
The force snesor 12 is installed respectively at hip and knee, further include for detecting movement velocity/acceleration attitude transducer 13,
Left and right shank rod piece, left and right thigh rod piece and the backrest install the attitude transducer 13 respectively, further include for detecting a left side
The capacitance type sensor 14 of muscle tone, left and right thigh rod piece 3 and left and right shank at right thigh rod piece 3 and left and right shank rod piece 4
It is respectively equipped with connection bandage 9 on rod piece 4, is fixed with and the non-contacting capacitance type sensor of human skin on each connection bandage 9
14, each capacitance type sensor 14 includes at least one electrode patch；
Control system 15 includes the full connection network operations module formed based on CNN machine learning algorithm, control system 15
Connection sensory perceptual system and the data that can receive to perceive system acquisition, the connection of control system 15 can drive exoskeleton robot sheet
The drive system 16 of body movement.
Wherein, the material for connecting bandage can be woven material, but not limited to this.
The sensory perceptual system of provided exoskeleton robot is made of multiple sensors, is mainly used to detect human-computer interaction work
With with feedback system operating status comprising for reflecting the force feedback information and position feedback information of motion intention, such as can
The data such as plantar pressure, joint power/torque, speed/acceleration are acquired, the motion intention identification of capacitance sensing is particularly based on,
Muscle tension sensor is changed to flexible wearing, contactless in muscle contact, and according to installation site positioning and human parameters
Detection, can detect and adjust automatically contact position and intensity, increase substantially adaptability, compare grinding based on electromyography signal
Study carefully, not only can achieve same accuracy of identification, but also overcome the shortcomings that surface electromyogram signal measurement must reach over to skin, is
The research of muscle deformation detection field provides a completely new method, can more accurately recognize and prediction human motion；
Further, multi-modal bio signal detection is carried out based on the sensory perceptual system, it is higher and general by movement, force feedback equally accurate
Property strong sensor, it is established that the physiological parameter model of human motion, and by CNN deep learning algoritic module makes the control be
System obtains the motion intention of exoskeleton robot wearer, and then establishes accurate human-computer interaction mechanism.The exoskeleton robot,
On the one hand will be significantly reduced to human nerve member and electromyography signal real-time capture demand, on the other hand based on precision it is higher and
The learning-oriented interaction mechanism of versatile signal feedback mechanism and manual intelligent can be such that ectoskeleton machine copes with well
The site environment of strange complexity.
Wherein, about the specific setting position of electrode patch, one of following two mode can be used:
The first, electrode patch is fixedly installed on the inner surface of connection bandage 9, and electrode patch is away from the inner surface
Side is equipped with one layer of silicon rubber；The thickness that one layer of silicon rubber is placed can be 2.5 millimeters.
Second, electrode patch is fixedly installed between the inner surface and the outer surface of connection bandage 9, directly by electrode patch
It is placed in inside connection bandage 9.
The arrangement of electrode patch makes it be not directly contacted with the skin of human body, can accurately measure muscle shape deformation
Amount, the measurement of muscle capacitor, muscle density, EMG (surface electromyogram signal) obtain muscle tone.
In the specific embodiment, each condenser type includes six electrode patch, when connection bandage 9 surrounds circle, six
A electrode patch is evenly distributed along circular, circumferential direction.Number is not limited to six.
Six electrode patch are evenly distributed along circular, circumferential direction, muscle that can be more on complete detection thigh shank everywhere
Tension.
Further, in the specific embodiment, drive system 16 includes knee joint motor driver and hip joint motor
Driver, knee joint motor driver are connected and installed in the motor at knee joint, and hip joint motor driver is connected and installed in hip
The motor of joint.
In the specific embodiment, as shown in figure 3, further including filtering processing module 17 and CNN machine learning module 18, filter
Wave processing module 17 is connected with CNN machine learning module 18, and filtering processing module 17 connects sensory perceptual system.Sensory perceptual system acquisition
Data information can reduce the error that generates in acquisition data procedures after filtering processing module 17 filters, and then pass through CNN
Machine learning module 18 forms full connection network operations module based on CNN machine learning algorithm.
As shown in figure 4, the quantity of the thin film strain-gauge sensor 10 in each shoes is seven, in the shoes at corresponding foot thumb
Equipped with a thin film strain-gauge sensor 10, set in the shoes of corresponding heel there are two thin film strain-gauge sensor 10 side by side,
It is equipped with a thin film strain-gauge sensor 10 at the front of corresponding heel, sets that there are three side by side at the rear of corresponding foot thumb
Thin film strain-gauge sensor 10.
The quantity of thin film strain-gauge sensor 10 in each shoes is seven, and is arranged in different location, be can be detected more
Add accurate plantar pressure, is based on the plantar pressure data, can definitely distinguish different motion mode.
As shown in figure 5, being set in backrest there are five attitude transducer 13, two side-by-side attitude transducer is equipped at corresponding waist
13, it corresponds to and is equipped with three attitude transducers 13 triangular in shape at chest.
It is set in backrest there are five attitude transducer 13, two side-by-side attitude transducer 13, corresponding chest is equipped at corresponding waist
Three attitude transducers 13 triangular in shape are equipped at dried meat equally can be detected more accurate speed/acceleration, and being based on should
Data can definitely distinguish different motion mode.
In conjunction with Fig. 6, the present invention also provides the detection method that a kind of detection human motion of exoskeleton robot is intended to, method packets
Include following steps:
Step 1: several groups data of the exoskeleton robot under different motion mode are chosen and construct database, every group of number
According to type include plantar pressure, joint angles, joint power/torque, speed/acceleration, muscle tone, plantar pressure is a left side
The pressure in right vola, joint angles are the angle in four joints of left and right knee hip, and joint power/torque is four joints of left and right knee hip
Power/torque, speed/acceleration is left and right shank, speed/acceleration at five key points of left and right thigh and upper trunk total nine
Degree, muscle tone are left and right thigh and the total muscle tone everywhere of left and right shank, and the data of plantar pressure constitute a decimal
According to matrix, the data of four joint angles constitute a four row small data matrix of a column, and the data of speed/acceleration are constituted at nine
Power/torque data of one nine row small data matrix of a column, four joints constitute a four row small data matrix of a column, everywhere
The data of muscle tone constitute a four row small data matrix of a column, are filtered place to each small data matrix detected
Reason, later, five small data matrixes form a big matrix according to diagonal line Heterogeneous Permutation；
Step 2: the database of several big matrix buildings is inputted into CNN machine learning module, to CNN machine learning module
It is trained, the CNN machine learning module after training can form the full connection network of a matrix type formula, the full connection of formation
The line number of the matrix type formula of network is consistent with the line number of big matrix；
Step 3: making according to pre-launch and detect one group of new data, the type and step 1 of the data of one group of new data
The type of every group of data of kind is identical, and one group of new data is input to trained CNN machine learning module to detect under human body
The motor pattern of one step.
In this method, CNN machine learning module is trained by a large amount of data, a large amount of data are ectoskeleton machines
Data of the people under different motion mode using CNN machine learning algorithm, are trained CNN machine learning based on these data
Module forms it into the full connection network of a matrix type formula, in turn, is made according to pre-launch and detects one group of new data,
One group of new data is input in trained CNN machine learning module, can detect that the motor pattern of human body next step.Together
When, it before input data, is filtered, by filtering processing, the error generated in acquisition data procedures can be reduced.It should
Detection method can produce more reliable, more acurrate, more complete information, and judge, estimate according to these information by operation
And decision, so as to the motion intention of timely and accurate detection to human body.
Further, in this method,
In step 2, when the several groups data under different motion mode are inputted CNN machine learning module, ectoskeleton machine
The different motion mode of device people delimit label, as follows: walking mode label is 1, and mode tag is 2 upstairs, downstairs mode tag
It is 3, mode tag of squatting down is 4, and mode tag of sitting down is 5, and mode tag of standing up is 6,
In step 3, one group of new data is input in the full connection network of matrix type formula, and above-mentioned label can be calculated
In numerical value so that determine human body next step motor pattern.
Different motion mode is subjected to delimitation label, includes label numerical value by the numerical value obtained after operation, it can be according to mark
Label numerical value determines the motor pattern of human body next step corresponding to the label numerical value.
In this method, a kind of preferred embodiment is as follows:
In step 1, plantar pressure is the pressure of seven key points on each vola, amounts to ten plantar pressures everywhere, joint
Angle is the angle in four joints of left and right knee hip, and speed/acceleration is five left and right shank, left and right thigh and upper trunk key points
Speed/acceleration at total nine, joint power/torque are the one dimension force in four joints of left and right knee hip, and muscle tone is left and right
Thigh and the total muscle tone everywhere of left and right shank, 14 plantar pressures constitute a column Ariadne small data matrix,
Four joint angles constitute a four row small data matrix of a column, and speed/acceleration constitutes a nine row small data of a column at nine
Matrix, four joint power/torque constitute a four row small data matrix of a column, and it is small to constitute four row of a column for muscle tone everywhere
Data matrix is filtered each small data matrix detected, and later, five small data matrixes are according to diagonal line mistake
Position arrangement forms the big matrix with 30 five-element 35 column,
In step 2, the line number of the matrix type formula of the full connection network of formation is 30 five-element,
In step 3, there is the matrix of 30 five-element with one group of new data matching that 30 five-element 35 arrange big matrix
The full connection network of type formula, each data in big matrix are intended to calculate with formula in full connection network, big after the completion of calculating
Matrix dimensionality reduction row is at the matrix of two column, 30 five-element, and wherein first row is shown as label numerical value or zero, when being shown as zero, does not hold
Any movement of row when being shown as label numerical value, executes the corresponding motor pattern of label numerical value, and secondary series shows to be Xiang Yingyun
The parameter moved in dynamic model formula, kinematic parameter are transferred to control system, and control system obtains executing accordingly and orders and then control
Drive system driving robot does corresponding movement.
The pressure of key point at seven, while the speed/acceleration of five key points of upper trunk are detected in each vola, can make
The numerical value that must be detected can more accurately correspond to certain motor pattern, so that the effect of training CNN machine learning module is more preferable.
Specifically, in use (being illustrated with walking mode), following step starting row until then, human body and equipment meeting
There are some pre-launch to make, which can be detected these movements by sensory perceptual system, can obtain five groups of individual sample vectors
That is small data matrix, wherein speed/acceleration data vector 9, constitute the matrix of nine row of a column such asPlantar pressure
Data 14 are such asThe matrix of a column Ariadne is constituted, power/torque data 4 are such asConstitute the square of four row of a column
Battle array, 4 data of joint angles are such asThe matrix etc. for constituting four row of a column, there are also muscle tone data, constitute a column four
Row small data matrix is filtered for independent each vector later, and vector each in this way can be smoothened (two neighboring
Can become between number more like), data fusion is spliced into a complete big matrix (dimension can be very high) later, is similar to this
SampleForm the big matrix of 30 five-element 35 column；Later to joining entirely in the matrix machine
Open network calculates, because full connection network is also the formula of a matrix type, the element being equivalent in each matrix will
Calculate that one time (first row of Quan Liantong network is taken as input layer, last column is output with each formula
Layer, centre are all hidden layer, and the calculating of matrix in layer is exactly the calculating arranged, output in fact
Layer meeting dimensionality reduction, becomes only two column), meeting dimensionality reduction, can obtain final matrix of consequence, matrix of consequence has 2 after calculating terminates
The matrix of column large number of rows, wherein first row is either 1 entirely or is 0,1 to represent and execute movement (only 1 machine talent conference row entirely
Walk), 0 represents invalid result (robot does nothing, and recalculates intention detection), has difference in different modes
Data represent the classification of motion such as 2 upstairs, 3 be downstairs, 4 be squat down, 5 be sit down, 6 be to stand up, secondary series representative movement ginseng
Number, with walking for such as spatial key point coordinate, stride, leg speed, step height, the parameter of secondary series can be sent to control system
System, control system can do by myself to be calculated by these parameters executes order accordingly.
In the embodiment, each capacitance type sensor includes an electrode patch, without being limited thereto, for example be can also be used
Each capacitance type sensor includes the exoskeleton robot of six electrode patch to be detected, at this point, respective muscle tension
The line number for the small data matrix that detection data is formed can be different from the line number of small data matrix in the above method, correspondingly, institute's shape
At complete big matrix line number and columns also can be different.But detecting step is still consistent, it is evident that its accuracy detected is wanted
Better than the accuracy in the above method.
In this method, further, in step 1, the detection method of the several groups data under different motion mode can be adopted
With the method detected in above-mentioned steps three, divide repeated detection of the exoskeleton robot under different motion mode.It can certainly
By buy existing data come using.
Embodiment of above is only exemplary embodiments of the present invention, is not used in the limitation present invention, protection of the invention
Range is defined by the claims.Those skilled in the art within the spirit and scope of the present invention, make the present invention
Various modifications or equivalent replacements are also fallen within the protection scope of the present invention.
Claims (11)
1. a kind of exoskeleton robot, which is characterized in that including exoskeleton robot ontology and be installed on the ectoskeleton machine
Sensory perceptual system, control system and drive system on human body；
The exoskeleton robot ontology includes upper support structure and lower limb structure, and the lower limb structure is symmetrically set on described
The left and right sides below portion's support construction, the upper support structure include backrest, and the lower limb structure includes hip rod piece, big
Leg rod piece, shank rod piece, foot member, the hip rod piece connect the backrest, further include connecting the hip rod piece and institute
State the hip joint of thigh rod piece, the connection thigh rod piece and the knee joint of the shank rod piece, connect the shank rod piece and
The ankle-joint of the foot member；
The sensory perceptual system includes the thin film strain-gauge sensor for detecting plantar pressure, the thin film strain-gauge sensor peace
It further include the encoder for detecting joint angles at the hip and knee of left and right, the left and right in shoes loaded on the foot member
The encoder is installed respectively at hip and knee, further includes for detecting power/torque force snesor at the hip and knee of left and right, institute
It states and the force snesor is installed respectively at the hip and knee of left and right, further include for detecting movement velocity/acceleration posture sensing
Device, the left and right shank rod piece, the left and right thigh rod piece and the backrest install the attitude transducer respectively, further include using
The capacitance type sensor of muscle tone, the left and right thigh rod piece and institute at detection left and right thigh rod piece and left and right shank rod piece
It states and is respectively equipped with connection bandage on the shank rod piece of left and right, be fixed on each connection bandage and the non-contacting institute of human skin
Capacitance type sensor is stated, each capacitance type sensor includes at least one electrode patch；
The control system includes the full connection network operations module formed based on CNN machine learning algorithm, the control system
It connects the sensory perceptual system and the data of the sensory perceptual system acquisition can be received, the control system connection can drive described
The drive system of exoskeleton robot ontology movement.
2. exoskeleton robot according to claim 1, which is characterized in that the electrode patch is fixedly installed in the company
It connects on the inner surface of bandage, the electrode patch is equipped with one layer of silicon rubber away from the side of the inner surface；
Alternatively, the electrode patch is fixedly installed between the inner surface and the outer surface of the connection bandage.
3. exoskeleton robot according to claim 2, which is characterized in that each capacitance type sensor includes six
A electrode patch, when the connection bandage surrounds circle, six electrode patch are uniform along the circular, circumferential direction
Arrangement.
4. exoskeleton robot according to claim 1, which is characterized in that the drive system includes that knee joint motor drives
Dynamic device and hip joint motor driver, the knee joint motor driver is connected and installed in the motor at the knee joint, described
Hip joint motor driver is connected and installed in the motor at the hip joint.
5. exoskeleton robot according to claim 1, which is characterized in that further include filtering processing module and CNN machine
Study module, the filtering processing module are connected with the CNN machine learning module, described in the filtering processing module connection
Sensory perceptual system.
6. exoskeleton robot according to claim 1, which is characterized in that the thin film strain-gauge in each shoes
The quantity of sensor is seven, and a thin film strain-gauge sensor, corresponding foot are equipped in the shoes at corresponding foot thumb
It is set in the shoes of heel there are two the thin film strain-gauge sensor side by side, is equipped at the front of corresponding heel one thin
Membrane strain piece sensor corresponds to thin film strain-gauge sensor side by side there are three setting at the rear of foot thumb.
7. exoskeleton robot according to claim 1, which is characterized in that the posture passes there are five setting in the backrest
Sensor corresponds to and is equipped with attitude transducer described in two side-by-side at waist, is equipped with three appearances triangular in shape at corresponding chest
State sensor.
8. a kind of detection method that exoskeleton robot detection human motion is intended to, which is characterized in that the method includes as follows
Step:
Step 1: choosing several groups data of the exoskeleton robot under different motion mode and construct database, every group of data
Type includes plantar pressure, joint angles, joint power/torque, speed/acceleration, muscle tone, and the plantar pressure is a left side
The pressure in right vola, the joint angles are the angle in four joints of left and right knee hip, and the joint power/torque is left and right knee hip
Power/the torque in four joints, the speed/acceleration are left and right shank, five key points of left and right thigh and upper trunk total nine
The speed/acceleration at place, the muscle tone amount to muscle tone everywhere for left and right thigh and left and right shank, plantar pressure
Data constitute a small data matrix, and the data of four joint angles constitute a four row small data matrix of a column, and speed at nine/
The data of acceleration constitute a nine row small data matrix of a column, and power/torque data in four joints constitute a column four
Row small data matrix, the data of muscle tone constitute a four row small data matrix of a column everywhere, described to each of detecting
Small data matrix is filtered, and later, five small data matrixes form a big square according to diagonal line Heterogeneous Permutation
Battle array；
Step 2: the database of several big matrix buildings is inputted into CNN machine learning module, to the CNN machine
Study module is trained, and the CNN machine learning module after training can form the full connection network of a matrix type formula,
The line number of the matrix type formula of the full connection network formed is consistent with the line number of the big matrix；
Step 3: making according to pre-launch and detect one group of new data, the type and the step of the data of one group of new data
A kind of rapid type of every group of data is identical, by one group of new data be input to the trained CNN machine learning module with
Detect the motor pattern of human body next step.
9. detection method according to claim 8, which is characterized in that
In the step 2, when the several groups data under different motion mode are inputted the CNN machine learning module,
The different motion mode of the exoskeleton robot delimit label, as follows: walking mode label is 1, and mode tag is 2 upstairs,
Downstairs mode tag is 3, and mode tag of squatting down is 4, and mode tag of sitting down is 5, and mode tag of standing up is 6,
In the step 3, one group of new data is input in the full connection network of the matrix type formula, can be calculated
The motor pattern of numerical value and then determining human body next step in above-mentioned label.
10. detection method according to claim 9, which is characterized in that
In the step 1, the plantar pressure is the pressure of seven key points on each vola, amounts to ten plantar pressures everywhere,
The joint angles are the angle in four joints of left and right knee hip, and the speed/acceleration is left and right shank, left and right thigh and upper body
Speed/acceleration at dry five key points total nine, the joint power/torque are the one-dimensional of four joints of left and right knee hip
Power, the muscle tone are left and right thigh and the total muscle tone everywhere of left and right shank, and 14 plantar pressures constitute one
One column Ariadne small data matrix, four joint angles constitute a four row small data matrix of a column, speed/acceleration structure at nine
At a nine row small data matrix of a column, four joint power/torque constitutes a four row small data matrix of a column, everywhere muscle
Power constitutes a four row small data matrix of a column, to each of detecting that the small data matrix is filtered, later, five
A small data matrix forms the big matrix with 30 five-element 35 column according to diagonal line Heterogeneous Permutation,
In the step 2, the line number of the matrix type formula of the full connection network of formation is 30 five-element,
In the step 3, the one group of new data matching for arranging big matrix with 30 five-element 35 has 30 five-element's
The full connection network of the matrix type formula, each data in the big matrix are intended to and formula meter in the full connection network
Calculate, after the completion of calculating the big matrix dimensionality reduction row at the matrix of two column, 30 five-element, wherein first row be shown as label numerical value or
Person zero, when being shown as zero, does not execute any movement, when being shown as label numerical value, executes the corresponding movement of the label numerical value
Mode, secondary series show to be the parameter moved in corresponding sports mode, and the kinematic parameter is transferred to control system, the control
System processed obtains executing order accordingly and then controlling drive system that the robot is driven to do corresponding movement.
11. detection method according to claim 8, which is characterized in that
In the step 1, the detection method of the several groups data under different motion mode can be used to be detected in the step 3
Method, point repeated detection of the exoskeleton robot under different motion mode.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910227324.1A CN109940584A (en) | 2019-03-25 | 2019-03-25 | The detection method that a kind of exoskeleton robot and its detection human motion are intended to |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910227324.1A CN109940584A (en) | 2019-03-25 | 2019-03-25 | The detection method that a kind of exoskeleton robot and its detection human motion are intended to |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109940584A true CN109940584A (en) | 2019-06-28 |
Family
ID=67011628
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910227324.1A Pending CN109940584A (en) | 2019-03-25 | 2019-03-25 | The detection method that a kind of exoskeleton robot and its detection human motion are intended to |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109940584A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110236539A (en) * | 2019-07-22 | 2019-09-17 | 宿州学院 | A kind of human body lower limbs surface electromyogram signal acquisition and pattern recognition system |
| CN111251276A (en) * | 2020-01-20 | 2020-06-09 | 南方科技大学 | Power assisting method and device based on gesture, server and storage medium |
| CN111267071A (en) * | 2020-02-14 | 2020-06-12 | 上海航天控制技术研究所 | Multi-joint combined control system and method for exoskeleton robot |
| CN111419632A (en) * | 2020-03-31 | 2020-07-17 | 合肥工业大学 | Redundant constraint flexible cable driven lower limb training parallel rehabilitation robot and control method thereof |
| CN111506190A (en) * | 2020-03-31 | 2020-08-07 | 哈尔滨工业大学 | Real-time continuous prediction method for human motion intention based on myoelectricity |
| CN111611887A (en) * | 2020-05-08 | 2020-09-01 | 武汉工程大学 | User intention estimation method and system for man-machine hybrid coordinated movement |
| CN111660297A (en) * | 2020-06-16 | 2020-09-15 | 南京航空航天大学 | Lasso driven wearable cooperation robot |
| CN112836617A (en) * | 2021-01-28 | 2021-05-25 | 北京理工大学前沿技术研究院 | IPSO-BPNN-based long-term human body lower limb movement prediction method |
| CN113043248A (en) * | 2021-03-16 | 2021-06-29 | 东北大学 | Transportation and assembly whole-body exoskeleton system based on multi-source sensor and control method |
| CN113143298A (en) * | 2020-03-31 | 2021-07-23 | 重庆牛迪普力科技合伙企业(有限合伙) | Limb skeletal muscle force-exerting state detection device and method and force-exerting state identification equipment |
| CN113180643A (en) * | 2021-04-25 | 2021-07-30 | 燕山大学 | Exoskeleton assistance detection device and evaluation method thereof |
| CN113459163A (en) * | 2021-07-12 | 2021-10-01 | 北京哈崎机器人科技有限公司 | Output end encoder, joint module, robot and rotation angle determination method |
| CN113576845A (en) * | 2021-08-02 | 2021-11-02 | 哈尔滨工程大学 | Human body consciousness recognition control device and method applied to exoskeleton robot |
| CN113910206A (en) * | 2021-12-15 | 2022-01-11 | 深圳市迈步机器人科技有限公司 | Exoskeleton assistance system combined with multiple sensors and assistance detection method thereof |
| CN114028775A (en) * | 2021-12-08 | 2022-02-11 | 福州大学 | Ankle joint movement intention identification method and system based on sole pressure |
| CN114191260A (en) * | 2021-11-25 | 2022-03-18 | 天津大学 | Highly integrated brain-controlled nerve electrical stimulation exoskeleton robot system and control method |
| CN116690567A (en) * | 2023-06-21 | 2023-09-05 | 湖南大学 | Lower limb exoskeleton driving method based on anthropomorphic nerve model |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107126344A (en) * | 2017-07-05 | 2017-09-05 | 天津科技大学 | Lower limb walking function rehabilitation exoskeleton rehabilitation robot and control system and method |
| CN107536613A (en) * | 2016-06-29 | 2018-01-05 | 深圳光启合众科技有限公司 | Robot and its human body lower limbs Gait Recognition apparatus and method |
-
2019
- 2019-03-25 CN CN201910227324.1A patent/CN109940584A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107536613A (en) * | 2016-06-29 | 2018-01-05 | 深圳光启合众科技有限公司 | Robot and its human body lower limbs Gait Recognition apparatus and method |
| CN107126344A (en) * | 2017-07-05 | 2017-09-05 | 天津科技大学 | Lower limb walking function rehabilitation exoskeleton rehabilitation robot and control system and method |
Non-Patent Citations (2)
| Title |
|---|
| 奚如如: "套索驱动半被动下肢助力外骨骼研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
| 龙亿: "下肢外骨骼人体运动预测与人机协调控制技术研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110236539B (en) * | 2019-07-22 | 2021-08-31 | 宿州学院 | Human lower limb surface electromyographic signal acquisition and pattern recognition system |
| CN110236539A (en) * | 2019-07-22 | 2019-09-17 | 宿州学院 | A kind of human body lower limbs surface electromyogram signal acquisition and pattern recognition system |
| CN111251276A (en) * | 2020-01-20 | 2020-06-09 | 南方科技大学 | Power assisting method and device based on gesture, server and storage medium |
| CN111267071A (en) * | 2020-02-14 | 2020-06-12 | 上海航天控制技术研究所 | Multi-joint combined control system and method for exoskeleton robot |
| CN111506190A (en) * | 2020-03-31 | 2020-08-07 | 哈尔滨工业大学 | Real-time continuous prediction method for human motion intention based on myoelectricity |
| CN113143298A (en) * | 2020-03-31 | 2021-07-23 | 重庆牛迪普力科技合伙企业(有限合伙) | Limb skeletal muscle force-exerting state detection device and method and force-exerting state identification equipment |
| CN111419632A (en) * | 2020-03-31 | 2020-07-17 | 合肥工业大学 | Redundant constraint flexible cable driven lower limb training parallel rehabilitation robot and control method thereof |
| CN111611887A (en) * | 2020-05-08 | 2020-09-01 | 武汉工程大学 | User intention estimation method and system for man-machine hybrid coordinated movement |
| CN111611887B (en) * | 2020-05-08 | 2023-04-11 | 武汉工程大学 | User intention estimation method and system for man-machine hybrid coordinated movement |
| WO2021223502A1 (en) * | 2020-05-08 | 2021-11-11 | 武汉工程大学 | Human-machine hybrid coordinated motion user intention estimation method and system |
| CN111660297A (en) * | 2020-06-16 | 2020-09-15 | 南京航空航天大学 | Lasso driven wearable cooperation robot |
| CN112836617A (en) * | 2021-01-28 | 2021-05-25 | 北京理工大学前沿技术研究院 | IPSO-BPNN-based long-term human body lower limb movement prediction method |
| CN112836617B (en) * | 2021-01-28 | 2023-02-03 | 北京理工大学前沿技术研究院 | IPSO-BPNN-based long-term human body lower limb movement prediction method |
| CN113043248A (en) * | 2021-03-16 | 2021-06-29 | 东北大学 | Transportation and assembly whole-body exoskeleton system based on multi-source sensor and control method |
| CN113043248B (en) * | 2021-03-16 | 2022-03-11 | 东北大学 | Transportation and assembly whole-body exoskeleton system based on multi-source sensor and control method |
| CN113180643A (en) * | 2021-04-25 | 2021-07-30 | 燕山大学 | Exoskeleton assistance detection device and evaluation method thereof |
| CN113180643B (en) * | 2021-04-25 | 2022-09-02 | 燕山大学 | Exoskeleton assistance detection device and evaluation method thereof |
| CN113459163A (en) * | 2021-07-12 | 2021-10-01 | 北京哈崎机器人科技有限公司 | Output end encoder, joint module, robot and rotation angle determination method |
| CN113576845A (en) * | 2021-08-02 | 2021-11-02 | 哈尔滨工程大学 | Human body consciousness recognition control device and method applied to exoskeleton robot |
| CN114191260A (en) * | 2021-11-25 | 2022-03-18 | 天津大学 | Highly integrated brain-controlled nerve electrical stimulation exoskeleton robot system and control method |
| CN114191260B (en) * | 2021-11-25 | 2023-12-15 | 天津大学 | Highly integrated brain-controlled nerve electric stimulation exoskeleton robot system and control method |
| CN114028775A (en) * | 2021-12-08 | 2022-02-11 | 福州大学 | Ankle joint movement intention identification method and system based on sole pressure |
| CN113910206A (en) * | 2021-12-15 | 2022-01-11 | 深圳市迈步机器人科技有限公司 | Exoskeleton assistance system combined with multiple sensors and assistance detection method thereof |
| CN116690567A (en) * | 2023-06-21 | 2023-09-05 | 湖南大学 | Lower limb exoskeleton driving method based on anthropomorphic nerve model |
| CN116690567B (en) * | 2023-06-21 | 2024-03-12 | 湖南大学 | Lower limb exoskeleton driving method based on anthropomorphic nerve model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109940584A (en) | The detection method that a kind of exoskeleton robot and its detection human motion are intended to | |
| CN110811553B (en) | Detection method for assistance efficiency of load exoskeleton | |
| CN110916679A (en) | Human body lower limb pose gait detection device and method | |
| Liu et al. | Deep spatial-temporal model for rehabilitation gait: Optimal trajectory generation for knee joint of lower-limb exoskeleton | |
| Chinmilli et al. | A review on wearable inertial tracking based human gait analysis and control strategies of lower-limb exoskeletons | |
| CN110420029A (en) | A kind of walking step state wireless detecting system based on Multi-sensor Fusion | |
| Bonnet et al. | Fast determination of the planar body segment inertial parameters using affordable sensors | |
| CN113043248B (en) | Transportation and assembly whole-body exoskeleton system based on multi-source sensor and control method | |
| Yang et al. | Novel soft smart shoes for motion intent learning of lower limbs using LSTM with a convolutional autoencoder | |
| Zeng et al. | Research on a gait detection system and recognition algorithm for lower limb exoskeleton robot | |
| Ding et al. | Control of walking assist exoskeleton with time-delay based on the prediction of plantar force | |
| CN110974242A (en) | Gait abnormal degree evaluation method for wearable device and wearable device | |
| Li et al. | Reconstructing walking dynamics from two shank-mounted inertial measurement units | |
| CN202932930U (en) | Wearable sensor measurement device for acting forces and moments related to lower limb joints of human body | |
| Hayashibe et al. | Muscle strength and mass distribution identification toward subject-specific musculoskeletal modeling | |
| Shull et al. | An overview of wearable sensing and wearable feedback for gait retraining | |
| CN116115217A (en) | Human lower limb gait phase estimation method based on depth network | |
| Ogata et al. | Estimating movements of human body for the shirt-type wearable device mounted on the strain sensors based on convolutional neural networks | |
| Chen et al. | An adaptive gait learning strategy for lower limb exoskeleton robot | |
| Peng | A novel motion detecting strategy for rehabilitation in smart home | |
| CN114224325A (en) | Gait analysis system and method for calculating joint moment and angle by using inertial sensor | |
| Wu et al. | Recognization of stance phase using flexible pressure sensors | |
| Wahab et al. | Microsystem based portable shoe integrated instrumentation using ultrasonic for gait analysis measurement | |
| Asif et al. | Analysis of Human Gait Cycle With Body Equilibrium Based on Leg Orientation | |
| Hnat et al. | Estimating center of mass kinematics during perturbed human standing using accelerometers |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190628 |