CN112932898B - On-demand auxiliary rehabilitation robot based on Bayesian optimization - Google Patents
On-demand auxiliary rehabilitation robot based on Bayesian optimization Download PDFInfo
- Publication number
- CN112932898B CN112932898B CN202110119541.6A CN202110119541A CN112932898B CN 112932898 B CN112932898 B CN 112932898B CN 202110119541 A CN202110119541 A CN 202110119541A CN 112932898 B CN112932898 B CN 112932898B
- Authority
- CN
- China
- Prior art keywords
- robot
- training
- round
- auxiliary
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- 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
- A61H1/00—Apparatus for passive exercising; Vibrating apparatus ; Chiropractic devices, e.g. body impacting devices, external devices for briefly extending or aligning unbroken bones
- A61H1/02—Stretching or bending or torsioning apparatus for exercising
- A61H1/0274—Stretching or bending or torsioning apparatus for exercising for the upper limbs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic 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
- A61H2201/50—Control means thereof
- A61H2201/5007—Control means thereof computer controlled
Abstract
A Bayesian optimization-based on-demand auxiliary rehabilitation robot training method comprises the steps of collecting performance indexes of a user in a random preset track following task for many times, evaluating the performance of the user in each round in a cost function mode, carrying out on-demand auxiliary algorithm hyperparametric optimization based on a Bayesian optimization algorithm and the task performance before the user in each subsequent round of tasks, and adjusting the on-demand auxiliary effect of the next task. Because the difference of the exercise capacities of different testees is large, and the learning capacities in the training process are different, the exercise capacity of the user cannot be adapted in a traditional training mode, in order to improve the rehabilitation training effect of the testees and improve the adaptability of rehabilitation strategies to different testees, the method can effectively improve the variability of tasks, can more quickly realize the adaptability of the equipment to the exercise capacity of the user through a shorter cycle process, reduces the workload of therapists, and effectively improves the rehabilitation efficiency.
Description
Technical Field
The invention relates to the technical field of rehabilitation robots, rehabilitation training and machine learning, in particular to an on-demand auxiliary rehabilitation robot based on Bayesian optimization.
Background
According to statistics, the incidence of stroke in China continuously increases in the last 30 years, the incidence of stroke rapidly rises along with the continuous acceleration of social aging and urbanization processes, the proportion of stroke accounting for disease death of residents in China is over 20 percent at present, more than 75 percent of survivors suffer from functional disorders such as hemiplegic paralysis and the like, and the limb dysfunction directly affects the daily life activities of patients. A large number of clinical verifications show that the exercise rehabilitation therapy, namely, the rehabilitation physician can assist the affected limb for a plurality of times for a long time to exercise the affected limb, and has positive effects on the exercise capacity rehabilitation of the patient. However, the one-to-one rehabilitation method is not universal due to the complex rehabilitation method, high repeatability, high cost and the like, and a comprehensive rehabilitation plan which provides long-term, stable, quantitative and accurate motion stimulation for the upper limbs of the stroke patient is one of important contents of rehabilitation treatment. In order to and solve the problem of upper limb rehabilitation therapy, the concept of a rehabilitation robot has therefore been proposed. Robot-assisted rehabilitation training has been shown by researchers to reduce athletic injuries and to have the effect of improving athletic performance. Rehabilitation therapy is gradually emphasized by people, more new rehabilitation equipment is continuously developed, new rehabilitation therapy technology is gradually integrated into a rehabilitation therapy scheme, and the application of the robot technology in the aspect of rehabilitation therapy is deeper.
The on-demand auxiliary strategy is a rehabilitation strategy proposed by researchers in recent years, and is helpful for inducing neural plasticity based on the improvement of the participation degree of a user in the training process, and aims to stimulate the participation of the user by assisting arm movement and accordingly improve the strength of rehabilitation therapy.
Therefore, in 2020.03.24, the applicant applies for a Bayesian optimization-based method for adjusting the complexity of the reference trajectory for active training of the upper limb rehabilitation robot, and the invention discloses a Bayesian optimization-based method for adjusting the complexity of the reference trajectory for active training of the upper limb rehabilitation robot, which sequentially comprises the following steps: establishing a training task model, selecting an athletic performance index, determining task track complexity, learning a functional relation between the athletic performance index and a task track complexity parameter, and realizing self-adaptive adjustment of a training task. Because the exercise learning ability of the tested person is likely to change in the training process, the invention monitors the exercise performance of the tested person in each round of training, adaptively adjusts the difficulty of the training task according to the exercise performance of the tested person, increases the variability for the task and can improve the rehabilitation training effect of the tested person. The invention monitors the motion performance of different testees, can realize individualized self-adaptive adjustment of the training task for each tester, can reduce tedious and fussy workload of a therapist/technician for adjusting the task difficulty for different testees in real time, and improves the rehabilitation training efficiency. When the designed reference track is fixed, only the on-demand auxiliary position control parameters are adjusted, the result of any experimental process does not have any influence on the track, and the on-demand auxiliary related content is not mentioned, the form and the calculation method of the joint on-demand auxiliary force field control are important points, the application describes that the training to be tested is based on a teaching active training form, the optimization target is carried out according to the track complexity, the drawing of the track in the auxiliary force field according to the test is not considered as the basis, the track following error of the user in the process and the work done in the process are used as evaluation indexes, and the optimization of the strength of the auxiliary force field according to the evaluation indexes is not considered.
Disclosure of Invention
In order to solve the problems, the invention provides an on-demand auxiliary rehabilitation robot based on Bayesian optimization, the method can effectively improve the variability of tasks, and can more quickly realize the adaptability of equipment to the user movement capacity through a shorter cyclic process, thereby reducing the workload of therapists and effectively improving the rehabilitation efficiency.
The invention provides an on-demand auxiliary rehabilitation robot based on Bayesian optimization, which comprises the following specific steps:
step 1, designing a track following task: the target track is designed into a quasi-sinusoidal curve formed by two semicircles, but only five reference points are uniformly displayed on the track in the whole track to serve as guide points, and the tail end of the robot is controlled to follow the track according to the reference points in the process of the operation to be tested;
step 2, self-adaptive auxiliary strategy according to needs: in the training process, the robot self-adaptively adjusts the force for correcting the track according to the track error, and according to the position error delta d of the tested robot in the operation process, the force and the direction applied by the robot to a user are adjusted in real time by the correction force according to an auxiliary strategy as required;
step 3, evaluation mechanism: after each round of training is finished, an evaluation mechanism is established, and the position data of the current round of operation and the tested machine are checkedMeasuring the force of the robot end operation and the auxiliary force of the robot, calculating the evaluation result according to the evaluation function, and recording the result as J i ;
Step 4, random process: in the random process, a trajectory tracking task is performed for a plurality of times according to the design, in each training round, the hyper-parameter lambda is a random value in a selected range, namely, the adaptive adjustment mechanism is different in each round of the random process, and the most real user performance can be obtained according to the user capacity under the condition of different assistance;
step 5, Bayesian optimization process: after the random process is finished, according to the previous n times of expression of the user and the hyper-parameter corresponding result D ← (lambda) of the auxiliary strategy on demand i ,J i ) Bayesian optimization is carried out, the numerical value of the optimal hyperparameter lambda of the next on-demand auxiliary strategy is obtained, next training is carried out according to the numerical value, after a new round of evaluation result is obtained, new data are merged into the data set D, then Bayesian optimization is carried out according to the new data set to obtain the hyperparameter lambda of the next round, and after the process is repeated for a plurality of times, the training is finished.
As a further improvement of the present invention, the adaptive on-demand assistance strategy described in step 2 is as follows:
f=f max *[1-exp(-(Δd/λ) 2 )]
wherein lambda is a hyper-parameter of an auxiliary strategy according to needs, delta d is a deviation distance between the position of the tail end of the robot operated by the tested and a preset track, and theta is an included angle between a connecting line of the tail end of the robot operated by the tested and the circle center of the current semicircular track and a horizontal plane, and a method for calculating the output force of the tail end of the robot according to the track where the tail end is located comprises the following steps:
f x =f*cosθ*sig(f)
f y =f*sinθ*sig(f)
wherein sig (f) is the symbolic value of the adjustment force of the robot.
As a further improvement of the present invention, in the evaluation mechanism described in step 3, the operation evaluation to be tested is divided into two parts: the method comprises the following steps of (1) calculating a tested track following error and work of a user in the operation process, wherein the method comprises the following steps:
wherein y is i Is the longitudinal axis value of the actual position, y e The expected longitudinal axis value corresponding to each position;
the work calculation method in the operation process of the tested object is as follows:
E u =E h -E r
wherein E h Work to be performed on the end of the robot, E r The robot does work on the tail end grab handle in the following calculation modes:
for the operation performance of each round of the tested object, data quantization is carried out on the operation performance through an evaluation function, and the calculation mode is as follows:
J=E u -β*S err 。
as a further improvement of the invention, in the Bayesian optimization process described in step 5, the performance of the tested object in each round of the fixed stage is recorded and the evaluation result is calculated by the method of the evaluation mechanism described in step 3, and lambda is calculated in each round i ,J i All the data samples are recorded and stored in a data sample set D, the data samples are fitted through a fitting model, then the corresponding J maximum value of lambda in a value range is obtained according to the fitted model, the lambda value is used as a hyper-parameter in an on-demand auxiliary training strategy in the next round of training, the next round of training is carried out, a new J value is obtained after the training, the lambda and the J of the new round are merged into the data sample set D, the new lambda value is obtained through fitting again according to a new data set and is used as the hyper-parameter of the next round, and the new lambda value is obtained for multiple timesThe optimization training process is repeated until the training is finished.
The invention relates to an on-demand auxiliary rehabilitation robot based on Bayesian optimization, which has the following specific design points:
1. the self-adaptive adjustment strategy is suitable for the training task of the upper limb rehabilitation robot, and self-adaptive adjustment for assisting rehabilitation according to needs is carried out according to the performance of a user, so that the robot can adapt to the human;
2. by adopting a Bayesian optimization mode, an auxiliary strategy more suitable for a user can be explored in a shorter training period, excessive cycle periods are not needed, the burden of the user is reduced, and a useless training process is avoided as much as possible;
3. compared with the traditional method, the on-demand auxiliary rehabilitation robot training method based on Bayesian optimization can improve the motion capability of the user in a short time more quickly, stimulate the neural perception of the user on force and motion control, promote the participation of the patient better and improve the rehabilitation interest of the user.
Drawings
FIG. 1 is a schematic diagram of an algorithm framework;
fig. 2 is a schematic diagram of a trajectory following task and assisting force adjustment.
Detailed Description
The invention is described in further detail below with reference to the following detailed description and accompanying drawings:
the invention provides an on-demand auxiliary rehabilitation robot based on Bayesian optimization, which can effectively improve the task variability, can more quickly realize the adaptability of equipment to the user movement capacity through a shorter cycle process, reduce the workload of therapists and effectively improve the rehabilitation efficiency.
As shown in fig. 1, an adaptive on-demand training method based on bayesian optimization according to an embodiment of the present invention includes the following steps:
1. a familiarity stage:
the explanation of the trail following task to be tested is carried out by the therapist/technician, including the shape of the expected trail, the position of the guide point, the starting and stopping conditions and the like, and the user can carry out about 3-5 cycle periods in the familiar stage;
2. a random stage:
randomly selecting parameter values in a given lambda value range, substituting the parameter values into a self-adaptive auxiliary adjustment algorithm according to needs:
f=f max *[1-exp(-(Δd/λ) 2 )]
wherein lambda is a hyper-parameter of the auxiliary strategy according to needs, delta d is a deviation distance between the tail end position of the robot operated by the tested robot and a preset track, and theta is an included angle between a connecting line of the tail end of the robot operated by the tested robot and the current semicircular track circle center and a horizontal plane as the operating plane of the robot is a two-dimensional plane.
In the random stage, ensuring that the selected lambda parameters are not repeated every time, then recording the tested track coordinate data, the tested output force data, the output force of the robot and the like of each round, and after each round is finished, calculating according to an evaluation function:
J=E u -β*S err
the method for calculating the track following error of the tested track comprises the following steps:
wherein y is i Is the longitudinal axis value of the actual position, y e The desired vertical axis value for each position.
The work calculation method in the operation process of the tested object is as follows:
E u =E h -E r
wherein E h Work to be performed on the end of the robot, E r The calculation modes for the robot to do work on the tail-end grab handle are respectively as follows:
3. and (3) an optimization stage:
will randomly step each round of (lambda) i ,J i ) The data samples are recorded and stored in a data sample set D, the data samples are fitted through a fitting model (a Gaussian mixture model or a random forest and the like), then a corresponding J maximum value of lambda in a value range is obtained according to the fitted model, the lambda value is used as a hyper-parameter in an auxiliary training strategy according to needs in the next round of training, the next round of training is carried out, a new J value is obtained after the training, the new round of lambda, J is merged into the data sample set D, the new lambda value is obtained through fitting according to a new data set again and is used as the hyper-parameter of the next round, and the optimization training process is repeated for a plurality of times until the training is finished.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention in any way, but any modifications or equivalent variations made according to the technical spirit of the present invention are within the scope of the present invention as claimed.
Claims (2)
1. An on-demand auxiliary rehabilitation robot based on Bayesian optimization comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein the computer program is used for realizing the training method of the on-demand auxiliary rehabilitation robot based on Bayesian optimization when being loaded to the processor, and the method comprises the following specific steps:
step 1, designing a track following task: the target track is designed into a quasi-sinusoidal curve formed by two semicircles, but only five reference points are uniformly displayed on the track in the whole track to serve as guide points, and the tail end of the robot is controlled to follow the track according to the reference points in the process of the operation to be tested;
step 2, self-adaptive auxiliary strategy according to needs: in the training process, the robot self-adaptively adjusts the force for correcting the track according to the track error, and the force and the direction applied by the robot to a user are adjusted in real time according to the position error delta d of the tested robot in the operation process and an auxiliary strategy as required by the correction force;
the self-adaptive on-demand auxiliary strategy in the step 2 is as follows:
f=f max *[1-exp(-(△d/λ) 2 )]
wherein f is max The value is the maximum auxiliary force and is a fixed value, lambda is an over-parameter of an auxiliary strategy according to needs, delta d is a deviation distance between the position of the tail end of the robot operated by the test and a preset track, and because the operation plane of the robot is a two-dimensional plane, theta is an included angle between a connecting line of the tail end of the robot operated by the test and the current semicircular track circle center and a horizontal plane, the method for calculating the output force of the tail end of the robot according to the track of the tail end is as follows:
f x =f*cosθ*sig(f)
f y =f*sinθ*sig(f)
wherein f is x And f y The force components of the auxiliary force which are calculated in real time and decomposed in the x direction and the y direction are not fixed numerical values and change in real time according to a related calculation formula, and sig (f) is a symbol value of the adjusting force of the robot;
step 3, evaluation mechanism: after each round of training is finished, an evaluation mechanism is established, the position data of the current round of operation, the force of the tested robot end operation and the auxiliary force of the robot are measured, an evaluation result is calculated according to an evaluation function and recorded as J i ;
In the evaluation mechanism described in step 3, the operation evaluation to be tested is divided into two parts: the method for calculating the tested track following error comprises the following steps of:
wherein y is i Is the longitudinal axis value of the actual position, y e The expected longitudinal axis value corresponding to each position; the work calculation method in the operation process of the tested object is as follows:
E u =E h -E r
wherein E h Made to the end of the robot for the purpose of being testedWork, E r The robot does work on the tail end grab handle in the following calculation modes:
wherein f is rx And f ry The force components in the x direction and the y direction applied by the robot to the human in the auxiliary process are not fixed values, and for the operation performance of each round of the tested object, data quantization is carried out on the force components through an evaluation function J, wherein the calculation mode of the evaluation function is as follows:
J=E u -β*S err ;
wherein beta is a fixed value
Step 4, random process: in the random process, a trajectory tracking task is performed for a plurality of times according to the design, in each training round, the hyper-parameter lambda is a random value in a selected range, namely, the adaptive adjustment mechanism is different in each round of the random process, and the most real user performance can be obtained according to the user capacity under the condition of different assistance;
step 5, Bayesian optimization process: after the random process is finished, according to the previous n times of expression of the user and the superparameter of the auxiliary strategy as required, corresponding result data set D ← (lambda) i ,J i ) And carrying out Bayesian optimization on the data set, acquiring the value of the optimal hyperparameter lambda of the next on-demand auxiliary strategy, carrying out next training according to the value, merging new data into the data set D after acquiring a new round of evaluation result, then carrying out Bayesian optimization according to the new data set to acquire the hyperparameter lambda of the next round, and repeating the process for a plurality of times to finish the training.
2. The on-demand auxiliary rehabilitation robot based on Bayesian optimization as recited in claim 1, wherein: step 5 theIn the bayesian optimization process, the tested performance in each round of the fixed stage is recorded and the evaluation result is calculated by the method of the evaluation mechanism in step 3, and the lambda of each round i ,J i All are recorded and stored in a data sample set D, the data samples are fitted through a fitting model, and then lambda is obtained according to the fitted model i At the corresponding J maximum value in the value range, the lambda is measured i And taking the value as a hyper-parameter in an on-demand auxiliary training strategy in the next round of training, carrying out the next round of training, obtaining a new J value after the training, merging the lambda and J of the new round into the data sample set D, fitting again according to the new data set to obtain a new lambda value as the hyper-parameter of the next round, and repeating the training process for multiple times until the training is finished.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110119541.6A CN112932898B (en) | 2021-01-28 | 2021-01-28 | On-demand auxiliary rehabilitation robot based on Bayesian optimization |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110119541.6A CN112932898B (en) | 2021-01-28 | 2021-01-28 | On-demand auxiliary rehabilitation robot based on Bayesian optimization |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112932898A CN112932898A (en) | 2021-06-11 |
CN112932898B true CN112932898B (en) | 2022-09-30 |
Family
ID=76238758
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110119541.6A Active CN112932898B (en) | 2021-01-28 | 2021-01-28 | On-demand auxiliary rehabilitation robot based on Bayesian optimization |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112932898B (en) |
Family Cites Families (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8280503B2 (en) * | 2008-10-27 | 2012-10-02 | Michael Linderman | EMG measured during controlled hand movement for biometric analysis, medical diagnosis and related analysis |
JP2012524636A (en) * | 2009-04-24 | 2012-10-18 | アドバンスド ブレイン モニタリング,インコーポレイテッド | Adaptive behavior trainer |
CN105054927B (en) * | 2015-07-16 | 2017-08-15 | 西安交通大学 | The biological quantitative estimation method for degree of being actively engaged in a kind of lower limb rehabilitation system |
WO2018017436A1 (en) * | 2016-07-22 | 2018-01-25 | President And Fellows Of Harvard College | Controls optimization for wearable systems |
CN110300542A (en) * | 2016-07-25 | 2019-10-01 | 开创拉布斯公司 | Use the method and apparatus of wearable automated sensor prediction muscle skeleton location information |
US10639510B2 (en) * | 2017-03-20 | 2020-05-05 | The Trustees Of Columbia University In The City Of New York | Human musculoskeletal support and training system methods and devices |
CN110303471B (en) * | 2018-03-27 | 2021-02-09 | 清华大学 | Power-assisted exoskeleton control system and control method |
CN108681396B (en) * | 2018-04-28 | 2021-07-06 | 北京机械设备研究所 | Human-computer interaction system and method based on brain-myoelectricity bimodal neural signals |
CN109381184A (en) * | 2018-10-15 | 2019-02-26 | 刘丹 | A kind of wearable smart machine control method that auxiliary is carried |
CN110400619B (en) * | 2019-08-30 | 2023-07-21 | 上海大学 | Hand function rehabilitation training method based on surface electromyographic signals |
CN111631923A (en) * | 2020-06-02 | 2020-09-08 | 中国科学技术大学先进技术研究院 | Neural network control system of exoskeleton robot based on intention recognition |
CN111816309B (en) * | 2020-07-13 | 2022-02-01 | 国家康复辅具研究中心 | Rehabilitation training prescription self-adaptive recommendation method and system based on deep reinforcement learning |
-
2021
- 2021-01-28 CN CN202110119541.6A patent/CN112932898B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN112932898A (en) | 2021-06-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109394476B (en) | Method and system for automatic intention recognition of brain muscle information and intelligent control of upper limbs | |
CN108785997B (en) | Compliance control method of lower limb rehabilitation robot based on variable admittance | |
Zollo et al. | Quantitative evaluation of upper-limb motor control in robot-aided rehabilitation | |
CN100594867C (en) | Apparel type robot for healing hand function and control system thereof | |
CN105919796B (en) | Traditional Chinese medicine massage robot system and its acupuncture point finding method | |
CN109091819A (en) | Upper limb rehabilitation robot control system | |
Fagg et al. | Kinetic trajectory decoding using motor cortical ensembles | |
CN107378944A (en) | A kind of multi-dimensional surface electromyographic signal prosthetic hand control method based on PCA | |
CN110675933B (en) | Finger mirror image rehabilitation training system | |
CN101961527A (en) | Rehabilitation training system and method combined with functional electric stimulation and robot | |
CN114366556B (en) | Multimode training control system and method for lower limb rehabilitation | |
US20230244909A1 (en) | Adaptive brain-computer interface decoding method based on multi-model dynamic integration | |
Yan | Tai chi practice reduces movement force variability for seniors | |
Khoshdel et al. | An optimized artificial neural network for human-force estimation: consequences for rehabilitation robotics | |
Franzke et al. | Exploring the relationship between EMG feature space characteristics and control performance in machine learning myoelectric control | |
CN112932898B (en) | On-demand auxiliary rehabilitation robot based on Bayesian optimization | |
CN113180993A (en) | Exoskeleton hand rehabilitation training system based on force feedback, storage medium and terminal | |
Dimitriou | Task-dependent modulation of spinal and transcortical stretch reflexes linked to motor learning rate. | |
CN113081671B (en) | Method for improving on-demand auxiliary rehabilitation training participation degree based on Bayesian optimization | |
Schmid et al. | Effect of fatigue on the precision of a whole-body pointing task | |
Hu et al. | Stiffness optimal modulation of a variable stiffness energy storage hip exoskeleton and experiments on its assistance effect | |
CN209253488U (en) | A kind of bionical class brain intelligent hand electric mechanical ectoskeleton and its control system entirely | |
Roman et al. | A Novel Hardware and Software Interface for a Grip Force Tracking System | |
Kulwa et al. | A Multidataset Characterization of Window-Based Hyperparameters for Deep CNN-Driven sEMG Pattern Recognition | |
Viekash et al. | Deep learning based muscle intent classification in continuous passive motion machine for knee osteoarthritis rehabilitation |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |