CN112336590A - Power-assisted exoskeleton movement intention and gait planning method based on multi-sensing information - Google Patents
Power-assisted exoskeleton movement intention and gait planning method based on multi-sensing information Download PDFInfo
- Publication number
- CN112336590A CN112336590A CN202011379078.0A CN202011379078A CN112336590A CN 112336590 A CN112336590 A CN 112336590A CN 202011379078 A CN202011379078 A CN 202011379078A CN 112336590 A CN112336590 A CN 112336590A
- Authority
- CN
- China
- Prior art keywords
- characteristic parameters
- leg
- gait
- signal characteristic
- acceleration
- 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
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H3/00—Appliances for aiding patients or disabled persons to walk about
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H2201/00—Characteristics of apparatus not provided for in the preceding codes
- A61H2201/16—Physical interface with patient
- A61H2201/1657—Movement of interface, i.e. force application means
- A61H2201/1659—Free spatial automatic movement of interface within a working area, e.g. Robot
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H2201/00—Characteristics of apparatus not provided for in the preceding codes
- A61H2201/50—Control means thereof
- A61H2201/5058—Sensors or detectors
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H2201/00—Characteristics of apparatus not provided for in the preceding codes
- A61H2201/50—Control means thereof
- A61H2201/5058—Sensors or detectors
- A61H2201/5084—Acceleration sensors
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61H—PHYSICAL THERAPY APPARATUS, e.g. DEVICES FOR LOCATING OR STIMULATING REFLEX POINTS IN THE BODY; ARTIFICIAL RESPIRATION; MASSAGE; BATHING DEVICES FOR SPECIAL THERAPEUTIC OR HYGIENIC PURPOSES OR SPECIFIC PARTS OF THE BODY
- A61H2230/00—Measuring physical parameters of the user
- A61H2230/08—Other bio-electrical signals
- A61H2230/085—Other bio-electrical signals used as a control parameter for the apparatus
Abstract
The invention provides a power-assisted exoskeleton movement intention and gait planning method based on multi-sensing information, which comprises the following steps: step 1, acquiring an acceleration signal and human oxygen consumption data of a human leg in the vertical direction, and acquiring a surface electromyographic signal of leg muscles; step 2, taking the acceleration signal characteristic parameter, the characteristic parameter of oxygen consumption data and the surface electromyographic signal characteristic parameter as the input of a training model; and step 3: establishing an LSTM deep learning network to obtain the corresponding relations between the surface electromyogram signal characteristic parameter and gait plan, the surface electromyogram signal characteristic parameter and exoskeleton assistance level, and the surface electromyogram signal characteristic parameter and motion mode, and outputting by taking the assistance level, the motion mode and the gait plan as models. The method can give out key parameters required by the power-assisted exoskeleton control system comprehensively, improves the resolving efficiency of real-time intentions, and can realize quick and accurate sharing, decision making and execution.
Description
Technical Field
The invention belongs to the technical field of man-machine cooperative control of exoskeleton robots, and particularly relates to a multi-sensor information-based power-assisted exoskeleton motion intention and gait planning method, which is used for active intention recognition, assistance efficiency grading and real-time gait planning of a power-assisted exoskeleton robot.
Background
With the continuous development of science and technology, the power-assisted exoskeleton robot is unprecedentedly developed in the field of military and civilian, and the exoskeleton has the most remarkable characteristic that the exoskeleton is in real-time interaction with a wearer in the whole process, the wearer is in real physical contact with the exoskeleton body, and the exoskeleton is a human-computer integrated system. Exoskeleton robot motion control needs to have high stability and robustness, and needs to have "naturalness" and "adaptability". "naturalness" means that a motion model is established based on a normal motion pattern of a human so that muscle activity in motion and subjective feeling of the human are close to the motion pattern of the human. The "adaptability" means that the action model is automatically matched and adjusted according to the physical conditions and habits of different users.
In the power-assisted exoskeleton system, a person is responsible for environment perception and behavior decision, motion instructions are provided for an exoskeleton, the exoskeleton needs to infer the motion intention of the person through a sensor, so that the person follows the motion of a human body, and the human-machine fusion system can work better in coordination, so that the core is to solve the problems of intention perception and cooperative control. The exoskeleton controller needs to recognize human body movement intention to drive the exoskeleton to move, movement information is obtained through movement data collected by the sensor systems, and movement characteristics are different due to different movement data collected by different sensor systems, so that methods for sensing the movement intention are different.
However, the existing exercise intention recognition method has the following problems: the method aims at the problem that human-computer interaction information from a single motion sensor is too comprehensive and cannot effectively express information such as motion intention and state of a human-computer system in the interaction process. Secondly, in the aspect of understanding and predicting human motion intention, the existing method adopts a single motion sensor to collect data, judges the motion state of multiple directions at present, and cannot accurately obtain future motion information, so that errors caused by mechanical delay cannot be compensated. In the aspect of surface electromyogram signal application, the existing algorithm structure is too simple, and electromyogram containing future movement intention information cannot be completely extracted. And fourthly, generating various information such as surface electromyogram signals, joint angle information and the like by utilizing the same behavior in the human-computer system interaction process, wherein the information is crucial to accurately estimating the human-computer motion state and the behavior intention of a rehabilitee, but the human motion intention really required by the power-assisted exoskeleton is used for identifying which information is contained but has no specific concept, and only aims at a certain specific motion amount such as joint angle information and the like at present. Therefore, when interpreting the connotation of the human body intention, the calculation efficiency of the real-time intention is considered, and after the motion state of the human-computer system and the human body movement intention are obtained, how to realize rapid and accurate sharing, decision making and execution is achieved, and related research is less.
Disclosure of Invention
In order to overcome the defects in the prior art, the inventor of the invention carries out intensive research and provides a power-assisted exoskeleton movement intention and gait planning method based on multi-sensor information, which is used for power-assisted exoskeleton robot movement intention identification, efficiency grading and real-time gait planning. The invention collects surface electromyographic signals (sEMG) directly related to muscle fatigue and human motion information, oxygen consumption data for sEMG biological signal calibration and an inertial sensor (IMU) kinematic signal, extracts characteristic parameters of the data to jointly form a motion mode multi-element characteristic vector, uses the motion mode multi-element characteristic vector as training data of an LSTM deep learning algorithm model, establishes the relationship between the sEMG signal and a motion mode, a gait plan and an assistance level, and can output three parameters which are most important for exoskeletal control of the human motion mode, the gait plan and the assistance level according to the relationship to control an exoskeleton robot to realize real-time follow-up and provide assistance, thereby completing the invention.
The technical scheme provided by the invention is as follows:
a power-assisted exoskeleton movement intention and gait planning method based on multi-sensing information comprises the following steps:
and step 3: motion mode model generation based on the LSTM algorithm: inputting the motion mode multi-primitive characteristic parameters into a training model, establishing an LSTM deep learning network, obtaining the corresponding relation between the surface electromyogram signal characteristic parameters and gait plan, the surface electromyogram signal characteristic parameters and exoskeleton assistance grade, and the surface electromyogram signal characteristic parameters and motion mode, and outputting the assistance grade, the motion mode and the gait plan as the model, wherein the gait plan comprises a support state and a pendulum state.
Further, in step 1, an acceleration signal in the vertical direction of the leg of the human body is acquired through an inertial sensor mounted on the leg of the calf.
Further, in step 1, acceleration signals and human oxygen consumption data of the leg of the human body in the vertical direction are acquired, single-leg data or double-leg data can be acquired when surface electromyographic signals of muscles of the leg are acquired, and each leg data is processed independently when each piece of data of the double legs is acquired.
Further, in step 2, in the extracting of the characteristic parameters, the initial points of the support state and the swing state are judged according to the change rule of the acceleration of the gait support state and the swing state in the vertical direction, and the support state and the swing state are assigned to two different rational numbers, that is, the acceleration signal characteristic parameters are obtained.
Further, in step 2, the absolute value integral average iemg), standard deviation rms, median frequency mf, average frequency mpf of the surface electromyographic signals of at least two channels of the eight channels of the semitendinosus, rectus femoris, vastus lateralis, peroneus longus, biceps femoris, tibialis anterior, gastrocnemius and soleus are selected as the characteristic parameters of the surface electromyographic signals based on the maximum value cwt of the wavelet coefficient.
Further, in step 2, the electromyographic signal integral value and standard deviation of the rectus femoris, the electromyographic signal average frequency and integral value of the gastrocnemius, the maximum wavelet coefficient value and the median frequency of the electromyographic signal of the tibialis anterior muscle are selected as surface electromyographic signal characteristic parameters.
Further, in the step 2, a moving windowing method is adopted to extract the characteristic parameters of the surface electromyographic signal in a segmented manner, when the sampling frequency is 2000-4000 Hz, the window length range is 500-1500 points, and the window displacement range is 10-50 points.
Further, in step 2, before feature parameter extraction is performed on the surface electromyographic signals of the leg muscles, filtering is performed on the surface electromyographic signals by adopting a wavelet filtering method.
Further, in step 3, different levels in the exoskeleton assistance levels correspond to different assistance efficiencies, and the assistance efficiencies are expressed as percentages of the maximum output torque of the exoskeleton.
According to the assistance exoskeleton movement intention and gait planning method based on multi-sensor information, the assistance exoskeleton movement intention and gait planning method has the following beneficial effects:
(1) the invention provides a human motion intention method comprising three dimensions of motion mode, assistance grade and gait planning, which can comprehensively give key parameters required by an assistance exoskeleton control system, improves the resolving efficiency of real-time intention, and can realize quick and accurate sharing, decision and execution;
(2) according to the invention, data obtained based on surface electromyographic signals, inertial sensors and oxygen consumption testing equipment are used as LSTM model input for model training, and compared with a traditional mode in which single man-machine interaction information from a motion sensor is used as input, the motion intention, state and other information of a man-machine system in an interaction process can be effectively expressed;
(3) according to the invention, LSTM model training is carried out based on the surface electromyographic signal, the surface electromyographic signal is a bioelectricity signal, the human motion intention is estimated by the bioelectricity signal in advance for 30-150 ms, the future motion information can be accurately predicted, and the time error caused by mechanical delay is compensated.
Drawings
Fig. 1 shows a flow chart of a method for assisting exoskeleton movement intent and gait planning based on multi-sensory information in accordance with the present invention;
FIG. 2 shows the position of attachment of the inertial sensor to the calf and the coordinate orientation;
FIG. 3 shows the change in vertical acceleration during one gait cycle;
FIG. 4 shows IMU gait feature results while walking;
FIG. 5 shows IMU gait feature results when ascending stairs;
fig. 6 shows filtered values of sEMG signals of each muscle and IMU gait characteristic parameters;
FIG. 7 shows preferred characteristic variables of a surface electromyographic signal;
FIG. 8 shows a schematic diagram of motion modality model generation based on the LSTM algorithm;
fig. 9 (a) shows the gait planning result, and (b) shows the error between the predicted value and the true value.
Detailed Description
The features and advantages of the present invention will become more apparent and appreciated from the following detailed description of the invention.
The invention provides a power-assisted exoskeleton movement intention and gait planning method based on multi-sensor information, which comprises the following steps as shown in figure 1:
the invention adopts the exoskeleton robot with the assistance of the lower limb and knee joint as a research object, and can acquire the kinematic parameters by acquiring the data of the inertial sensors (IMU) at the positions of the lower legs at two sides or one side. The inertial sensor can simultaneously measure vector change in any direction of space and decompose the vector change into component change in x, y and z directions. As shown in fig. 2, the x-axis is the vertical direction of the test object, the upward direction is positive, the y-axis is the forward direction, the forward direction is positive, the z-axis is the left-right direction of the movement, and the left direction is positive.
When the human body keeps an upright static state, the coordinate axes of the inertial sensor are superposed with the standard coordinate system, the acceleration of the x axis is equal to the gravity acceleration g, and the accelerations in the y axis direction and the z axis direction are equal to 0. When a tester moves, the three-axis output value of the inertial sensor changes, wherein the numerical value in the x-axis direction represents the vertical-direction acceleration value generated by the movement of the lower leg during the walking process of the tester, the numerical value in the y-axis direction represents the acceleration value in the advancing direction, and the numerical value in the z-axis direction represents the lateral acceleration value generated by the swinging in the left and right directions, and the acceleration data of the tester is obtained by integrating the values.
The inertial sensor is preferably arranged on the leg of a lower leg, and when a human body walks or goes upstairs and downstairs and the like, the acceleration amplitude in the x direction is larger than the acceleration amplitudes in the y direction and the z direction, for example, when the human body naturally walks, the acceleration amplitude of the lower leg in the x direction is-1.7-3.3 g, and the change is obvious, so that the acceleration signal in the x direction is selected for gait judgment.
In the present invention, surface electromyographic signals are applied to at least two of eight channels, one or both Semitendinosus (ST), rectus femoris (VR), lateral femoral muscle (VL), gastrocnemius (PL), biceps femoris (BM), Tibialis Anterior (TA), Gastrocnemius (GM), and Soleus (SM).
In the invention, when each data in the two legs is collected, each leg data is processed independently.
establishing an oxygen consumption grading standard, and extracting characteristic parameters of the acquired oxygen consumption data based on the oxygen consumption grading standard;
extracting characteristic parameters of surface electromyographic signals of leg muscles to obtain characteristic parameters of the surface electromyographic signals;
the acceleration signal characteristic parameter, the characteristic parameter of oxygen consumption data and the surface electromyographic signal characteristic parameter are jointly used as a motion mode multi-element characteristic parameter and used as the input of a training model.
After studying collected calf IMU data, the inventor finds that the amplitude change of the acceleration signal is directly related to the walking speed of a tester, the large waveform change represents the high walking speed, and conversely, the small waveform change represents the low walking speed. The acceleration signal in the x-axis direction is most characterized as being significant and more stable than the acceleration in the y-axis and z-axis directions. In addition, a gait cycle, defined as the time it takes for a side heel to land on the ground until the side heel lands again, is divided into two phases, the support phase and the swing phase. Taking the right leg as an example, fig. 3 shows a complete gait cycle with acceleration signals in the x-axis direction as scales, the support phase including standing flexion, standing extension, pre-swing, and the swing phase including swing flexion and swing extension. The starting point of the support phase is heel landing, and the starting point of the swing phase is toe off.
The tester walks and goes upstairs at a speed of 5km/h, acquires the acceleration of the x axis of the inertial sensor as shown in fig. 4(a) and 5(a), extracts characteristic parameters according to the change rule of the acceleration of the gait support state and the swing state in the x direction, judges the starting points of the support state and the swing state, assigns the support state to 0 and assigns the swing state to 1 (or other two different rational numbers), and obtains the curves shown in fig. 4(b) and 5 (b). Because the acceleration information obtained by the inertial sensor has larger time delay during real-time control, the gait phase can not be distinguished by the inertial sensor alone.
In the invention, the oxygen consumption index can be used for representing muscle fatigue, and different muscle fatigue degrees correspond to different grades of boosting efficiency. Through evaluation, the oxygen consumption is high, the muscle fatigue of the human body is high, and the required power assisting efficiency is high. The method establishes an oxygen consumption grade division standard, extracts characteristic parameters of the acquired oxygen consumption data based on the oxygen consumption grade division standard, uses the characteristic parameters as input of model training, is used for calibration of the assistance grade of sEMG signal extraction, and further cooperates with sEMG signal characteristics to implement assistance grade discrimination.
Because the sEMG signal has strong randomness, information capable of effectively representing the type of muscle activity is often fused with interference in the acquired electromyographic signals, and useful signals need to be extracted, identified and utilized. For the sEMG signals acquired by multiple channels, on the premise of comprehensively considering the practical application effect and the calculation complexity, the surface electromyogram signal characteristics comprise time domain characteristic parameters based on absolute value integral mean (iemg) and standard deviation (rms), frequency domain characteristic parameters based on median frequency (mf) and mean frequency (mpf), and time-frequency domain characteristic parameters based on the maximum value (cwt) of the wavelet coefficient.
In the invention, the sEMG signal characteristic parameters are extracted in a segmented manner by adopting a time-frequency domain moving windowing method (data point number of parameter Fourier transform), and whether the prediction real-time performance and precision requirements can be met in time is directly determined by proper channels, window lengths and window shift. The filtered value of the sEMG signal after wavelet filtering and the IMU gait characteristic parameter are shown in fig. 6, and it can be seen from fig. 6 that semitendinosus muscle, biceps femoris muscle are in phase with tibialis anterior muscle, rectus femoris muscle, vastus femoris muscle are in phase with vastus medialis muscle, and gastrocnemius muscle are in phase with gastrocnemius muscle, and according to the strength of the sEMG signal of the muscle, the sEMG signals of three muscles of the rectus femoris muscle, tibialis anterior muscle and gastrocnemius muscle are preferably extracted to obtain the sEMG signal characteristic parameter. When the sampling frequency of the electromyographic signals is 2000-4000 Hz, the window length range is 500-1500 points, and the window displacement range is 10-50 points. More preferably, the sEMG signal characteristic parameters include an integral value and a standard deviation of the electromyographic signal of the rectus femoris, an average frequency and an integral value of the electromyographic signal of the gastrocnemius, and a maximum value and a median frequency of the wavelet coefficient of the electromyographic signal of the tibialis anterior, and the periodicity and the smoothness of the curves of the characteristic parameters are superior to those of other characteristic parameters, as shown in fig. 7.
And step 3: motion mode model generation based on the LSTM algorithm: inputting the multi-element characteristic parameters of the motion mode into a training model, establishing an LSTM deep learning network, obtaining the corresponding relation between the sEMG signal characteristic parameters and gait planning, between the sEMG signal characteristic parameters and exoskeleton assistance levels, and between the sEMG signal characteristic parameters and the motion mode, and outputting the assistance levels, the motion mode and the gait planning as the model, wherein the gait planning comprises a support state and pendulum dynamics.
The long-time memory (LSTM) network is a deep learning model, and the structure of the LSTM network has good time series modeling expression capacity. The LSTM deep network is adopted for modeling, sequence information with time correlation and space correlation in the sEMG signals can be effectively extracted, and the model can keep strong generalization capability, so that the requirement of exoskeleton control is met. The specific process is shown in fig. 8, the sEMG signal characteristic quantity in the multi-element characteristic vector of the motion mode is used as the input of the LSTM algorithm, and the corresponding relations between the sEMG signal characteristic quantity and the gait plan, between the sEMG signal characteristic quantity and the exoskeleton power-assisted level, and between the sEMG signal characteristic quantity and the motion mode are respectively established, that is, an intention decoding numerical value with three dimensions of the motion mode, the gait characteristic and the power-assisted level is output, and the intention decoding numerical value can be directly applied to exoskeleton motion control.
The corresponding relation between the sEMG signal characteristic parameters and the exoskeleton assistance levels is established through an LSTM algorithm model, the exoskeleton is guided to provide corresponding auxiliary torque for a human body, and a division mode of the exoskeleton assistance levels is shown in Table 1. Different levels in the exoskeleton assistance levels correspond to different assistance efficiencies, the assistance efficiencies are related to muscle fatigue degrees, the muscle fatigue degrees are represented through oxygen consumption data, corresponding assistance efficiencies should be given under different muscle fatigue degrees, otherwise, the assistance efficiency is low, the assistance effect is not obvious, the assistance efficiency is large, radical behaviors are generated on a human body, and discomfort of the human body is caused. In the present invention, the efficiency of the boost is expressed as a percentage of the maximum output torque of the exoskeleton.
TABLE 1 assistance levels for lower extremity exoskeleton
In the present invention, suitable motion modes include, but are not limited to, uphill walking, downhill walking, upstairs, downstairs, flat-bottomed walking, and the like.
Examples
The assistance exoskeleton movement intention and gait planning method based on multi-sensing information provided by the invention is adopted to carry out actual test. In the experiment, a tester walks on a running machine at a speed of 5km/h and a gradient of 10 degrees, 4 groups of experimental data are collected, each group is about 1min, 3 groups of the 4 groups of data are selected as training data, and the other 1 group of data are used as test data to predict the assistance level, the motion mode and the gait planning. Fig. 9 shows the training result of selecting the combination of the electromyographic signal integral value and standard deviation of the rectus femoris, the electromyographic signal average frequency and integral value of the gastrocnemius, the maximum value of the wavelet coefficient of the electromyographic signal of the tibialis anterior and the median frequency characteristic value, and respectively gives the comparison between the model measured value and the predicted value of the combined data set and the corresponding error value. The model predicts that the movement mode is walking on an uphill slope, and as can be seen from fig. 9, the predicted value is about 50ms before the measured value, so that the mechanical and communication delay in the exoskeleton control system can be compensated, and the exoskeleton system can follow the human body without time difference.
The invention has been described in detail with reference to specific embodiments and illustrative examples, but the description is not intended to be construed in a limiting sense. Those skilled in the art will appreciate that various equivalent substitutions, modifications or improvements may be made to the technical solution of the present invention and its embodiments without departing from the spirit and scope of the present invention, which fall within the scope of the present invention. The scope of the invention is defined by the appended claims.
Those skilled in the art will appreciate that those matters not described in detail in the present specification are well known in the art.
Claims (9)
1. A multi-sensor information based power-assisted exoskeleton movement intention and gait planning method is characterized by comprising the following steps:
step 1, acquiring human intention information data: acquiring acceleration signals of the legs of a human body in the vertical direction and oxygen consumption data of the human body, and acquiring surface electromyographic signals of muscles of the legs;
step 2, generating a motion mode multi-element characteristic parameter: extracting characteristic parameters according to the change rule of the acceleration of the gait support state and the swing state in the vertical direction to obtain acceleration signal characteristic parameters; establishing an oxygen consumption grading standard, and extracting characteristic parameters of the acquired oxygen consumption data based on the oxygen consumption grading standard; extracting characteristic parameters of surface electromyographic signals of leg muscles to obtain characteristic parameters of the surface electromyographic signals; taking the acceleration signal characteristic parameter, the characteristic parameter of oxygen consumption data and the surface electromyographic signal characteristic parameter as the motion mode multi-element characteristic parameter as the input of the training model;
and step 3: motion mode model generation based on the LSTM algorithm: inputting the motion mode multi-primitive characteristic parameters into a training model, establishing an LSTM deep learning network, obtaining the corresponding relation between the surface electromyogram signal characteristic parameters and gait plan, the surface electromyogram signal characteristic parameters and exoskeleton assistance grade, and the surface electromyogram signal characteristic parameters and motion mode, and outputting the assistance grade, the motion mode and the gait plan as the model, wherein the gait plan comprises a support state and a pendulum state.
2. The method according to claim 1, wherein in step 1, the acceleration signal of the human leg in the vertical direction is obtained by an inertial sensor mounted on the leg of the lower leg.
3. The method according to claim 1, wherein in step 1, acceleration signals of the human leg in the vertical direction and human oxygen consumption data are acquired, single-leg data or double-leg data can be acquired when surface electromyographic signals of leg muscles are acquired, and each leg data is processed independently when each piece of double-leg data is acquired.
4. The method according to claim 1, wherein in step 2, in the extracting of the characteristic parameters according to the change rule of the acceleration in the vertical direction of the gait support state and the swing state, the starting points of the support state and the swing state are judged, and the support state and the swing state are assigned to two different rational numbers, that is, the acceleration signal characteristic parameters are obtained.
5. The method as claimed in claim 1, wherein in step 2, the integral mean iemg, standard deviation rms, median frequency mf, mean frequency mpf of the absolute values of the surface electromyographic signals of at least two channels of the eight channels of semitendinosus, rectus femoris, vastus lateralis, longus peroneus, biceps femoris, tibialis anterior, gastrocnemius and soleus are selected as the characteristic parameters of the surface electromyographic signals based on the maximum value cwt of the wavelet coefficient.
6. The method according to claim 5, wherein in step 2, the integral value and standard deviation of the electromyographic signals of the rectus femoris, the average frequency and integral value of the electromyographic signals of the gastrocnemius, the maximum value and the median frequency of the wavelet coefficients of the electromyographic signals of the tibialis anterior muscle are selected as the surface electromyographic signal characteristic parameters.
7. The method according to claim 5, wherein in the step 2, a moving windowing method is adopted to extract the surface electromyogram signal characteristic parameters in a segmented manner, when the sampling frequency is 2000-4000 Hz, the window length range is 500-1500 points, and the window shift amount range is 10-50 points.
8. The method according to claim 1, wherein in step 2, the surface electromyographic signals of the leg muscles are filtered by a wavelet filtering method before being subjected to feature parameter extraction.
9. The method of claim 1, wherein in step 3, different levels of exoskeleton assistance correspond to different assistance efficiencies, the assistance efficiencies being expressed as a percentage of the maximum output torque of the exoskeleton.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011379078.0A CN112336590A (en) | 2020-11-30 | 2020-11-30 | Power-assisted exoskeleton movement intention and gait planning method based on multi-sensing information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011379078.0A CN112336590A (en) | 2020-11-30 | 2020-11-30 | Power-assisted exoskeleton movement intention and gait planning method based on multi-sensing information |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112336590A true CN112336590A (en) | 2021-02-09 |
Family
ID=74366227
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011379078.0A Pending CN112336590A (en) | 2020-11-30 | 2020-11-30 | Power-assisted exoskeleton movement intention and gait planning method based on multi-sensing information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112336590A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112842825A (en) * | 2021-02-24 | 2021-05-28 | 郑州铁路职业技术学院 | Training device for lower limb rehabilitation recovery |
| CN113180643A (en) * | 2021-04-25 | 2021-07-30 | 燕山大学 | Exoskeleton assistance detection device and evaluation method thereof |
| CN113261974A (en) * | 2021-06-07 | 2021-08-17 | 吉林大学 | Sports fatigue monitoring method based on multiple physiological signals |
| CN113829339A (en) * | 2021-08-02 | 2021-12-24 | 上海大学 | Exoskeleton movement coordination method based on long-time and short-time memory network |
| CN114948609A (en) * | 2022-04-12 | 2022-08-30 | 北京航空航天大学 | Walking aid auxiliary device and method for paralytic |
| CN115687898A (en) * | 2022-12-30 | 2023-02-03 | 苏州大学 | Gait parameter adaptive fitting method based on multi-mode signals |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103536293A (en) * | 2013-11-06 | 2014-01-29 | 杭州共远科技有限公司 | Method for limiting time mode gait data analysis of gait training correction instrument measuring system |
| CN103984962A (en) * | 2014-05-30 | 2014-08-13 | 河北工业大学 | Exoskeleton walking mode identification method based on electromyographic signals |
| CN107397649A (en) * | 2017-08-10 | 2017-11-28 | 燕山大学 | A kind of upper limbs exoskeleton rehabilitation robot control method based on radial base neural net |
-
2020
- 2020-11-30 CN CN202011379078.0A patent/CN112336590A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103536293A (en) * | 2013-11-06 | 2014-01-29 | 杭州共远科技有限公司 | Method for limiting time mode gait data analysis of gait training correction instrument measuring system |
| CN103984962A (en) * | 2014-05-30 | 2014-08-13 | 河北工业大学 | Exoskeleton walking mode identification method based on electromyographic signals |
| CN107397649A (en) * | 2017-08-10 | 2017-11-28 | 燕山大学 | A kind of upper limbs exoskeleton rehabilitation robot control method based on radial base neural net |
Non-Patent Citations (1)
| Title |
|---|
| 郭子铭等: "基于长短期记忆模型的外骨骼实时步态分类", 《广西大学学报》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112842825A (en) * | 2021-02-24 | 2021-05-28 | 郑州铁路职业技术学院 | Training device for lower limb rehabilitation recovery |
| 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 |
| CN113261974A (en) * | 2021-06-07 | 2021-08-17 | 吉林大学 | Sports fatigue monitoring method based on multiple physiological signals |
| CN113829339A (en) * | 2021-08-02 | 2021-12-24 | 上海大学 | Exoskeleton movement coordination method based on long-time and short-time memory network |
| CN113829339B (en) * | 2021-08-02 | 2023-09-15 | 上海大学 | Exoskeleton movement coordination method based on long-short-term memory network |
| CN114948609A (en) * | 2022-04-12 | 2022-08-30 | 北京航空航天大学 | Walking aid auxiliary device and method for paralytic |
| CN115687898A (en) * | 2022-12-30 | 2023-02-03 | 苏州大学 | Gait parameter adaptive fitting method based on multi-mode signals |
| CN115687898B (en) * | 2022-12-30 | 2023-07-11 | 苏州大学 | Gait parameter self-adaptive fitting method based on multi-mode signals |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112336590A (en) | Power-assisted exoskeleton movement intention and gait planning method based on multi-sensing information | |
| CN112754468B (en) | Human body lower limb movement detection and identification method based on multi-source signals | |
| CN110141239B (en) | Movement intention recognition and device method for lower limb exoskeleton | |
| CN110537922B (en) | Human body walking process lower limb movement identification method and system based on deep learning | |
| CN110916679B (en) | Human body lower limb pose gait detection device and method | |
| CN109259739B (en) | Myoelectricity estimation method of wrist joint movement moment | |
| CN106156524A (en) | A kind of online gait planning system and method for Intelligent lower limb power assisting device | |
| CN104983489B (en) | Road conditions recognition methods during artificial leg walking | |
| CN111643321B (en) | Exoskeleton joint angle prediction method and system based on sEMG signals | |
| CN108549856A (en) | A kind of human action and road conditions recognition methods | |
| CN108338791A (en) | The detection device and detection method of unstable motion data | |
| CN110236550A (en) | A kind of body gait prediction meanss based on multi-modal deep learning | |
| CN111506189B (en) | Motion mode prediction and switching control method for complex motion of human body | |
| Ahamed et al. | Fuzzy inference system-based recognition of slow, medium and fast running conditions using a triaxial accelerometer | |
| CN113043248B (en) | Transportation and assembly whole-body exoskeleton system based on multi-source sensor and control method | |
| CN110319840A (en) | Conjugate gradient attitude algorithm method towards abnormal gait identification | |
| CN112487902B (en) | Exoskeleton-oriented gait phase classification method based on TCN-HMM | |
| Li et al. | An enhanced EMG-driven Musculoskeletal model based on non-negative matrix factorization | |
| Li et al. | Prediction of knee joint moment by surface electromyography of the antagonistic and agonistic muscle pairs | |
| CN208591046U (en) | The detection device of unstable motion data | |
| Jin | Design of intelligent perception module based on wireless sensor network and basketball sports attitude | |
| Zhang et al. | Continuous estimation of knee angles from decomposition of single channel surface electromyography signals | |
| Du et al. | Design on exoskeleton robot intellisense system based on multi-dimensional information fusion | |
| Chen et al. | A novel gait prediction method for soft exosuit base on limit cycle and neural network | |
| Palaniappan et al. | Averaged EMG Profile and Artificial Neural Network Based Walking Speed Estimation by GAIT Analysis |
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 | ||
| CB03 | Change of inventor or designer information | ||
| CB03 | Change of inventor or designer information |
Inventor after: Chen Jing Inventor after: Yu Zhiyuan Inventor after: Huang Yuping Inventor after: Zhu Xiaorong Inventor after: Tao Yunfei Inventor before: Chen Jing Inventor before: Yu Zhiyuan Inventor before: Huang Yuping Inventor before: Zhu Xiao Inventor before: Tao Yunfei |