WO2024113241A1 - Procédé et système d'estimation d'angle d'articulation de main inter-utilisateur, et dispositif informatique - Google Patents

Procédé et système d'estimation d'angle d'articulation de main inter-utilisateur, et dispositif informatique Download PDF

Info

Publication number
WO2024113241A1
WO2024113241A1 PCT/CN2022/135520 CN2022135520W WO2024113241A1 WO 2024113241 A1 WO2024113241 A1 WO 2024113241A1 CN 2022135520 W CN2022135520 W CN 2022135520W WO 2024113241 A1 WO2024113241 A1 WO 2024113241A1
Authority
WO
WIPO (PCT)
Prior art keywords
user
model
joint angle
hand
data
Prior art date
Application number
PCT/CN2022/135520
Other languages
English (en)
Chinese (zh)
Inventor
耿艳娟
龙昱丞
李光林
Original Assignee
中国科学院深圳先进技术研究院
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 中国科学院深圳先进技术研究院 filed Critical 中国科学院深圳先进技术研究院
Priority to PCT/CN2022/135520 priority Critical patent/WO2024113241A1/fr
Publication of WO2024113241A1 publication Critical patent/WO2024113241A1/fr

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/316—Modalities, i.e. specific diagnostic methods
    • A61B5/389—Electromyography [EMG]
    • A61B5/397—Analysis of electromyograms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data

Definitions

  • the present application relates to the field of medical robot control technology, and in particular to a method, system and computer device for estimating hand joint angles across users.
  • EMG Surface electromyography
  • cross-user intention recognition methods in surface muscle electrical signal control can be divided into two categories: neural network-based model transfer methods and feature-based model transfer methods.
  • the neural network is first trained using sufficient labeled data during training. Then, part of the model parameters are fixed by freezing part of the network weights, and the unfrozen model parameters are updated using labeled or unlabeled target data.
  • the feature-based model structure attempts to obtain features that are invariant in the feature domain through feature domain alignment.
  • the neural network extracts features from the source domain and the target domain to calculate the loss of the feature domain, with the aim of reducing the mismatch of feature distribution in the implicit space.
  • the knowledge transfer method based on the neural network model is widely used because it is relatively simple to implement.
  • the electromyographic signal features extracted by the model between different users are also quite different, and the universality of these features between different users cannot be guaranteed. Therefore, the effect of using the knowledge transfer method based on the neural network model is often unsatisfactory after the model is corrected, and the corrected model is prone to lose the generalization ability for the original user.
  • the parameters of the model are updated by learning the information of the original user and the target user at the same time, which can extract features that are applicable to different users and have little to do with individual differences. Therefore, the corrected model will not lose its robustness and generalization ability for other users, but the correction takes a long time and the user acceptance is low.
  • One of the purposes of this application is to provide a method for estimating hand joint angles across users, comprising the following steps:
  • hand activity signal data of multiple users wherein the hand activity signal data includes electromyographic signals and joint angle signals;
  • the calibrated multi-user model estimates the finger joint angles of the new user.
  • the hand activity signal data including electromyographic signals and joint angle signals
  • the electromyographic signals come from the finger extensors, finger flexors, biceps and triceps, and a circle of muscles on the forearm 2 to 6 cm away from the elbow
  • the electromyographic signal sampling frequency is 1000-2000 Hz
  • the joint angle signals are collected through data gloves, and the sampling frequency is 20-40 Hz.
  • the step of pre-processing the obtained hand activity signal data of multiple users is also included.
  • the step of preprocessing the hand activity signal data of multiple users specifically includes the following steps:
  • a fourth-order Butterworth filter is used to bandpass filter the electromyographic signal, and then the electromyographic signal is baseline corrected and the noise of the electromyographic signal is removed, and the u-law logarithmic scaling method is used to amplify the electromyographic signal; the joint angle signal is resampled to 2000 Hz to synchronize the electromyography and joint angle sequences in time, and the joint angle signal is smoothed using a low-pass filter; the maximum and minimum values of the electromyographic signal and the joint angle signal are recorded and collected for normalization of training and test data.
  • the step of constructing a multi-user model according to the hand activity signal data specifically includes the following steps:
  • the surface electromyographic signals and joint angle training data of multiple users are obtained, and a supervised mean square loss function is used to perform model training to obtain a multi-user neural network model.
  • the multi-user neural network model includes a feature fusion module, two convolution modules and a multi-layer perceptron module, wherein: the feature extraction module is composed of a convolution or self-attention mechanism, and the feature fusion module is composed of a convolution.
  • the step of calibrating the multi-user model using training data of a new user specifically includes the following steps:
  • the calibration of the multi-user model is completed.
  • the feature region discriminator includes a structure of a densely connected residual module, wherein the densely connected residual module is composed of a convolutional layer, a densely connected layer, a conversion layer and a residual connection layer, wherein the previous convolutional layers are used to extract features, the densely connected layer is directly connected to the convolutional layer, the convolutional layer is used to concatenate the extracted features as input to the subsequent convolutional layers, the transition layer uses 1 ⁇ 1 convolution to compress the number of features and compress the network, and the residual connection layer is used to give the network a deeper structure and richer feature representation.
  • the densely connected residual module is composed of a convolutional layer, a densely connected layer, a conversion layer and a residual connection layer, wherein the previous convolutional layers are used to extract features, the densely connected layer is directly connected to the convolutional layer, the convolutional layer is used to concatenate the extracted features as input to the subsequent convolutional layers, the transition layer uses 1 ⁇ 1 convolution to compress the number of features
  • the step of estimating the finger joint angles of a new user by the calibrated multi-user model specifically includes the following steps: inputting a test data set into the calibrated multi-user model, comparing the continuously estimated joint angle curve with the actual joint angle curve obtained by the joint angle sensor, and using three performance indicators, namely, Pearson correlation coefficient, root mean square error and determination coefficient, as evaluation criteria for the regression task.
  • the second object of the present application is to provide a system for estimating hand joint angles across users, comprising:
  • a data acquisition unit used to obtain hand activity signal data of multiple users, wherein the hand activity signal data includes electromyographic signals and joint angle signals;
  • a multi-user model building unit used to build a multi-user model according to the hand activity signal data
  • a calibration unit configured to calibrate the multi-user model using training data of a new user
  • the estimation unit is used to estimate the finger joint angles of the new user based on the calibrated multi-user model.
  • the third purpose of the present application is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for estimating the angles of hand joints across users is implemented.
  • the cross-user hand joint angle estimation method, system and computer device obtained hand activity signal data of multiple users, build a multi-user model based on the hand activity signal data, calibrate the multi-user model through the training data of the new user, and estimate the finger joint angles of the new user by the calibrated multi-user model.
  • the above method, system and computer device use the surface electromyography signals and finger joint angle data of multiple users to establish the multi-user model, and then adopt the newly proposed adversarial transfer learning strategy to calibrate the multi-user model using part of the training data of the new user.
  • the calibrated model can estimate the joint angles of the hand joints of a person during continuous movement in real time. At the same time, due to the application of data-driven algorithms, the complex implementation process of establishing an electromyography to a physiological model is avoided.
  • FIG1 is a flowchart of the steps of a method for estimating hand joint angles across users provided in an embodiment of the present application.
  • FIG2 is a schematic diagram of the model correction and cross-user knowledge transfer strategy provided in this embodiment of the application.
  • FIG3 is a schematic diagram of the structure of the densely connected residual module provided in this embodiment of the application.
  • FIG4 is a schematic diagram of the structure of a user hand joint angle estimation system provided in an embodiment of the present application.
  • FIG. 5 is a schematic diagram of the structure of the computer device provided in this embodiment of the application.
  • first and second are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, a feature defined as “first” or “second” may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of “plurality” is two or more, unless otherwise clearly and specifically defined.
  • Figure 1 is a step flow chart of a method for estimating hand joint angles across users provided in Example 1 of the present application, including the following steps S110 to S140, and the implementation method of each step is described in detail below.
  • Step S110 Acquire hand activity signal data of multiple users, including electromyographic signals and joint angle signals.
  • the hand activity signal data include electromyographic signals and joint angle signals
  • the electromyographic signals come from the finger extensors, finger flexors, biceps and triceps, and a circle of muscles on the forearm 2 to 6 cm away from the elbow
  • the electromyographic signal sampling frequency is 1000-2000Hz
  • the joint angle signals are collected through data gloves, and the sampling frequency is 20-40Hz.
  • the EMG signals were collected using the EMG differential electrodes in the Delsys EMG acquisition system, and the joint angle signals were collected using the CyberGlove II data gloves.
  • the method further includes a step of preprocessing the acquired hand activity signal data of multiple users.
  • the step of preprocessing the hand activity signal data of multiple users specifically includes the following steps:
  • a fourth-order Butterworth filter is used to bandpass filter the electromyographic signal.
  • the electromyographic signal is then baseline corrected and noise is removed. Because some channels of surface electromyographic signals have small amplitudes, the u-law logarithmic scaling method is used to amplify the electromyographic signal; the joint angle signal is resampled to 2000Hz to synchronize the electromyographic and joint angle sequences in time, and the joint angle signal is smoothed using a low-pass filter to avoid step jitter in the signal, making it more like a normal human body movement curve; the maximum and minimum values of the electromyographic signal and the joint angle signal are recorded and collected for normalization of training and test data.
  • Step S120 constructing a multi-user model according to the hand movement signal data.
  • the step of constructing a multi-user model based on the hand activity signal data specifically includes the following steps: obtaining surface electromyography signals and joint angle training data of multiple users, using a supervised mean square loss function to perform model training, and obtaining a multi-user neural network model.
  • the multi-user neural network model includes a feature fusion module, two convolution modules and a multi-layer perceptron module, wherein: the feature extraction module is composed of convolution or self-attention mechanism, and the feature fusion module is composed of convolution.
  • Step S130 calibrating the multi-user model using training data of new users.
  • FIG. 2 shows the process of model training and model correction using the adversarial transfer learning strategy (where the dotted line indicates the optimization direction of the model parameters during the knowledge transfer process), which consists of a multi-user neural network model (Mnet-s), a new user neural network model (Nnet-t), and a feature region discriminator (DFD).
  • the multi-user neural network model is used to extract features from the training data of multiple users
  • the new user neural network model is used to extract features from the training data of new users
  • the feature region discriminator is used to minimize the feature distance between multi-user features and new user features during the model correction process.
  • the step of calibrating the multi-user model using the training data of the new user specifically includes the following steps:
  • Step S131 construct a new user neural network model with the same structure as the multi-user neural network model, and initialize it with the weights of the trained multi-user neural network model so that the initial parameters are the same as the parameters of the trained multi-user neural network model.
  • the multi-user neural network model was trained with surface EMG signals and joint angle data from multiple subjects using a supervised mean square loss function for model training.
  • Step S132 Setting a feature region discriminator, the parameters of the multi-user neural network model are frozen, and the parameters of the new user neural network model and the feature region discriminator are corrected and optimized to minimize the feature distance between the feature domains of the multi-users and the feature domain of the new user.
  • the cross-user model transfer learning process is trained through the following formulas (1)-(5), where L total is the total loss function used to optimize the new user neural network model and feature region discriminator.
  • LD represents the loss function of the feature region discriminator, which is used to represent the relative distance between the feature domains of the multi-user neural network model and the new user neural network model.
  • Fs represents the characteristics of the multi-user neural network model
  • Ft represents the characteristics of the new user neural network model
  • LN represents the loss function of the new user neural network model
  • Lmap represents the degree of mapping from the features of the multi-user neural network model to the features of the new user neural network model.
  • L subject is used to improve the performance of the loss function and make the new user neural network model more sensitive to the characteristics of new users during the transfer learning process.
  • Nnet(x) represents the finger joint angle estimated by the model
  • the label represents the finger joint angle
  • N represents the number of finger joints
  • ⁇ is a weighting coefficient
  • Step S133 When the feature region discriminator has difficulty distinguishing two feature domains and finding the minimum distance between the two feature domains, the calibration of the multi-user model is completed.
  • the densely connected residual module consists of a convolutional layer, a densely connected layer, a transition layer, and a residual connection.
  • the convolutional layer is used to extract features
  • the densely connected layer is a direct connection between convolutional layers.
  • the features extracted by the previous convolutional layers are concatenated as the input of the subsequent convolutional layers. This can improve the information propagation within the neural network and thus speed up the convergence of the model.
  • the transition layer uses 1 ⁇ 1 convolution to compress the number of features and compress the network.
  • the final residual connection structure gives the network a deeper structure and a richer feature representation.
  • the migration method proposed in this embodiment corrects the parameters by learning the information of the original user and the target user, and can extract features that are applicable to different users and have little to do with individual differences. Therefore, the corrected model can reuse more features of the original user, obtain higher estimation accuracy, and have better robustness and generalization ability; the knowledge migration algorithm proposed in this embodiment has a short correction time, which is conducive to actual deployment.
  • the algorithm proposed in the present invention achieves model correction by minimizing the difference in feature distribution between different users, and does not require model retraining, so the correction time is short and user acceptance is high.
  • Step S140 the calibrated multi-user model estimates the finger joint angles of the new user.
  • the step of estimating the finger joint angles of a new user by the calibrated multi-user model specifically includes the following steps: inputting a test data set into the calibrated multi-user model, comparing the continuously estimated joint angle curve with the actual joint angle curve obtained by the joint angle sensor, and using three performance indicators, namely, Pearson correlation coefficient, root mean square error and determination coefficient, as evaluation criteria for the regression task.
  • the Pearson correlation coefficient (CC) is used to measure the linear correlation between the estimated value of the finger joint angle and the corresponding actual data.
  • the calculation formula is as follows.
  • RMSE Root mean square error
  • R2 Coefficient of determination
  • N represents the sample size
  • pi represents the sample points of the predicted finger joint angles
  • gi represents the sample points of the actual finger joint angles.
  • the cross-user hand joint angle estimation method provided in the present application utilizes surface electromyography signals and finger joint angle data of multiple users to establish a multi-user model, and then adopts the newly proposed adversarial transfer learning strategy to calibrate the multi-user model using partial training data of new users.
  • the calibrated model can estimate the joint angles of a person's hand joints during continuous motion in real time.
  • the complex implementation process of establishing an electromyography to a physiological model is avoided.
  • the cross-user hand joint angle estimation method provided by this application can be used for industrial and aerospace robot control and operation, and can manipulate the manipulator to grasp objects of different sizes based on human surface electromyography information.
  • the hand joint angle continuously estimated by surface electromyography signals can make the entire control process natural and dexterous, and can achieve efficient interaction between people and manipulators.
  • the cross-user model transfer technology can improve the generalization performance of the hand state estimation algorithm for new users that have never been seen. At the same time, since there is no need to retrain the model, it can improve the rapid deployment capability of the model in the real world.
  • the cross-user hand joint angle estimation method provided in this application can provide more intelligent and humanized rehabilitation training equipment for patients with motor dysfunction, introduce electromyographic pattern recognition technology into the control link of the rehabilitation training system, and train the model through the patient's residual surface electromyographic signals and the virtual hand joint angle information when the patient imagines completing the action, so as to help patients with motor dysfunction to complete rehabilitation training more actively.
  • the model pre-trained with the initial data is promoted to new users, and the generalization is improved while the time for model recalibration is greatly reduced. This technology greatly reduces the burden on users.
  • Figure 4 provides a cross-user hand joint angle estimation system for this embodiment 2, including: a data acquisition unit 110, used to obtain hand activity signal data of multiple users, the hand activity signal data including electromyographic signals and joint angle signals; a multi-user model construction unit 120, used to construct a multi-user model based on the hand activity signal data; a calibration unit 130, used to calibrate the multi-user model through the training data of the new user; an estimation unit 140, used to estimate the finger joint angles of the new user using the calibrated multi-user model.
  • a data acquisition unit 110 used to obtain hand activity signal data of multiple users, the hand activity signal data including electromyographic signals and joint angle signals
  • a multi-user model construction unit 120 used to construct a multi-user model based on the hand activity signal data
  • a calibration unit 130 used to calibrate the multi-user model through the training data of the new user
  • an estimation unit 140 used to estimate the finger joint angles of the new user using the calibrated multi-user model.
  • the detailed working mode of the estimation system provided in this embodiment can be found in the method for estimating the hand joint angles across users provided in the embodiment, which will not be described in detail here.
  • the cross-user hand joint angle estimation system utilizes surface electromyographic signals and finger joint angle data of multiple users to establish a multi-user model, and then adopts the newly proposed adversarial transfer learning strategy to calibrate the multi-user model using partial training data of new users.
  • the calibrated model can estimate the joint angles of a person's hand joints during continuous motion in real time.
  • the complex implementation process of establishing an electromyographic to physiological model is avoided.
  • FIG5 is a schematic diagram of the structure of a computer device according to an embodiment of the present application.
  • the computer device 50 includes a processor 51 and a memory 52 coupled to the processor 51 .
  • the memory 52 stores program instructions for implementing the above-mentioned method for estimating the hand joint angles across users.
  • the processor 51 is used to execute the program instructions stored in the memory 52 to implement the method for estimating the angles of hand joints across users.
  • the processor 51 may also be referred to as a CPU (Central Processing Unit).
  • the processor 51 may be an integrated circuit chip having signal processing capabilities.
  • the processor 51 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • DSP digital signal processor
  • ASIC application-specific integrated circuit
  • FPGA field-programmable gate array
  • the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.
  • the experiment uses a sliding window to generate a surface electromyography signal sequence and a joint angle sequence with a window length of 2000 sampling points.
  • the sliding window step size is 100 sampling points.
  • the joint angle and surface electromyography signal data in each sliding window will be used as a sample data, and the joint angle vector dimension represents the estimated number of joint angles.
  • a total of 35 users' data are collected, of which 28 users' data are used to train the multi-user model, and the remaining 7 users are used as new users for correction and testing algorithm performance.
  • For the data of new users 80% of the data samples are used as correction models, and 20% of the data samples are used as test data to test algorithm performance.
  • the present invention uses the joint and electromyographic data of 35 subjects when completing the grasping action of objects of different sizes to train, correct and test the network model.
  • the data of 28 subjects are used to establish a multi-user model, and the remaining 7 subjects are used as new users for model correction and testing. 80% of the data of each new user is used for model migration correction, and the other 20% is used for model performance testing.
  • the three regression model performance indicators of CC, RMSE and R2 of the proximal interphalangeal joints and metacarpophalangeal joints of the hand are tested, as shown in Table 1.
  • the present invention uses the network Long Short-Term Memory Network Sparse Pseudo-Input Gaussian Process CC 0.859 ⁇ 0.012 0.830 ⁇ 0.011 0.775 ⁇ 0.009 RMSE 7.344 ⁇ 0.242 7.973 ⁇ 0.219 8.953 ⁇ 0.267 R2 0.761 ⁇ 0.035 0.711 ⁇ 0.040 0.643 ⁇ 0.047

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Physics & Mathematics (AREA)
  • Surgery (AREA)
  • Animal Behavior & Ethology (AREA)
  • Biomedical Technology (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Medical Informatics (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Veterinary Medicine (AREA)
  • Human Computer Interaction (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)

Abstract

La présente demande concerne un procédé et un système d'estimation d'angle d'articulation de main inter-utilisateur, et un dispositif informatique. Le procédé consiste à : acquérir des données de signal d'activité de main d'une pluralité d'utilisateurs ; construire un modèle multiutilisateur en fonction des données de signal d'activité de main ; étalonner le modèle multiutilisateur au moyen de données d'apprentissage d'un nouvel utilisateur ; et estimer, au moyen du modèle multiutilisateur étalonné, un angle d'articulation de doigt du nouvel utilisateur. Selon le procédé, le système et le dispositif informatique, un modèle multiutilisateur est établi au moyen de signaux d'électromyographie de surface et de données d'angle d'articulation de doigt d'une pluralité d'utilisateurs, puis, au moyen d'une politique d'apprentissage de transfert antagoniste nouvellement proposée, le modèle multiutilisateur est étalonné à l'aide de certaines données d'apprentissage d'un nouvel utilisateur, et le modèle étalonné permet d'estimer l'angle d'articulation d'une articulation de main d'une personne en temps réel dans un processus de mouvement continu ; un algorithme commandé par des données est en outre utilisé, ce qui permet d'éviter le processus de mise en œuvre complexe consistant à établir un modèle électromyographique à physiologique.
PCT/CN2022/135520 2022-11-30 2022-11-30 Procédé et système d'estimation d'angle d'articulation de main inter-utilisateur, et dispositif informatique WO2024113241A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/135520 WO2024113241A1 (fr) 2022-11-30 2022-11-30 Procédé et système d'estimation d'angle d'articulation de main inter-utilisateur, et dispositif informatique

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/135520 WO2024113241A1 (fr) 2022-11-30 2022-11-30 Procédé et système d'estimation d'angle d'articulation de main inter-utilisateur, et dispositif informatique

Publications (1)

Publication Number Publication Date
WO2024113241A1 true WO2024113241A1 (fr) 2024-06-06

Family

ID=91322699

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/135520 WO2024113241A1 (fr) 2022-11-30 2022-11-30 Procédé et système d'estimation d'angle d'articulation de main inter-utilisateur, et dispositif informatique

Country Status (1)

Country Link
WO (1) WO2024113241A1 (fr)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108874149A (zh) * 2018-07-28 2018-11-23 华中科技大学 一种基于表面肌电信号连续估计人体关节角度的方法
US20200034978A1 (en) * 2018-01-25 2020-01-30 Ctrl-Labs Corporation Real-time processing of handstate representation model estimates
CN114548165A (zh) * 2022-02-18 2022-05-27 中国科学技术大学 一种可跨用户的肌电模式分类方法
CN114707539A (zh) * 2022-02-23 2022-07-05 中国科学院深圳先进技术研究院 手部关节角度估计方法、估计装置、存储介质和设备
CN115137351A (zh) * 2022-07-22 2022-10-04 安徽大学 一种基于肌电信号的上肢肘关节角度估计方法及系统

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200034978A1 (en) * 2018-01-25 2020-01-30 Ctrl-Labs Corporation Real-time processing of handstate representation model estimates
CN108874149A (zh) * 2018-07-28 2018-11-23 华中科技大学 一种基于表面肌电信号连续估计人体关节角度的方法
CN114548165A (zh) * 2022-02-18 2022-05-27 中国科学技术大学 一种可跨用户的肌电模式分类方法
CN114707539A (zh) * 2022-02-23 2022-07-05 中国科学院深圳先进技术研究院 手部关节角度估计方法、估计装置、存储介质和设备
CN115137351A (zh) * 2022-07-22 2022-10-04 安徽大学 一种基于肌电信号的上肢肘关节角度估计方法及系统

Similar Documents

Publication Publication Date Title
CN112005198A (zh) 基于多个输入的手部状态重建
CN110946556A (zh) 基于可穿戴式体感网的帕金森静息态震颤评估方法
Bao et al. Toward robust, adaptiveand reliable upper-limb motion estimation using machine learning and deep learning–A survey in myoelectric control
Ma et al. A bi-directional LSTM network for estimating continuous upper limb movement from surface electromyography
CN113749670A (zh) 使用相干算法的局部噪声识别
WO2023159674A1 (fr) Appareil d'estimation et procédé d'estimation d'angle d'articulation de main, et support de stockage
US11166667B2 (en) System for identifying information represented by biological signals
Zhou et al. Adapting random forest classifier based on single and multiple features for surface electromyography signal recognition
CN113901891A (zh) 帕金森病握拳任务的评估方法及系统、存储介质及终端
Bao et al. CNN confidence estimation for rejection-based hand gesture classification in myoelectric control
Dousty et al. Towards clustering hand grasps of individuals with spinal cord injury in egocentric video
WO2024113241A1 (fr) Procédé et système d'estimation d'angle d'articulation de main inter-utilisateur, et dispositif informatique
Briouza et al. EMG signal classification for human hand rehabilitation via two machine learning techniques: kNN and SVM
Chen et al. SEMG-based gesture recognition using GRU with strong robustness against forearm posture
CN115966015A (zh) 跨用户手部关节角度估计方法、系统及计算机设备
Unanyan et al. A real-time fail-safe algorithm for decoding of myoelectric signals to control a prosthetic arm
Putri et al. Retracted: Implementation of Neural Network Classification for Diabetes Mellitus Prediction System through Iridology Image
WO2024103371A1 (fr) Procédé d'estimation d'angle d'articulation de la main
CN115363594A (zh) 一种基于循环神经网络的实时心脏疾病筛查方法
Biswas et al. CORDIC framework for quaternion-based joint angle computation to classify arm movements
Ali et al. Robust arrhythmia classifier using hybrid multilayered perceptron network
CN113729728A (zh) 心脏标测中心脏结构的自动检测
CN113907866A (zh) 用于确定导管的位置的系统和方法
Chen et al. A SEMG-angle model based on HMM for human robot interaction
CN111035390A (zh) 一种基于人手腕背伸动作的手功能康复定量评估方法