CN114931389B - Myoelectric signal identification method based on residual error network and graph convolution network - Google Patents
Myoelectric signal identification method based on residual error network and graph convolution network Download PDFInfo
- Publication number
- CN114931389B CN114931389B CN202210449287.0A CN202210449287A CN114931389B CN 114931389 B CN114931389 B CN 114931389B CN 202210449287 A CN202210449287 A CN 202210449287A CN 114931389 B CN114931389 B CN 114931389B
- Authority
- CN
- China
- Prior art keywords
- network
- electromyographic signal
- muscle
- graph
- residual
- 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
- 238000000034 method Methods 0.000 title claims abstract description 11
- 230000003183 myoelectrical effect Effects 0.000 title description 6
- 238000005096 rolling process Methods 0.000 claims abstract description 7
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims abstract description 4
- 210000003205 muscle Anatomy 0.000 claims description 56
- 239000011159 matrix material Substances 0.000 claims description 32
- 238000004364 calculation method Methods 0.000 claims description 9
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 8
- 230000009466 transformation Effects 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 6
- 238000011176 pooling Methods 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 4
- 230000002195 synergetic effect Effects 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 230000015556 catabolic process Effects 0.000 abstract description 4
- 238000006731 degradation reaction Methods 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 18
- 230000004913 activation Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000000977 initiatory effect Effects 0.000 description 2
- 230000033001 locomotion Effects 0.000 description 2
- 230000036982 action potential Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000002567 electromyography Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000001087 myotubule Anatomy 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
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]
-
- 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
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- 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
- A61B5/7225—Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
-
- 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
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Public Health (AREA)
- Surgery (AREA)
- Veterinary Medicine (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Physiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychiatry (AREA)
- Power Engineering (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention relates to an electromyographic signal identification method based on a residual error network and a graph convolution network, which comprises the following steps of: step S1, acquiring a multichannel electromyographic signal, and preprocessing the multichannel electromyographic signal to obtain a data set to be trained; step S2, selecting a time window meeting real-time performance and a corresponding sliding window to sample electromyographic signals and constructing a surface electromyographic signal image, and step S3, constructing a residual network and a structural model of a graph rolling network and training based on the surface electromyographic signal image; and S4, identifying the surface electromyographic signals based on the structural model of the trained residual error network and the graph convolution network. The invention can better learn the characteristics of the electromyographic signals, and simultaneously solves the degradation problem of the deep convolution network model because the residual network structure is reserved.
Description
Technical Field
The invention relates to an electromyographic signal identification method based on a residual error network and a graph convolution network.
Background
The surface electromyographic signals (Surface Electromyography) are superposition of action potentials of the motion units in a plurality of muscle fibers in time and space, and can reflect relevant motions of muscles by identifying the surface electromyographic signals, so that the surface electromyographic signal identification is widely applied to various fields of medical treatment, military, education, life and the like, for example, by means of electromyographic signal gesture identification, the electromyographic signals are used as control signals of artificial limbs, and real-time control of the artificial limbs is realized.
The traditional electromyographic signal identification mainly depends on a machine learning method, and commonly comprises a support vector machine, a random deep forest, a neural network and the like. Deep learning, which is a branch of machine learning, has been widely used in myoelectricity recognition in recent years. The convolutional neural network model has higher recognition rate than the traditional machine learning method, but as the depth of the network increases, the convolutional neural network can generate a model degradation problem, and a residual network can be introduced to solve the problem. However, the residual network is proposed based on picture classification, and is improved in order to improve the accuracy and the applicability of the residual network in the electromyographic signal identification field. Muscle synergy is also an important point in surface electromyographic signal research, and muscle synergy analysis is to obtain a muscle synergy element matrix and an activation coefficient matrix by decomposing the activation degrees of a plurality of muscle channels. However, the activation degree information of the muscle is not directly obtained, and the activation degree of the muscle needs to be obtained through acquiring surface electromyographic signal information of the muscle and corresponding processing.
Disclosure of Invention
In view of the above, the present invention aims to provide a myoelectric signal identification method based on a residual network and a graph convolution network, which can better learn the characteristics of the myoelectric signal, and meanwhile, the degradation problem of a deep convolution network model is solved due to the residual network structure being reserved.
In order to achieve the above purpose, the invention adopts the following technical scheme:
an electromyographic signal identification method based on a residual network and a graph convolution network comprises the following steps:
step S1, acquiring a multichannel electromyographic signal, and preprocessing the multichannel electromyographic signal to obtain a data set to be trained;
s2, selecting a time window meeting preset requirements and a corresponding sliding window to sample electromyographic signals, and constructing a surface electromyographic signal image;
s3, constructing a structural model of a residual error network and a graph convolution network, and training based on surface electromyographic signal images;
and S4, identifying the surface electromyographic signals based on the structural model of the trained residual error network and the graph convolution network.
Further, the pretreatment specifically comprises: and carrying out baseline filtering and power frequency filtering treatment on the multichannel electromyographic signals and off-line marking.
Further, the residual network structure is composed of an initial convolution layer, a plurality of residual blocks and a full connection layer, and is specifically as follows: the initial convolution kernel is changed to 1×n, the maximum pooling layer is deleted, the convolution kernel size in each residual block is changed to 1×n, and the step size of each convolution kernel is adjusted.
Further, the graph convolution network initiation unit is a joint voting network.
Further, the step S4 specifically includes:
The electromyographic signal characteristics X epsilon H multiplied by W multiplied by C, H, W and C are respectively the height, the broadband and the channel number of the electromyographic signal characteristics, the characteristic matrix of each muscle is obtained by voting weight calculation on the electromyographic signal characteristics, and the voting weight matrix is expressed as follows:
W＝Φ(ψ(X)) (1)
Wherein ψ (·) is the transformation function implemented by a1×1 convolution, Φ is the spatial softmax normalization, W e h×w×n, where N is the number of muscle blocks; w K E H W is the voting weight matrix of the kth muscle; f K represents the characteristics of the kth muscle, expressed as a weighted average of all electromyographic signal characteristics, calculated as follows:
where X i represents the electromyographic signal characteristic of the ith channel, Is a transform function implemented by a1 x 1 convolutional layer, W Ki is the corresponding element in the voting weight matrix W K. The overall characteristics of all muscles are defined as:
since the cooperative relationship between the muscles is unknown, the connection graph G between the muscles is defined as full connection, i.e., it is assumed that each muscle has a cooperative relationship with the rest of the muscles;
The new muscle cooperative characteristic F e is automatically extracted from the arrived muscle characteristics of the voting network through the graph rolling network, and the calculation formula is as follows:
Wherein a e = a + I, because graph G is a fully connected graph, a is a matrix with diagonal elements all 0 and the remaining elements all 1, I is a unitary matrix, matrix D is a degree matrix of a, W e is a trainable transformation matrix, σ (·) is a nonlinear function;
After obtaining the new muscle cooperative characteristics, carrying out the inverse operation of the joint voting to obtain the muscle cooperative characteristics with the same dimension as the electromyographic signal characteristics X, wherein the calculation formula is as follows:
Wherein the method comprises the steps of For the new characteristics of the kth muscle obtained through the graph convolution network, W ik is equal to the voting weight matrix W Ki obtained before; ρ (·) is a nonlinear function with BN layer;
Finally, combining the electromyographic signal characteristic X obtained by the residual network with the muscle synergic characteristic F q to obtain a new enhancement characteristic X e, wherein the formula is as follows:
where τ (·) is a nonlinear function with BN layer.
Compared with the prior art, the invention has the following beneficial effects:
The invention can better learn the characteristics of the electromyographic signals and learn the muscle cooperative characteristics end to end, and simultaneously solves the degradation problem of the deep convolution network model because the residual network structure is reserved.
Drawings
FIG. 1 is a diagram of a residual network and graph roll-up network architecture model of the present invention;
FIG. 2 is a diagram of improved residual network model structure and parameters in accordance with one embodiment of the present invention;
FIG. 3 is a diagram of a rolling network model of the present invention in one embodiment.
Detailed Description
The invention will be further described with reference to the accompanying drawings and examples.
The invention provides an electromyographic signal identification method based on a residual error network and a graph convolution network, which comprises the following steps of:
Step S1, acquiring a multichannel electromyographic signal, carrying out baseline filtering and power frequency filtering treatment on the multichannel electromyographic signal, and carrying out off-line marking to obtain a data set to be trained;
s2, selecting a time window meeting real-time performance and a corresponding sliding window to sample the electromyographic signals, and constructing a surface electromyographic signal image;
s3, constructing a structural model of a residual error network and a graph convolution network, and training based on surface electromyographic signal images;
and S4, identifying the surface electromyographic signals based on the structural model of the trained residual error network and the graph convolution network.
Referring to fig. 1, the present embodiment provides a structure model for combining a modified residual network and a graph roll-up network to construct the residual network and the graph roll-up network,
Electromyographic signal features X are extracted from the residual network, and the graph rolling network (Graph Convolutional Network, GCN) extracts muscle synergic features F q.
Combining the electromyographic signal characteristic and the muscle cooperative characteristic to obtain a new enhancement characteristic X e.
And then carrying out global average pooling on the enhancement features, then sending the enhancement features into a neural network for electromyographic signal recognition, and finally outputting electromyographic signal recognition results, such as electromyographic signal gesture classification results.
In this embodiment, the residual network structure is composed of an initial convolution layer, a plurality of residual blocks and a full connection layer, and specifically includes the following steps: the initial convolution kernel is changed to 1×n, the maximum pooling layer is deleted, the convolution kernel size in each residual block is changed to 1×n, and the step size of each convolution kernel is adjusted.
Preferably, the specific parameter structure is exemplified in fig. 2. In the figure, the initial convolution kernel is 1×9, the step size is 1, p=1 represents that the padding modes involved in convolution are all same, and s is the step size. In the figure, 64×5conv+bn+relu represents a convolution layer, its output channel is 64, the convolution kernel size is 1*5, BN represents using batch normalization, RELU represents the activation function as relu function; in the figure, there are two modes of solid line jump and broken line jump, the broken line jump represents the same dimension between the input and the output to be directly combined, the combined characteristic is activated by relu functions, the solid line jump represents different dimensions between the input and the output, the input is combined with the output after the convolution layer with the convolution kernel of 1 multiplied by 1 and the step length of 1 and the padding mode of same is used for carrying out the up-down dimension, and the combined characteristic is activated by relu functions. The convolution layers in the figure comprise an initial convolution layer with a convolution kernel of 1 multiplied by 9 and 16 residual block convolution layers with a convolution kernel of 1 multiplied by 5, the residual block convolution layers are fed into a full-connection layer after global average pooling, an activation function of the full-connection layer is a softmax function, and finally a myoelectric signal recognition result, such as a myoelectric signal gesture classification result, is output.
In this embodiment, referring to fig. 3, the graph convolution network initiation unit is a joint voting network.
The electromyographic signal characteristics X epsilon H multiplied by W multiplied by C, H, W and C are respectively the height, the broadband and the channel number of the electromyographic signal characteristics, the characteristic matrix of each muscle is obtained by voting weight calculation on the electromyographic signal characteristics, and the voting weight matrix is expressed as follows:
W＝Φ(ψ(X)) (1)
Where ψ (·) is the transformation function implemented by a1×1 convolution, Φ is the spatial softmax normalization, W e h×w×n, where N is the number of muscle blocks. W K E H W is the voting weight matrix of the kth muscle. f K represents the characteristics of the kth muscle, expressed as a weighted average of all electromyographic signal characteristics, calculated as follows:
where X i represents the electromyographic signal characteristic of the ith channel, Is a transform function implemented by a1 x 1 convolutional layer, W Ki is the corresponding element in the voting weight matrix W K. The overall characteristics of all muscles are defined as:
Since the cooperative relationship between muscles is unknown, the linkage map G between muscles is defined as full connection, i.e., it is assumed that each muscle has a cooperative relationship with the rest of the muscles. The new muscle cooperative characteristic F e is automatically extracted from the arrived muscle characteristics of the voting network through the graph rolling network, and the calculation formula is as follows:
Where a e = a + I, because graph G is a fully connected graph, a is a matrix with diagonal elements all 0 and the remaining elements all 1, I is an identity matrix, matrix D is a degree matrix of a, W e is a trainable transformation matrix, σ (·) is a nonlinear function.
After obtaining the new muscle cooperative characteristics, carrying out the inverse operation of the joint voting to obtain the muscle cooperative characteristics with the same dimension as the electromyographic signal characteristics X, wherein the calculation formula is as follows:
Wherein the method comprises the steps of For the new feature of the kth muscle obtained via the graph convolution network, W ik is equal to the voting weight matrix W Ki obtained previously. ρ (·) is a nonlinear function with BN layer (the activation function used in the present invention is ReLU)
And finally, combining the electromyographic signal characteristic X obtained by the residual network with the muscle synergistic characteristic F q to obtain a new enhancement characteristic X e. The formula is as follows:
Where τ (·) is a nonlinear function with BN layer, the activation function used in this example is preferably ReLU.
The foregoing description is only of the preferred embodiments of the invention, and all changes and modifications that come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.
Claims (2)
1. The electromyographic signal identification method based on the residual network and the graph convolution network is characterized by comprising the following steps of:
step S1, acquiring a multichannel electromyographic signal, and preprocessing the multichannel electromyographic signal to obtain a data set to be trained;
s2, selecting a time window meeting preset requirements and a corresponding sliding window to sample electromyographic signals, and constructing a surface electromyographic signal image;
s3, constructing a structural model of a residual error network and a graph convolution network, and training based on surface electromyographic signal images;
s4, identifying the surface electromyographic signals based on the structural model of the trained residual error network and the graph convolution network;
the residual network structure consists of an initial convolution layer, a plurality of residual blocks and a full connection layer, and is specifically as follows: changing the initial convolution kernel into 1 Xn, deleting the maximum pooling layer, changing the convolution kernel size in each residual block into 1 Xn, and simultaneously adjusting the step length of each convolution kernel;
the initial unit of the graph convolution network is a joint voting network;
the step S4 specifically includes:
The electromyographic signal characteristics X epsilon H multiplied by W multiplied by C, H, W and C are respectively the height, the broadband and the channel number of the electromyographic signal characteristics, the characteristic matrix of each muscle is obtained by voting weight calculation on the electromyographic signal characteristics, and the voting weight matrix is expressed as follows:
W ＝ Φ (ψ (X )) (1)
Wherein ψ (·) is the transformation function implemented by a1×1 convolution, Φ is the spatial softmax normalization, W e h×w×n, where N is the number of muscle blocks; w K E H W is the voting weight matrix of the kth muscle; f K represents the characteristics of the kth muscle, expressed as a weighted average of all electromyographic signal characteristics, calculated as follows:
where X i represents the electromyographic signal characteristic of the ith channel, Is a transformation function implemented by a1 x 1 convolutional layer, W Ki is the corresponding element in the voting weight matrix W K; the overall characteristics of all muscles are defined as:
since the cooperative relationship between the muscles is unknown, the connection graph G between the muscles is defined as full connection, i.e., it is assumed that each muscle has a cooperative relationship with the rest of the muscles;
The new muscle cooperative characteristic F e is automatically extracted from the arrived muscle characteristics of the voting network through the graph rolling network, and the calculation formula is as follows:
Wherein a e = a + I, because graph G is a fully connected graph, a is a matrix with diagonal elements all 0 and the remaining elements all 1, I is a unitary matrix, matrix D is a degree matrix of a, W e is a trainable transformation matrix, σ (·) is a nonlinear function;
After obtaining the new muscle cooperative characteristics, carrying out the inverse operation of the joint voting to obtain the muscle cooperative characteristics with the same dimension as the electromyographic signal characteristics X, wherein the calculation formula is as follows:
Cik＝WikFe k (5)
Wherein F e k is the new feature of the kth muscle obtained by the graph rolling network, and W ik is equal to the voting weight matrix W Ki obtained before; ρ (·) is a nonlinear function with BN layer;
Finally, combining the electromyographic signal characteristic X obtained by the residual network with the muscle synergic characteristic F q with the same latitude as the electromyographic signal characteristic X to obtain a new enhancement characteristic X e, wherein the formula is as follows:
Xe＝τ([X,Fq]) (7)
wherein τ (·) is a nonlinear function with BN layer;
And then carrying out global average pooling on the enhancement features, then sending the enhancement features into a neural network for electromyographic signal identification, and finally outputting an electromyographic signal identification result.
2. The electromyographic signal identification method based on a residual network and a graph convolution network according to claim 1, wherein the preprocessing specifically comprises: and carrying out baseline filtering and power frequency filtering treatment on the multichannel electromyographic signals and off-line marking.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210449287.0A CN114931389B (en) | 2022-04-27 | 2022-04-27 | Myoelectric signal identification method based on residual error network and graph convolution network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210449287.0A CN114931389B (en) | 2022-04-27 | 2022-04-27 | Myoelectric signal identification method based on residual error network and graph convolution network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114931389A CN114931389A (en) | 2022-08-23 |
| CN114931389B true CN114931389B (en) | 2024-07-05 |
Family
ID=82861829
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210449287.0A Active CN114931389B (en) | 2022-04-27 | 2022-04-27 | Myoelectric signal identification method based on residual error network and graph convolution network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114931389B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108388348A (en) * | 2018-03-19 | 2018-08-10 | 浙江大学 | A kind of electromyography signal gesture identification method based on deep learning and attention mechanism |
| CN110598676A (en) * | 2019-09-25 | 2019-12-20 | 南京邮电大学 | Deep learning gesture electromyographic signal identification method based on confidence score model |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111870242A (en) * | 2020-08-03 | 2020-11-03 | 南京邮电大学 | Intelligent gesture action generation method based on electromyographic signals |
| CN112244851A (en) * | 2020-11-13 | 2021-01-22 | 山东中科先进技术研究院有限公司 | Muscle movement recognition method and surface electromyogram signal acquisition device |
| CN113934302B (en) * | 2021-10-21 | 2024-02-06 | 燕山大学 | Myoelectric gesture recognition method based on SeNet and gating time sequence convolution network |
-
2022
- 2022-04-27 CN CN202210449287.0A patent/CN114931389B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108388348A (en) * | 2018-03-19 | 2018-08-10 | 浙江大学 | A kind of electromyography signal gesture identification method based on deep learning and attention mechanism |
| CN110598676A (en) * | 2019-09-25 | 2019-12-20 | 南京邮电大学 | Deep learning gesture electromyographic signal identification method based on confidence score model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114931389A (en) | 2022-08-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110287805B (en) | Micro-expression identification method and system based on three-stream convolutional neural network | |
| CN109346063B (en) | Voice data enhancement method | |
| CN109685716B (en) | Image super-resolution reconstruction method for generating countermeasure network based on Gaussian coding feedback | |
| CN111584029B (en) | Electroencephalogram self-adaptive model based on discriminant confrontation network and application of electroencephalogram self-adaptive model in rehabilitation | |
| CN109031421B (en) | A kind of stack velocity spectrum pick-up method and processing terminal based on deeply study | |
| CN112818764A (en) | Low-resolution image facial expression recognition method based on feature reconstruction model | |
| Montazerin et al. | ViT-HGR: vision transformer-based hand gesture recognition from high density surface EMG signals | |
| DE69124231T2 (en) | ARTIFICIAL NEURONAL PLANT | |
| CN112766413A (en) | Bird classification method and system based on weighted fusion model | |
| CN111598213A (en) | Network training method, data identification method, device, equipment and medium | |
| CN111616706A (en) | Surface electromyogram signal classification method and system based on convolutional neural network | |
| CN112215339A (en) | Medical data expansion method based on generation of countermeasure network | |
| CN112233199A (en) | fMRI visual reconstruction method based on discrete characterization and conditional autoregression | |
| CN114820828A (en) | Vision transform-based image compressed sensing reconstruction method and system | |
| CN115105094A (en) | Attention and 3D dense connection neural network-based motor imagery classification method | |
| CN113569735B (en) | Complex input feature graph processing method and system based on complex coordinate attention module | |
| CN114931389B (en) | Myoelectric signal identification method based on residual error network and graph convolution network | |
| CN113032535A (en) | Visual question and answer method and device for assisting visually impaired people, computing equipment and storage medium | |
| Li et al. | A novel motor imagery EEG recognition method based on deep learning | |
| CN112307996A (en) | Fingertip electrocardiogram identity recognition device and method | |
| CN117150346A (en) | EEG-based motor imagery electroencephalogram classification method, device, equipment and medium | |
| CN112418068A (en) | On-line training effect evaluation method, device and equipment based on emotion recognition | |
| CN116596046A (en) | Method for reconstructing image by utilizing electroencephalogram signals and visual features | |
| CN116503499A (en) | Sketch drawing generation method and system based on cyclic generation countermeasure network | |
| DE102019204118A1 (en) | Method for transferring a feature of a first image to a second image |
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 |