CN115568856A - Multi-finger force continuous estimation method based on collaborative mapping reconstruction - Google Patents
Multi-finger force continuous estimation method based on collaborative mapping reconstruction Download PDFInfo
- Publication number
- CN115568856A CN115568856A CN202211264621.1A CN202211264621A CN115568856A CN 115568856 A CN115568856 A CN 115568856A CN 202211264621 A CN202211264621 A CN 202211264621A CN 115568856 A CN115568856 A CN 115568856A
- Authority
- CN
- China
- Prior art keywords
- force
- muscle
- global
- task
- finger
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000013507 mapping Methods 0.000 title claims abstract description 14
- 210000003205 muscle Anatomy 0.000 claims abstract description 122
- 230000004913 activation Effects 0.000 claims abstract description 33
- 230000009471 action Effects 0.000 claims abstract description 24
- 238000012417 linear regression Methods 0.000 claims abstract description 17
- 238000012549 training Methods 0.000 claims abstract description 13
- 238000007781 pre-processing Methods 0.000 claims abstract description 9
- 239000011159 matrix material Substances 0.000 claims description 66
- 230000000875 corresponding effect Effects 0.000 claims description 35
- 230000006690 co-activation Effects 0.000 claims description 24
- 239000013598 vector Substances 0.000 claims description 24
- 238000004422 calculation algorithm Methods 0.000 claims description 20
- 230000033001 locomotion Effects 0.000 claims description 20
- 238000012360 testing method Methods 0.000 claims description 15
- 238000004458 analytical method Methods 0.000 claims description 14
- 230000008602 contraction Effects 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 13
- 230000002195 synergetic effect Effects 0.000 claims description 13
- 238000000354 decomposition reaction Methods 0.000 claims description 12
- 238000012937 correction Methods 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 6
- 230000004118 muscle contraction Effects 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 4
- 239000000284 extract Substances 0.000 claims description 4
- 101800004637 Communis Proteins 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000001360 synchronised effect Effects 0.000 abstract description 5
- 210000001364 upper extremity Anatomy 0.000 abstract description 2
- 230000002730 additional effect Effects 0.000 abstract 1
- 210000003811 finger Anatomy 0.000 description 47
- 230000008859 change Effects 0.000 description 2
- 238000002567 electromyography Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 210000003169 central nervous system Anatomy 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003064 k means clustering Methods 0.000 description 1
- 210000004932 little finger Anatomy 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003387 muscular Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000012567 pattern recognition method Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/22—Ergometry; Measuring muscular strength or the force of a muscular blow
- A61B5/224—Measuring muscular strength
- A61B5/225—Measuring muscular strength of the fingers, e.g. by monitoring hand-grip force
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/103—Detecting, measuring or recording devices for testing the shape, pattern, colour, size or movement of the body or parts thereof, for diagnostic purposes
- A61B5/11—Measuring movement of the entire body or parts thereof, e.g. head or hand tremor, mobility of a limb
- A61B5/1107—Measuring contraction of parts of the body, e.g. organ, muscle
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/22—Ergometry; Measuring muscular strength or the force of a muscular blow
- A61B5/224—Measuring muscular strength
- A61B5/227—Measuring muscular strength of constricting muscles, i.e. sphincters
-
- 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]
-
- 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
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7271—Specific aspects of physiological measurement analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/011—Arrangements for interaction with the human body, e.g. for user immersion in virtual reality
- G06F3/015—Input arrangements based on nervous system activity detection, e.g. brain waves [EEG] detection, electromyograms [EMG] detection, electrodermal response detection
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Veterinary Medicine (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Public Health (AREA)
- Physiology (AREA)
- Physical Education & Sports Medicine (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Neurology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychiatry (AREA)
- Signal Processing (AREA)
- Dermatology (AREA)
- Artificial Intelligence (AREA)
- Neurosurgery (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- General Physics & Mathematics (AREA)
- Dentistry (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
Abstract
A multi-finger force continuous estimation method based on collaborative mapping reconstruction includes the steps that firstly, multi-channel upper limb surface electromyographic signals and finger tip force signals when a hand stably grabs an object through six actions are synchronously collected, and envelopes of the two signals are obtained after preprocessing; extracting global muscle synergy and corresponding global muscle synergy activation coefficients from the surface electromyogram signal envelope; then extracting global force coordination and a corresponding global force coordination activation coefficient from the finger force signal envelope; then taking the global muscle cooperative activation coefficient as input and the global force cooperative activation coefficient as output, training a multiple linear regression model, and carrying out synchronous and continuous estimation on multi-finger force through cooperative mapping and reconstruction; finally, the influence of the finger error activation on the multi-finger force estimation result is reduced through an additional action identification link; the invention can realize the synchronous and continuous estimation of the force of multiple fingers under six different actions and has higher accuracy.
Description
Technical Field
The invention relates to a finger force estimation method, in particular to a multi-finger force continuous estimation method based on collaborative mapping reconstruction.
Background
The surface electromyogram (sEMG) is a bioelectric signal generated when muscles contract, and is widely applied to the field of human-computer interaction due to the characteristics of no wound, easy acquisition, large information capacity, strong real-time performance and the like; the finger motion information decoded from the electromyographic signals on the surface of the upper limb can be used for controlling rehabilitation equipment such as an intelligent prosthetic hand or hand exoskeleton and the like.
The existing electromyographic control interface technology mostly adopts a pattern recognition method, for example, chinese invention patent CN202210837431.8 proposes an electromyographic biofeedback device and system based on pattern recognition, wherein game devices are controlled by extracting the characteristics of the electromyographic signal active segment of a user and recognizing the pattern into a plurality of actions, but the output of the game devices is a predefined discrete control instruction, so that the smooth control of a human-like person cannot be realized; the Chinese invention patent CN201711019198.8 provides a manipulator motion speed proportion control method based on electromyography, which improves the naturalness and the initiative of man-machine interaction by fitting the average power of the electromyography signals into a continuous manipulator speed control instruction, but can only realize the continuous decoding of a single motion freedom degree each time and cannot be directly used for the synchronous and continuous estimation of multi-finger force.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a multi-finger force continuous estimation method based on collaborative mapping reconstruction, which can realize the synchronous and continuous estimation of multi-finger force under six actions by extracting the global muscle collaboration and the global force collaboration when a human hand stably grabs an object (isometric contraction) through six actions and carrying out mapping reconstruction, and has higher accuracy.
In order to achieve the purpose, the invention adopts the technical scheme that:
a multi-finger force continuous estimation method based on collaborative mapping reconstruction comprises the following steps:
(1) Data acquisition and preprocessing: when an object to be tested stably grabs an object under three different muscle contraction forces by six actions, namely isometric contraction, multi-channel electromyographic signals of the surfaces of the internal muscle and the external muscle of a hand and fingertip force signals of five fingers are synchronously acquired, and multi-channel envelope signals X of the two signals are respectively obtained through preprocessing M (t) and X F (t)；
(2) Muscle synergy analysis: multichannel envelope signal X for sEMG M (t) first extracting task-specific muscle synergies W from the non-negative matrix factorization (NNMF) algorithm, respectively M (i) (i =1,2, …, 6) and task-shared muscle synergy W M Then adopting a K-means clustering algorithm to cooperate W from task-specific muscles M (i) Extract global muscle synergy from MC Extracting corresponding global muscle synergistic activation coefficient H by adopting non-negative least square method (NNLS) MC (t)；
(3) And (3) force synergy analysis: multichannel envelope signal X for fingertip force F (t) first, respectively extracting task specific forces from the data by using a nonnegative matrix factorization (NNMF) algorithm F (i) (i =1,2, …, 6) and tasking force synergy W F Then, cooperating W from task specific power by adopting a K mean value clustering algorithm F (i) Medium extraction global force synergy W FC Extracting corresponding global force cooperative activation coefficient H by adopting non-negative least square method (NNLS) FC (t)；
(4) Model training and testing: training a multiple linear regression model to coordinate the activation coefficient H of the global muscles MC (t) mapping to a Global force co-activation coefficient H FC (t) global force co-activation coefficient H obtained by regression FC (t) synergize with the global force W FC Reconstructing a real finger force signal, and then using the test set data for the test of the multiple linear regression model;
the multiple linear regression model willGlobal muscle co-activation coefficient H MC (t) as model input, global force co-activation coefficient H FC (t) as model output, assuming a linear relationship exists between the two:
H FC (t)＝α×H MC (t)
wherein alpha is a regression coefficient matrix;
the model test flow is as follows: global muscle synergy W is known MC Collaborate with global force W FC For test electromyographic signal envelope X M (t) extracting the corresponding global muscle co-activation coefficient H by using a non-negative least squares method (NNLS) MC (t), inputting the result into a trained multiple linear regression model to obtain a global force cooperative activation coefficient H FC (t) then coordinating W with a global force FC Multiplying by a global force co-activation coefficient H FC (t) to reconstruct the true finger force signal X F (t) is the estimated force of the plurality of fingers
(5) Correcting a finger force estimation result: the method comprises the steps of estimating the force of multiple fingers by using a multiple linear regression model, synchronously identifying the real activation state of each finger in a multi-finger grabbing task by adopting a task-specific muscle cooperation-based action identification method, and generating a corresponding correction vector psi according to the identification result i (i =1,2, …, 6) result of finger force estimation
And (5) correcting:
wherein
Is the final prediction result of the finger force.
In the step (1), the six actions include: thumb-index finger (TI), thumb-middle finger (TM), thumb-ring finger (TR), thumb-little finger (TL), thumb-index finger-middle finger (TIM), and all fingers (TIMRL);
three different muscle contraction forces include: 15% maximum autonomous contraction (MVC), 30% maximum autonomous contraction (MVC), and 45% maximum autonomous contraction (MVC);
the intrinsic muscles of the hand include: THE first dorsal interosseous muscle (FDI), THE dorsal interosseous muscle (DI), THE major thenar muscle (he) and THE minor thenar muscle (ADM);
the external muscles of the hand include: extensor Digitorum Communis (EDC) and Flexor Digitorum Superficialis (FDS);
the pre-processing includes 20-500Hz band-pass filtering, full-wave rectification, cut-off frequency 3Hz low-pass filtering and normalization.
In the step (2), the task-specific muscle is cooperated with W M (i) By enveloping X with electromyographic signals corresponding to individual movements M (i) (t) (i =1,2, …, 6) as the NNMF input, and matrix X is divided into M (i) (t) new matrix W obtained after decomposition M (i) Matrix W as a task specific muscle synergy M (i) The number of columns is the number of the muscle synergies specific to the task;
task consensus muscle synergy W M By enveloping X with electromyographic signals corresponding to respective movements M (i) (t) matrix X obtained after tandem M (t) as input to NNMF, and the matrix X M (t) new matrix W obtained after decomposition M As task-shared muscle coordination, matrix W M The number of lines is the number of the common muscle synergies of the tasks;
global muscle synergy W MC The extraction process comprises the following steps: synergy of task-shared muscles W M The column vectors are respectively used as the initial value of the mass center of each cluster, and a K mean clustering algorithm is adopted to cooperate with the special muscles of the task by W M (i) All column vectors of (A) are subjected to clustering analysis, and the value of K and the task share muscle cooperation W M The column number of the new matrix W is the same, and the new matrix W is obtained by recombining the centroid vectors of all the clusters after clustering MC As global muscle synergy, matrix W MC The number of lines is the number of global muscle synergies;
globalCoefficient of muscular co-activation H MC (t) extraction process: known training set electromyographic signal envelope X M (t) global muscle synergy W MC Extracting the global muscle synergistic activation coefficient H by adopting a non-negative least square method (NNLS) MC (t)。
In the step (3), the task-specific force is cooperated with W F (i) By enveloping X with the force signal corresponding to the individual movements F (i) (t) (i =1,2, …, 6) as the NNMF input, and matrix X is divided into F (i) (t) new matrix W obtained after decomposition F (i) As a task-specific force synergy, matrix W F (i) The number of columns is the number of the task specific force cooperation;
task-shared force synergy W F By enveloping X the force signal corresponding to each motion F (i) (t) matrix X obtained after tandem connection F (t) as input to NNMF, and X F (t) new matrix W obtained after decomposition F As a task-shared force synergy, matrix W F The number of the rows is the number of the task sharing force coordination;
global force synergy W FC The extraction process comprises the following steps: coordinating task consensus forces with W F The column vectors are respectively used as the initial value of the mass center of each cluster, and the K mean value clustering algorithm is adopted to cooperate with the specific force of the task to W F (i) All column vectors are subjected to clustering analysis, and the value of K and the task sharing force cooperate with W F The column number of the new matrix W is the same, and the new matrix W is obtained by recombining the centroid vectors of all the clusters after clustering FC As a global force synergy, matrix W FC The number of columns is the number of global force coordination;
global force co-activation coefficient H FC (t) extraction process: known training set force signal envelope X F (t) synergize with the global force W FC Extracting global force synergistic activation coefficient H by adopting non-negative least square method (NNLS) FC (t)。
In the step (5), the finger force estimation result correction vector Ψ corresponding to the six actions i Respectively as follows:
Ψ 1 ＝[1,1,0,0,0] T ；
Ψ 2 ＝[1,0,1,0,0] T ；
Ψ 3 ＝[1,0,0,1,0] T ；
Ψ 4 ＝[1,0,0,0,1] T ；
Ψ 5 ＝[1,1,1,0,0] T ；
Ψ 6 ＝[1,1,1,1,1] T ；
the action identification method based on task specific muscle cooperation comprises the following steps: sequential use of task specific muscle synergies W M (i) Electromyographic signal X to be identified M (t) reconstructing and calculating the reconstructed electromyographic signals respectively
And electromyographic signal X to be recognized M (t) similarity between S i (i =1,2, …, 6), W with the highest similarity M (i) Corresponding action category is taken as the electromyographic signal X to be recognized M (t) the corresponding action category.
The similarity S i Various evaluation indexes such as included angle cosine, euclidean distance, pearson correlation coefficient, decision coefficient and the like can be adopted.
The invention has the beneficial effects that:
1. the method uses the multiple linear regression model for collaborative mapping and force reconstruction, has simple principle, is easy to model and train, and can realize the synchronous and continuous estimation of the finger tip force of five fingers under six degrees of freedom.
2. According to the method, the global cooperative component with higher motion decoupling degree is extracted from the multichannel surface electromyographic signals and is used for fingertip force signal reconstruction, and the influence of signal crosstalk between adjacent channels on the multi-finger force estimation result is reduced.
3. The method combines finger force estimation and finger action identification, and reduces the influence of the false activation of individual fingers on the multi-finger force estimation result in the multi-finger grabbing task.
Drawings
FIG. 1 is an overall flow chart of the present invention.
FIG. 2 is a flow chart of the muscle synergy analysis of the present invention.
FIG. 3 is a flow chart of the force synergy analysis of the present invention.
FIG. 4 is a flowchart of a task-specific muscle coordination-based motion recognition method according to the present invention.
FIG. 5 is a graph of experimental results of multi-finger force estimation under six actions in accordance with the present invention.
Detailed Description
The invention is described in detail below with reference to the figures and examples.
Referring to fig. 1, a multi-finger force continuous estimation method based on collaborative mapping reconstruction includes the following steps:
(1) Data acquisition and preprocessing: when an object (isometric contraction) is tested to be stably grabbed under three different muscle contraction forces by six actions, multi-channel electromyographic signals of the surfaces of the internal muscle and the external muscle of a hand and fingertip force signals of five fingers are synchronously acquired, and multi-channel envelope signals X of the two signals are respectively obtained through preprocessing M (t) and X F (t)；
The six actions include: thumb-index finger (TI), thumb-middle finger (TM), thumb-ring finger (TR), thumb-little finger (TL), thumb-index finger-middle finger (TIM), and all fingers (TIMRL);
three different muscle contraction forces include: 15% maximum autonomous contraction (MVC), 30% maximum autonomous contraction (MVC), and 45% maximum autonomous contraction (MVC);
the intrinsic muscles of the hand include: THE first dorsal interosseous muscle (FDI), THE dorsal interosseous muscle (DI), THE major thenar muscle (he) and THE minor thenar muscle (ADM);
the external muscles of the hand include: extensor Digitorum Communis (EDC) and Flexor Digitorum Superficialis (FDS);
the pretreatment comprises 20-500Hz band-pass filtering, full-wave rectification, cutoff frequency 3Hz low-pass filtering and normalization;
(2) Muscle synergy analysis: according to the muscle synergy theory, the surface electromyographic signals can be approximately expressed as linear combinations of muscle synergy and corresponding activation coefficients; the muscle cooperation reflects the participation degree of each muscle in the action execution process, and the activation coefficient can be regarded as a time-varying control instruction sent from the central nervous system;
referring to fig. 2, a multi-channel envelope signal X for sEMG M (t) first extracting task specific muscle synergies W from each of them using a non-negative matrix factorization (NNMF) algorithm M (i) (i =1,2, …, 6) and task-shared muscle synergy W M Then adopting a K mean clustering algorithm to cooperate W from task-specific muscles M (i) Extract global muscle synergy from MC Extracting corresponding global muscle synergistic activation coefficient H by adopting non-negative least square method (NNLS) MC (t)；
Task specific muscle synergy W M (i) By enveloping X with electromyographic signals corresponding to individual movements M (i) (t) (i =1,2, …, 6) as the NNMF input, and matrix X is divided into M (i) (t) new matrix W obtained after decomposition M (i) Matrix W as a task specific muscle synergy M (i) The number of lines is the number of muscle synergies specific to the task;
task consensus muscle synergy W M By enveloping X with electromyographic signals corresponding to respective movements M (i) (t) matrix X obtained after tandem M (t) as input to NNMF, and matrix X M (t) new matrix W obtained after decomposition M As task-shared muscle coordination, matrix W M The number of lines is the number of the common muscle synergies of the tasks;
global muscle synergy W MC The extraction process comprises the following steps: synergy of task-shared muscles W M The column vectors are respectively used as the initial value of the mass center of each cluster, and a K mean clustering algorithm is adopted to cooperate with the special muscles of the task by W M (i) All column vectors of (A) are subjected to clustering analysis, and the value of K and the task share muscle cooperation W M The column number of the new matrix W is the same, and the new matrix W is obtained by recombining the centroid vectors of all the clusters after clustering MC As global muscle synergy, matrix W MC The number of lines is the number of global muscle synergies;
global muscle co-activation coefficient H MC (t) extraction process: since muscle coordination represents the degree of involvement of each muscle during action triggering, it can be considered thatThe structure is relatively stable and does not change along with time; to improve the computational efficiency of the algorithm, the training set electromyographic signal envelope X is known M (t) global muscle synergy W MC The global muscle synergistic activation coefficient H can be extracted by adopting a non-negative least square method (NNLS) MC (t)；
(3) Force synergy analysis: referring to FIG. 3, a multi-channel envelope signal X for fingertip force F (t) first, respectively extracting task specific forces from the data by using a nonnegative matrix factorization (NNMF) algorithm F (i) (i =1,2, …, 6) and task-shared force synergy W F Then, cooperating W from task specific power by adopting a K mean value clustering algorithm F (i) Middle extracted global force synergy W FC Extracting corresponding global force cooperative activation coefficient H by adopting non-negative least square method (NNLS) FC (t)；
Task specific force collaboration W F (i) By enveloping X with the force signal corresponding to the individual movements F (i) (t) (i =1,2, …, 6) as the NNMF input, and matrix X is divided into F (i) (t) new matrix W obtained after decomposition F (i) As a task-specific force synergy, matrix W F (i) The number of columns is the number of the task specific force cooperation;
task-shared force synergy W F By enveloping X the force signal corresponding to each motion F (i) (t) matrix X obtained after tandem F (t) as input to NNMF, and X F (t) new matrix W obtained after decomposition F As a task-shared force synergy, matrix W F The number of the rows is the number of the task sharing force coordination;
global force synergy W FC The extraction process comprises the following steps: coordinating the task sharing force obtained in the previous step with W F The column vectors of the data are respectively used as the initial value of the mass center of each cluster, and a K mean value clustering algorithm is adopted to cooperate with the specific force of the task to W F (i) All column vectors are subjected to clustering analysis, and the value of K and the task sharing force cooperate with W F The column number of the new matrix W is the same, and the new matrix W is obtained by recombining the centroid vectors of all the clusters after clustering FC As a global force synergy, matrix W FC The number of columns is the number of global force coordination;
global force co-activation coefficient H FC (t) extraction process: the force cooperation represents the contribution degree of each finger force when the object is stably grabbed, so that the structure of the robot is relatively stable and does not change along with time; to improve the computational efficiency of the algorithm, the training set force signal envelope X is known F (t) synergize with the global force W FC The non-negative least square method (NNLS) can be adopted to extract the global force synergistic activation coefficient H FC (t)；
(4) Model training and testing: training a multiple linear regression model to coordinate the activation coefficient H of the global muscles MC (t) mapping to a Global force co-activation coefficient H FC (t) global force co-activation coefficient H obtained by regression FC (t) and the global force obtained in step (3) cooperate with W FC Reconstructing a real finger force signal, and then using the test set data for the test of the multiple linear regression model;
the multiple linear regression model combines the global muscle co-activation coefficient H MC (t) as model input, global force co-activation coefficient H FC (t) as model output, assuming a linear relationship exists between the two:
H FC (t)＝α×H MC (t)
wherein alpha is a regression coefficient matrix;
the model test flow is as follows: knowing the global muscle synergy W extracted in step (2) MC Cooperating with the global force extracted in the step (3) to obtain W FC For test set electromyographic signal envelope X M (t) extracting the corresponding global muscle co-activation coefficient H by using a non-negative least squares method (NNLS) MC (t), inputting the global force synergistic activation coefficient into a trained multiple linear regression model to obtain a global force synergistic activation coefficient H FC (t) then coordinating W with a global force FC Multiplying by a global force co-activation coefficient H FC (t) to reconstruct the true finger force signal X F (t) is the estimated force of the plurality of fingers
(5) And (3) correcting a finger force estimation result: coupled motions of individual fingers (such as ring fingers and little fingers) in a multi-finger grabbing task easily cause misinterpretation of finger activation states, and in order to reduce the influence of finger false activation on a finger force estimation result, a finger force estimation result correction link is introduced after a regression prediction model to correct the finger force which is falsely activated;
the finger force estimation result correction comprises the following steps: when the multi-finger force is estimated by using a multiple linear regression model, a motion identification method based on task specific muscle cooperation is adopted to synchronously identify the real activation state of each finger in a multi-finger grabbing task, and a corresponding correction vector psi is generated according to the identification result i (i =1,2, …, 6) result of finger force estimation
And (5) correcting:
wherein
The final prediction result of the finger force is obtained;
correction vector psi of finger force estimation result corresponding to six actions i Respectively as follows:
Ψ 1 ＝[1,1,0,0,0] T ；
Ψ 2 ＝[1,0,1,0,0] T ；
Ψ 3 ＝[1,0,0,1,0] T ；
Ψ 4 ＝[1,0,0,0,1] T ；
Ψ 5 ＝[1,1,1,0,0] T ；
Ψ 6 ＝[1,1,1,1,1] T ；
referring to fig. 4, a task-specific muscle coordination based action recognition method: sequentially using the task-specific muscle synergies W obtained in the step (2) M (i) Electromyographic signals to be identifiedX M (t) reconstructing and calculating the reconstructed electromyographic signals respectively
And electromyographic signal X to be identified M (t) similarity between S i (i =1,2, …, 6), W with the highest similarity M (i) Corresponding action category is taken as the electromyographic signal X to be recognized M (t) the corresponding action category;
similarity S i Various evaluation indexes such as included angle cosine, euclidean distance, pearson correlation coefficient, decision coefficient and the like can be adopted, and the decision coefficient is adopted as the similarity evaluation index in the embodiment.
In this embodiment, the length of the signal corresponding to each motion is set to be 30s, the number of global muscle synergies extracted from the signal is 5, the number of global force synergies is 4, the experimental result of the multi-finger force estimation under six motions is shown in fig. 5, the average accuracy of the multi-finger force estimation under six motions can reach 83.7% by calculation of the decision coefficient, and the validity of the method provided by the present invention is verified.
The above examples are only for illustrating the technical idea and features of the present invention, and should not be construed as limiting the scope of the present invention. It will be appreciated by those skilled in the art that various modifications and changes may be made without departing from the spirit of the invention.
Claims (6)
1. A multi-finger force continuous estimation method based on collaborative mapping reconstruction is characterized by comprising the following steps:
(1) Data acquisition and preprocessing: when an object to be tested stably grabs an object under three different muscle contraction forces by six actions, namely isometric contraction, multi-channel electromyographic signals of the surfaces of the internal muscle and the external muscle of a hand and fingertip force signals of five fingers are synchronously acquired, and multi-channel envelope signals X of the two signals are respectively obtained through preprocessing M (t) and X F (t)；
(2) Muscle synergy analysis: multichannel envelope signal X for sEMG M (t) first use non-negative matrix factorization (NNMF)) The algorithm respectively extracts task-specific muscle cooperation W from the muscle coordination data M (i) (i =1,2, …, 6) and task-shared muscle synergy W M Then adopting a K mean clustering algorithm to cooperate W from task-specific muscles M (i) Extract global muscle synergy W MC Extracting corresponding global muscle synergistic activation coefficient H by adopting non-negative least square method (NNLS) MC (t)；
(3) Force synergy analysis: multichannel envelope signal X for fingertip force F (t) first, respectively extracting task specific forces from the data by using a nonnegative matrix factorization (NNMF) algorithm F (i) (i =1,2, …, 6) and tasking force synergy W F Then, cooperating W from task specific power by adopting a K mean value clustering algorithm F (i) Middle extracted global force synergy W FC Extracting corresponding global force cooperative activation coefficient H by adopting non-negative least square method (NNLS) FC (t)；
(4) Model training and testing: training a multiple linear regression model to coordinate the activation coefficient H of the global muscles MC (t) mapping to a Global force co-activation coefficient H FC (t) global force co-activation coefficient H obtained by regression FC (t) synergize with the global force W FC Reconstructing a real finger force signal, and then using the test set data for the test of the multiple linear regression model;
the multiple linear regression model combines the global muscle co-activation coefficient H MC (t) as model input, global force co-activation coefficient H FC (t) as model output, assuming a linear relationship exists between the two:
H FC (t)＝α×H MC (t)
wherein alpha is a regression coefficient matrix;
the model test flow is as follows: global muscle synergy W is known MC Cooperate with global force W FC For test set electromyographic signal envelope X M (t) extracting the corresponding global muscle co-activation coefficient H by using a non-negative least squares method (NNLS) MC (t), inputting the global force synergistic activation coefficient into a trained multiple linear regression model to obtain a global force synergistic activation coefficient H FC (t) then coordinating W with a global force FC Multiplying by a global force co-activation coefficient H FC (t) to reconstruct the true finger force signal X F (t) is the estimated force of the plurality of fingers
(5) And (3) correcting a finger force estimation result: the method comprises the steps of estimating the force of multiple fingers by using a multiple linear regression model, synchronously identifying the real activation state of each finger in a multi-finger grabbing task by adopting a task-specific muscle cooperation-based action identification method, and generating a corresponding correction vector psi according to the identification result i (i =1,2, …, 6) result of finger force estimation
And (5) correcting:
wherein
Is the final prediction result of the finger force.
2. The method of claim 1, wherein in step (1), the six actions comprise: thumb-index finger (TI), thumb-middle finger (TM), thumb-ring finger (TR), thumb-little finger (TL), thumb-index finger-middle finger (TIM), and all fingers (TIMRL);
three different muscle contraction forces include: 15% maximum autonomous contraction (MVC), 30% maximum autonomous contraction (MVC), and 45% maximum autonomous contraction (MVC);
the intrinsic muscles of the hand include: THE first dorsal interosseous muscle (FDI), THE dorsal interosseous muscle (DI), THE major thenar muscle (he) and THE minor thenar muscle (ADM);
the external muscles of the hand include: total Extensor Digitorum Communis (EDC) and superficial Flexor Digitorum (FDS);
the pre-processing includes 20-500Hz band-pass filtering, full-wave rectification, cut-off frequency 3Hz low-pass filtering and normalization.
3. The method of claim 1, wherein in step (2), task specific muscle coordination W M (i) By respectively corresponding single action electromyographic signal envelope X M (i) (t) (i =1,2, …, 6) as the NNMF input, and matrix X is divided into M (i) (t) new matrix W obtained after decomposition M (i) Matrix W as a task specific muscle synergy M (i) The number of lines is the number of muscle synergies specific to the task;
task consensus muscle synergy W M By enveloping X with electromyographic signals corresponding to respective movements M (i) (t) matrix X obtained after tandem M (t) as input to NNMF, and matrix X M (t) new matrix W obtained after decomposition M As task-shared muscle coordination, matrix W M The number of lines is the number of the common muscle synergies of the tasks;
global muscle synergy W MC The extraction process comprises the following steps: synergy of task-shared muscles W M The column vectors are respectively used as the initial value of the mass center of each cluster, and a K mean clustering algorithm is adopted to cooperate with the special muscles of the task by W M (i) All column vectors of (A) are subjected to clustering analysis, and the value of K and the task share muscle cooperation W M The column number of the new matrix W is the same, and the new matrix W is obtained by recombining the centroid vectors of all the clusters after clustering MC As global muscle synergy, matrix W MC The number of lines is the number of global muscle synergies;
global muscle co-activation coefficient H MC (t) extraction process: known training set electromyographic signal envelope X M (t) global muscle synergy W MC Extracting the global muscle synergistic activation coefficient H by adopting a non-negative least square method (NNLS) MC (t)。
4. The method of claim 1, wherein in step (3), the task-specific force is coordinated with W F (i) By respectively corresponding individual actionsForce signal envelope X F (i) (t) (i =1,2, …, 6) as the NNMF input, and matrix X is divided into F (i) (t) new matrix W obtained after decomposition F (i) As a task-specific force synergy, matrix W F (i) The number of columns is the number of the task specific force cooperation;
task-shared force synergy W F By enveloping X the force signal corresponding to each motion F (i) (t) matrix X obtained after tandem F (t) as input to NNMF, and X F (t) new matrix W obtained after decomposition F As a task-shared force synergy, matrix W F The number of the rows is the number of the task sharing force coordination;
global force synergy W FC The extraction process comprises the following steps: coordinating task consensus forces with W F The column vectors are respectively used as the initial value of the mass center of each cluster, and the K mean value clustering algorithm is adopted to cooperate with the specific force of the task to W F (i) All column vectors are subjected to clustering analysis, and the value of K and the task sharing force cooperate with W F The column number of the new matrix W is the same, and the new matrix W is obtained by recombining the centroid vectors of all the clusters after clustering FC As a global force synergy, matrix W FC The number of columns is the number of global force synergies;
global force co-activation coefficient H FC (t) extraction process: known training set force signal envelope X F (t) synergize with the global force W FC Extracting the global force synergistic activation coefficient H by adopting a non-negative least square method (NNLS) FC (t)。
5. The method as claimed in claim 1, wherein in step (5), the vector Ψ is modified by the finger force estimation results corresponding to six actions i Respectively as follows:
Ψ 1 ＝[1,1,0,0,0] T ；
Ψ 2 ＝[1,0,1,0,0] T ；
Ψ 3 ＝[1,0,0,1,0] T ；
Ψ 4 ＝[1,0,0,0,1] T ；
Ψ 5 ＝[1,1,1,0,0] T ；
Ψ 6 ＝[1,1,1,1,1] T ；
the action identification method based on task specific muscle cooperation comprises the following steps: sequential use of task specific muscle synergies W M (i) Electromyographic signal X to be identified M (t) reconstructing and calculating the reconstructed electromyographic signals respectively
And electromyographic signal X to be identified M (t) similarity between S i (i =1,2, …, 6), W with the highest similarity M (i) Corresponding action category is used as electromyographic signal X to be recognized M (t) the corresponding action category.
6. The method of claim 5, wherein the similarity S i And various evaluation indexes of included angle cosine, euclidean distance, pearson correlation coefficient and decision coefficient are adopted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211264621.1A CN115568856B (en) | 2022-10-17 | 2022-10-17 | Multi-finger force continuous estimation method based on collaborative mapping reconstruction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211264621.1A CN115568856B (en) | 2022-10-17 | 2022-10-17 | Multi-finger force continuous estimation method based on collaborative mapping reconstruction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115568856A true CN115568856A (en) | 2023-01-06 |
| CN115568856B CN115568856B (en) | 2024-05-28 |
Family
ID=84585060
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211264621.1A Active CN115568856B (en) | 2022-10-17 | 2022-10-17 | Multi-finger force continuous estimation method based on collaborative mapping reconstruction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115568856B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104337666A (en) * | 2014-11-05 | 2015-02-11 | 中山大学 | Multi-muscle collaborative myoelectricity feedback rehabilitation training system and method |
| CN106067178A (en) * | 2016-05-30 | 2016-11-02 | 杭州电子科技大学 | A kind of hand joint based on muscle synergistic activation model continuous method of estimation of motion |
| JP2018015405A (en) * | 2016-07-29 | 2018-02-01 | 日本電信電話株式会社 | Muscle activity analysis device, method, and program |
| KR20230061865A (en) * | 2021-10-29 | 2023-05-09 | 한국과학기술원 | Apparatus abnormal muscle synergy correction training and method of correcting muscle synergy correction using the same |
-
2022
- 2022-10-17 CN CN202211264621.1A patent/CN115568856B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104337666A (en) * | 2014-11-05 | 2015-02-11 | 中山大学 | Multi-muscle collaborative myoelectricity feedback rehabilitation training system and method |
| CN106067178A (en) * | 2016-05-30 | 2016-11-02 | 杭州电子科技大学 | A kind of hand joint based on muscle synergistic activation model continuous method of estimation of motion |
| JP2018015405A (en) * | 2016-07-29 | 2018-02-01 | 日本電信電話株式会社 | Muscle activity analysis device, method, and program |
| KR20230061865A (en) * | 2021-10-29 | 2023-05-09 | 한국과학기술원 | Apparatus abnormal muscle synergy correction training and method of correcting muscle synergy correction using the same |
Non-Patent Citations (1)
| Title |
|---|
| 桂奇政;孟明;马玉良;罗志增;: "基于肌肉协同激活模型的上肢关节运动连续估计", 仪器仪表学报, no. 06, 15 June 2016 (2016-06-15), pages 207 - 214 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115568856B (en) | 2024-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107378944B (en) | Multidimensional surface electromyographic signal artificial hand control method based on principal component analysis method | |
| Xu et al. | Advances and disturbances in sEMG-based intentions and movements recognition: A review | |
| Fang et al. | A multichannel surface EMG system for hand motion recognition | |
| CN107273798A (en) | A kind of gesture identification method based on surface electromyogram signal | |
| CN112732090B (en) | Muscle cooperation-based user-independent real-time gesture recognition method | |
| CN102306303B (en) | Electroencephalography signal characteristic extraction method based on small training samples | |
| Yang et al. | EMG pattern recognition using convolutional neural network with different scale signal/spectra input | |
| Zancanaro et al. | CNN-based approaches for cross-subject classification in motor imagery: From the state-of-the-art to DynamicNet | |
| Thomik et al. | Real-time movement prediction for improved control of neuroprosthetic devices | |
| Dwivedi et al. | EMG based decoding of object motion in dexterous, in-hand manipulation tasks | |
| Bai et al. | Multi-channel sEMG signal gesture recognition based on improved CNN-LSTM hybrid models | |
| Wang et al. | Deep convolutional neural network for decoding EMG for human computer interaction | |
| Zhao et al. | Online Decomposition of Surface Electromyogram into Individual Motor Unit Activities Using Progressive FastICA Peel-off | |
| CN109498362A (en) | A kind of hemiplegic patient's hand movement function device for healing and training and model training method | |
| CN112998725A (en) | Rehabilitation method and system of brain-computer interface technology based on motion observation | |
| CN115568856B (en) | Multi-finger force continuous estimation method based on collaborative mapping reconstruction | |
| Wang et al. | EMG-based hand gesture recognition by deep time-frequency learning for assisted living & rehabilitation | |
| Bhalerao et al. | Automatic detection of motor imagery EEG signals using swarm decomposition for robust BCI systems | |
| Sheng et al. | A hand gesture recognition using single-channel electrodes based on artificial neural network | |
| Lu et al. | Channel-distribution hybrid deep learning for sEMG-based gesture recognition | |
| Zhang et al. | Identification and Analysis of Limb Rehabilitation Signal Based on Wavelet Transform. | |
| CN114569143A (en) | Myoelectric gesture recognition method based on attention mechanism and multi-feature fusion | |
| CN114343679A (en) | Surface electromyogram signal upper limb action recognition method and system based on transfer learning | |
| Liu et al. | Real-time control of intelligent prosthetic hand based on the improved tcn | |
| Wolczowski et al. | Control of hand prosthesis using fusion of biosignals and information from prosthesis sensors |
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 |