CN111543985A - Brain control hybrid intelligent rehabilitation method based on novel deep learning model - Google Patents
Brain control hybrid intelligent rehabilitation method based on novel deep learning model Download PDFInfo
- Publication number
- CN111543985A CN111543985A CN202010364660.3A CN202010364660A CN111543985A CN 111543985 A CN111543985 A CN 111543985A CN 202010364660 A CN202010364660 A CN 202010364660A CN 111543985 A CN111543985 A CN 111543985A
- Authority
- CN
- China
- Prior art keywords
- electroencephalogram
- data set
- deep learning
- learning model
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- 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
- A61B5/369—Electroencephalography [EEG]
-
- 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
-
- 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
- A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
-
- 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
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B21/00—Exercising apparatus for developing or strengthening the muscles or joints of the body by working against a counterforce, with or without measuring devices
- A63B21/00181—Exercising apparatus for developing or strengthening the muscles or joints of the body by working against a counterforce, with or without measuring devices comprising additional means assisting the user to overcome part of the resisting force, i.e. assisted-active exercising
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B24/00—Electric or electronic controls for exercising apparatus of preceding groups; Controlling or monitoring of exercises, sportive games, training or athletic performances
- A63B24/0087—Electric or electronic controls for exercising apparatus of groups A63B21/00 - A63B23/00, e.g. controlling load
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- A—HUMAN NECESSITIES
- A63—SPORTS; GAMES; AMUSEMENTS
- A63B—APPARATUS FOR PHYSICAL TRAINING, GYMNASTICS, SWIMMING, CLIMBING, OR FENCING; BALL GAMES; TRAINING EQUIPMENT
- A63B2230/00—Measuring physiological parameters of the user
- A63B2230/08—Measuring physiological parameters of the user other bio-electrical signals
- A63B2230/10—Measuring physiological parameters of the user other bio-electrical signals electroencephalographic signals
- A63B2230/105—Measuring physiological parameters of the user other bio-electrical signals electroencephalographic signals used as a control parameter for the apparatus
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Veterinary Medicine (AREA)
- Pathology (AREA)
- Public Health (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Medical Informatics (AREA)
- Heart & Thoracic Surgery (AREA)
- Signal Processing (AREA)
- Theoretical Computer Science (AREA)
- Psychiatry (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Mathematical Physics (AREA)
- Physical Education & Sports Medicine (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Orthopedic Medicine & Surgery (AREA)
- Psychology (AREA)
- Power Engineering (AREA)
- Fuzzy Systems (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
- Rehabilitation Tools (AREA)
Abstract
A brain-controlled hybrid intelligent rehabilitation method based on a novel deep learning model is characterized in that a brain-computer interface formed by an electrode cap and portable electroencephalogram signal acquisition equipment is used for acquiring motor imagery electroencephalogram signals of a tested person, and a decoding model is used for decoding the motor intention of the tested person; controlling rehabilitation equipment to assist the testee in limb movement according to the movement intention of the testee; the decoding model is obtained by training a deep learning model by a training data set, and is continuously updated along with the promotion of rehabilitation training of a testee so as to adapt to the change of electroencephalogram characteristics of the testee; the training data set is derived from a motor imagery database, and the motor imagery database is used for storing collected electroencephalogram signal samples of the testee and carrying out time identification and marking of action labels. The invention can update the decoding model in real time to adapt to the electroencephalogram characteristic change of a patient in the training process, thereby effectively improving the classification effect of the electroencephalogram signals and achieving a more efficient rehabilitation training effect.
Description
Technical Field
The invention relates to a brain-controlled hybrid intelligent rehabilitation method. In particular to a brain control hybrid intelligent rehabilitation method based on a novel deep learning model.
Background
Deep learning is a branch of machine learning, and has been developed into a multi-field cross subject, which relates to multiple subjects such as probability theory, statistics, approximation theory, convex analysis, computational complexity theory and the like. The method tries to perform high-level abstraction on data by using a plurality of processing layers comprising complex structures or multiple nonlinear transformations, is a method based on characterization learning of data, is widely applied to a plurality of fields, and particularly has wide application prospects in a motor imagery brain-computer interface system. The brain-computer interface technology provides an effective way for realizing human-computer hybrid intelligence, and the brain-computer interface based on motor imagery is fused to a rehabilitation system, so that active rehabilitation training of affected limbs of a stroke patient can be facilitated, the participation degree of the patient is effectively improved, and the training efficiency is improved.
In general, training of a deep learning model requires extracting a corresponding relationship between event labels and data to train the model, so as to classify input new data to obtain corresponding event labels, and a large amount of labeled data is often required to train the model. In the classification algorithm for motor imagery, the electroencephalogram data of a patient are difficult to acquire, the effective data volume is small, and a high-quality model is difficult to train. Meanwhile, the individual difference of the electroencephalogram signals and the difference between individuals are large, namely the difference of the electroencephalogram signal characteristics of different individuals and the difference of the electroencephalogram signals of the same individual along with the propulsion of training are large, so that the classification performance is poor, and the accuracy of model training is further influenced.
Disclosure of Invention
The invention aims to solve the technical problem of providing a brain-controlled hybrid intelligent rehabilitation method based on a novel deep learning model, which can realize human-computer hybrid intelligent rehabilitation training.
The technical scheme adopted by the invention is as follows: a brain-controlled hybrid intelligent rehabilitation method based on a novel deep learning model comprises the steps of firstly obtaining motor imagery electroencephalogram signals of a tested person through a brain-computer interface formed by an electrode cap and portable electroencephalogram signal acquisition equipment, and decoding the motor intention of the tested person through a decoding model; controlling rehabilitation equipment to assist the testee in limb movement according to the movement intention of the testee; the decoding model is obtained by training a deep learning model by a training data set, and is continuously updated to adapt to the electroencephalogram characteristic change of the testee along with the promotion of rehabilitation training of the testee; the training data set is derived from a motor imagery database, and the motor imagery database is used for storing collected electroencephalogram signal samples of a testee and carrying out time identification and marking of action labels.
The deep learning model trained by the training data set comprises the following steps:
1) taking the current day as a reference, acquiring a historical electroencephalogram of the testee and determining electroencephalogram signal samples corresponding to the testee;
2) determining time identification information and label information of electroencephalogram signal samples to construct a training data set;
3) respectively confirming a plurality of sub-loss functions based on a plurality of corresponding relations in the training data set, and determining a final model loss function through weighted combination, wherein the plurality of corresponding relations in the training data set are EEG signal samples of each day and motor imagery action labels corresponding to the EEG signal samples;
4) constructing a deep learning model based on the model loss function and initializing;
5) taking the training data set as input and the motor imagery action label as output, optimizing parameters of the deep learning model through a minimization loss function to obtain a final decoding model, wherein the decoding model is used for decoding the electroencephalogram signal sample on the same day;
6) storing the current day electroencephalogram into a motor imagery database;
7) repeating the steps 1) to 6) on the basis of the following day, and updating the decoding model and the motor imagery database.
The electroencephalogram signal sample in the step 1) comprises electroencephalogram signal records of a subject lasting for multiple days, and when the historical electroencephalogram information of the subject is insufficient, the electroencephalogram information of other subjects is supplemented so as to supplement data.
The training data set in the step 2) comprises electroencephalogram signal samples, time identification information of the samples and label information; the time identification information is the number of acquisition days of the electroencephalogram sample, wherein the time identifications of the electroencephalogram information of other supplemented testees are uniformly defined as the total number of acquisition days of the electroencephalogram sample; the label information is the movement intention of the testee corresponding to the electroencephalogram sample; and then, filtering and channel optimization preprocessing are carried out on the multichannel electroencephalogram samples to obtain a pure training data set, and the moving labels are converted into one-hot codes for deep learning model training.
Each sub-loss function in the step 3) adopts a cross entropy loss function form to correspond to the motor imagery signal and the corresponding action label of the testee every day; the model loss function is a weighted combination of sub-loss functions of each day, the weight is determined according to the historical time, and the longer the historical time is, the smaller the weight is, so that the recent electroencephalogram characteristics are enhanced, and the long-term electroencephalogram characteristics are weakened.
Constructing a deep learning model and initializing, wherein the method comprises the steps of inputting the training data set into a time convolution layer containing a large convolution kernel to obtain a filtered time data set; inputting the time data set into a space convolution layer to obtain a single-channel time data set after space filtering; inputting the single-channel time data set into a nonlinear function layer containing baseline drift removal and square removal to obtain an energy data set; sequentially inputting the energy data set into the time convolution layer and the average pooling layer to obtain a fusion data set; and finally, inputting the fusion data set to a full connection layer to obtain a final output data set, initializing parameters of the deep learning model after the deep learning model is constructed, and combining a loss function to obtain the initialized deep learning model.
The first time convolution layer is used for extracting information of different filtering bands, the space convolution layer is used for carrying out space filtering on filtered data, the nonlinear function layer is used for extracting power spectrum energy of a single-channel time data set, the second time convolution layer and the average pooling layer are used for fusing energy information, and the full connection layer is used for identifying characteristics.
The model parameter optimization process of step 5), comprising: inputting the training data set into an initialized deep learning model to obtain an output result; calculating a loss value based on the output result and the corresponding action tag; and optimizing the parameters of the deep learning model by adopting a gradient descent algorithm according to the obtained loss value to obtain a final decoding model.
Step 7) the updating of the decoding model and the motor imagery database comprises: storing electroencephalogram information of the testees collected on the same day into a database for a new decoding model optimization process on the next day.
The brain-control hybrid intelligent rehabilitation method based on the novel deep learning model can realize human-computer hybrid intelligent rehabilitation training, and can be matched with the change of the electroencephalogram characteristics of a patient through the real-time updating of the decoding model of the equipment. The method is based on a deep learning model, the final model loss is obtained by acquiring the information of the current-day and historical electroencephalograms of patients and combining a plurality of sub-loss functions, and the optimal decoding model is obtained by optimizing model parameters through the minimized loss function. By combining the decoding model, the brain-control hybrid intelligent rehabilitation device can realize human-computer interaction active rehabilitation training, and the real-time updating of the decoding model is in accordance with the electroencephalogram characteristic change of a patient, so that the rehabilitation training effect is improved.
Drawings
FIG. 1 is a block diagram of a brain-controlled hybrid intelligent rehabilitation method based on a novel deep learning model;
FIG. 2 is a flow chart of a brain-controlled hybrid intelligent rehabilitation method based on a novel deep learning model according to the invention;
fig. 3 is a deep learning model architecture of the motor imagery decoding in the present invention.
Detailed Description
The brain control hybrid intelligent rehabilitation method based on the novel deep learning model is described in detail below with reference to the embodiment and the accompanying drawings.
As shown in fig. 1, the brain-controlled hybrid intelligent rehabilitation method based on the novel deep learning model of the invention firstly obtains motor imagery electroencephalogram signals of a subject through a brain-computer interface formed by an electrode cap and a portable electroencephalogram signal acquisition device, and decodes motor intentions of the subject through a decoding model; controlling rehabilitation equipment to assist the testee in limb movement according to the movement intention of the testee; the decoding model is obtained by training a deep learning model by a training data set, and is continuously updated to adapt to the electroencephalogram characteristic change of the testee along with the promotion of rehabilitation training of the testee; the training data set is derived from a motor imagery database, and the motor imagery database is used for storing collected electroencephalogram signal samples of a testee and carrying out time identification and marking of action labels.
The brain-computer interface can adopt a Neuroscan electroencephalogram acquisition system, an EmotivEpoc wireless portable electroencephalogram system or a NeuSen W series wireless electroencephalogram acquisition system.
As shown in fig. 2, the training of the deep learning model by the training data set includes:
1) taking the current day as a reference, acquiring a historical electroencephalogram of the testee and determining electroencephalogram signal samples corresponding to the testee;
the electroencephalogram signal sample comprises electroencephalogram signal records of a tested person lasting for a plurality of days, T is set as the lasting days, and when the historical electroencephalogram information of the tested person is less than T days, the electroencephalogram information of other tested persons is supplemented for supplementing data.
2) Determining time identification information and label information of electroencephalogram signal samples to construct a training data set;
the training data set comprises electroencephalogram signal samples, time identification information of the samples and label information; the time identification information is the acquisition days of the electroencephalogram sample, wherein the time identification of the electroencephalogram information of other supplemented testees is uniformly defined as the total days of acquiring the electroencephalogram sample, namely T days; the label information is the movement intention of the testee corresponding to the electroencephalogram sample; and then, filtering and channel optimization preprocessing are carried out on the multichannel electroencephalogram samples to obtain a pure training data set, and the moving labels are converted into one-hot codes for deep learning model training.
3) Respectively confirming a plurality of sub-loss functions based on a plurality of corresponding relations in the training data set, and determining a final model loss function through weighted combination, wherein the plurality of corresponding relations in the training data set are EEG signal samples of each day and motor imagery action labels corresponding to the EEG signal samples;
each sub-loss function is a motor imagery signal and a corresponding action tag which correspond to the testee every day in a cross entropy loss function form; the sub-loss function formula is as follows:
wherein losssIs the loss s days ago,
is a real label s days ago,
the predicted value s days ago, i represents the number of samples, and j represents the number of categories;
the model loss function is a weighted combination of sub-loss functions of each day, and the loss function of the model is calculated by adopting the following formula:
The weight is determined according to the historical time, and the longer the historical time is, the smaller the weight is, so as to strengthen the recent electroencephalogram characteristics and weaken the long-term electroencephalogram characteristics.
4) Constructing a deep learning model based on the model loss function and initializing;
as shown in fig. 3, the constructing of the deep learning model and the initialization include inputting the training data set into a time convolution layer containing a large convolution kernel to obtain a filtered time data set; inputting the time data set into a space convolution layer to obtain a single-channel time data set after space filtering; inputting the single-channel time data set into a nonlinear function layer containing baseline drift removal and square removal to obtain an energy data set; sequentially inputting the energy data set into the time convolution layer and the average pooling layer to obtain a fusion data set; and finally, inputting the fusion data set to a full connection layer to obtain a final output data set, initializing parameters of the deep learning model after the deep learning model is constructed, and combining a loss function to obtain the initialized deep learning model.
The first time convolution layer is used for extracting information of different filtering bands, the space convolution layer is used for carrying out space filtering on filtered data, the nonlinear function layer is used for extracting power spectrum energy of a single-channel time data set, the second time convolution layer and the average pooling layer are used for fusing energy information, and the full connection layer is used for identifying characteristics.
5) And (3) taking the training data set as input and the motor imagery action label as output, optimizing parameters of the deep learning model through a minimization loss function to obtain a final decoding model, wherein the decoding model is used for decoding the electroencephalogram signal sample on the same day.
The model parameter optimization process comprises the following steps: inputting the training data set into an initialized deep learning model to obtain an output result; calculating a loss value based on the output result and the corresponding action tag; and optimizing the parameters of the deep learning model by adopting a gradient descent algorithm according to the obtained loss value to obtain a final decoding model.
6) Storing the current day electroencephalogram into the motor imagery database
7) Repeating the steps 1) to 6) on the basis of the following day, and updating the decoding model and the motor imagery database, wherein the updating of the decoding model and the motor imagery database comprises the following steps: storing electroencephalogram information of the testees collected on the same day into a database for a new decoding model optimization process on the next day.
The above description of the present invention and the embodiments is not limited thereto, and the description of the embodiments is only one of the implementation manners of the present invention, and any structure or embodiment similar to the technical solution without inventive design is within the protection scope of the present invention without departing from the inventive spirit of the present invention.
Claims (9)
1. A brain-controlled hybrid intelligent rehabilitation method based on a novel deep learning model is characterized in that motor imagery electroencephalogram signals of a testee are obtained through a brain-computer interface formed by an electrode cap and portable electroencephalogram signal acquisition equipment, and the motor intention of the testee is decoded by a decoding model; controlling rehabilitation equipment to assist the testee in limb movement according to the movement intention of the testee; the decoding model is obtained by training a deep learning model by a training data set, and is continuously updated to adapt to the electroencephalogram characteristic change of the testee along with the promotion of rehabilitation training of the testee; the training data set is derived from a motor imagery database, and the motor imagery database is used for storing collected electroencephalogram signal samples of a testee and carrying out time identification and marking of action labels.
2. The brain-controlled hybrid intelligent rehabilitation method based on the novel deep learning model as claimed in claim 1, wherein the training of the deep learning model by the training data set comprises the following steps:
1) taking the current day as a reference, acquiring a historical electroencephalogram of the testee and determining electroencephalogram signal samples corresponding to the testee;
2) determining time identification information and label information of electroencephalogram signal samples to construct a training data set;
3) respectively confirming a plurality of sub-loss functions based on a plurality of corresponding relations in the training data set, and determining a final model loss function through weighted combination, wherein the plurality of corresponding relations in the training data set are EEG signal samples of each day and motor imagery action labels corresponding to the EEG signal samples;
4) constructing a deep learning model based on the model loss function and initializing;
5) taking the training data set as input and the motor imagery action label as output, optimizing parameters of the deep learning model through a minimization loss function to obtain a final decoding model, wherein the decoding model is used for decoding the electroencephalogram signal sample on the same day;
6) storing the current day electroencephalogram into a motor imagery database;
7) repeating the steps 1) to 6) on the basis of the following day, and updating the decoding model and the motor imagery database.
3. The brain-controlled hybrid intelligent rehabilitation method based on the novel deep learning model as claimed in claim 2, wherein the electroencephalogram signal samples in step 1) comprise electroencephalogram signal records of the subject lasting for a plurality of days, and when the historical electroencephalogram information of the subject is insufficient, the electroencephalogram information of other subjects is supplemented to supplement data.
4. The brain-controlled hybrid intelligent rehabilitation method based on the novel deep learning model as claimed in claim 2, wherein the training data set of step 2) comprises electroencephalogram signal samples and time identification information and label information of the samples; the time identification information is the number of acquisition days of the electroencephalogram sample, wherein the time identifications of the electroencephalogram information of other supplemented testees are uniformly defined as the total number of acquisition days of the electroencephalogram sample; the label information is the movement intention of the testee corresponding to the electroencephalogram sample; and then, filtering and channel optimization preprocessing are carried out on the multichannel electroencephalogram samples to obtain a pure training data set, and the moving labels are converted into one-hot codes for deep learning model training.
5. The brain-controlled hybrid intelligent rehabilitation method based on the novel deep learning model according to claim 2, characterized in that each sub-loss function in step 3) is a motor imagery signal and a corresponding action tag corresponding to a subject every day in a cross entropy loss function form; the model loss function is a weighted combination of sub-loss functions of each day, the weight is determined according to the historical time, and the longer the historical time is, the smaller the weight is, so that the recent electroencephalogram characteristics are enhanced, and the long-term electroencephalogram characteristics are weakened.
6. The brain-controlled hybrid intelligent rehabilitation method based on the novel deep learning model as claimed in claim 2, wherein the step 4) of constructing the deep learning model and initializing comprises inputting the training data set into a time convolution layer containing a large convolution kernel to obtain a filtered time data set; inputting the time data set into a space convolution layer to obtain a single-channel time data set after space filtering; inputting the single-channel time data set into a nonlinear function layer containing baseline drift removal and square removal to obtain an energy data set; sequentially inputting the energy data set into the time convolution layer and the average pooling layer to obtain a fusion data set; and finally, inputting the fusion data set to a full connection layer to obtain a final output data set, initializing parameters of the deep learning model after the deep learning model is constructed, and combining a loss function to obtain the initialized deep learning model.
7. The brain-controlled hybrid intelligent rehabilitation method based on the novel deep learning model as claimed in claim 6, wherein the first time convolution layer is used for extracting information of different filter bands, the spatial convolution layer is used for performing spatial filtering on the filtered data, the nonlinear function layer is used for extracting power spectrum energy of a single-channel time data set, the second time convolution layer and the average pooling layer are used for fusing energy information, and the full connection layer is used for feature recognition.
8. The brain-controlled hybrid intelligent rehabilitation method based on the novel deep learning model as claimed in claim 2, wherein the model parameter optimization process of step 5) comprises: inputting the training data set into an initialized deep learning model to obtain an output result; calculating a loss value based on the output result and the corresponding action tag; and optimizing the parameters of the deep learning model by adopting a gradient descent algorithm according to the obtained loss value to obtain a final decoding model.
9. The brain-controlled hybrid intelligent rehabilitation method based on the novel deep learning model as claimed in claim 2, wherein the step 7) of updating the decoding model and the motor imagery database comprises: storing electroencephalogram information of the testees collected on the same day into a database for a new decoding model optimization process on the next day.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010364660.3A CN111543985A (en) | 2020-04-30 | 2020-04-30 | Brain control hybrid intelligent rehabilitation method based on novel deep learning model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010364660.3A CN111543985A (en) | 2020-04-30 | 2020-04-30 | Brain control hybrid intelligent rehabilitation method based on novel deep learning model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111543985A true CN111543985A (en) | 2020-08-18 |
Family
ID=71999841
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010364660.3A Pending CN111543985A (en) | 2020-04-30 | 2020-04-30 | Brain control hybrid intelligent rehabilitation method based on novel deep learning model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111543985A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114767120A (en) * | 2022-04-25 | 2022-07-22 | 上海韶脑传感技术有限公司 | Depth learning-based selection method for motor imagery electroencephalogram channels of unilateral limb patients |
| CN116089798A (en) * | 2023-02-07 | 2023-05-09 | 华东理工大学 | Decoding method and device for finger movement |
-
2020
- 2020-04-30 CN CN202010364660.3A patent/CN111543985A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114767120A (en) * | 2022-04-25 | 2022-07-22 | 上海韶脑传感技术有限公司 | Depth learning-based selection method for motor imagery electroencephalogram channels of unilateral limb patients |
| CN114767120B (en) * | 2022-04-25 | 2024-05-10 | 上海韶脑传感技术有限公司 | Single-side limb patient motor imagery electroencephalogram channel selection method based on deep learning |
| CN116089798A (en) * | 2023-02-07 | 2023-05-09 | 华东理工大学 | Decoding method and device for finger movement |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111544855B (en) | Pure idea control intelligent rehabilitation method based on distillation learning and deep learning and application | |
| CN111584029B (en) | Electroencephalogram self-adaptive model based on discriminant confrontation network and application of electroencephalogram self-adaptive model in rehabilitation | |
| CN110472649B (en) | Electroencephalogram emotion classification method and system based on multi-scale analysis and integrated tree model | |
| CN113951900B (en) | Motor imagery intention recognition method based on multi-mode signals | |
| CN109934125B (en) | Semi-supervised operation video flow identification method | |
| CN111407243B (en) | Pulse signal pressure identification method based on deep learning | |
| CN106725452A (en) | Based on the EEG signal identification method that emotion induces | |
| CN114564990B (en) | Electroencephalogram signal classification method based on multichannel feedback capsule network | |
| CN111544256A (en) | Brain-controlled intelligent full limb rehabilitation method based on graph convolution and transfer learning | |
| CN113505719B (en) | Gait recognition model compression system and method based on local-integral combined knowledge distillation algorithm | |
| CN111543985A (en) | Brain control hybrid intelligent rehabilitation method based on novel deep learning model | |
| CN108364062B (en) | Deep learning model construction method based on MEMD and application of deep learning model in motor imagery | |
| CN117520891A (en) | Motor imagery electroencephalogram signal classification method and system | |
| CN115238796A (en) | Motor imagery electroencephalogram signal classification method based on parallel DAMSCN-LSTM | |
| CN114209323A (en) | Method for recognizing emotion and emotion recognition model based on electroencephalogram data | |
| CN116225222A (en) | Brain-computer interaction intention recognition method and system based on lightweight gradient lifting decision tree | |
| CN113128353B (en) | Emotion perception method and system oriented to natural man-machine interaction | |
| CN114417836A (en) | Deep learning-based Chinese electronic medical record text semantic segmentation method | |
| CN114145744A (en) | Cross-device forehead electroencephalogram emotion recognition method and system | |
| CN113951883A (en) | Gender difference detection method based on electroencephalogram signal emotion recognition | |
| CN117158994A (en) | Brain electrical signal classification method and system for cerebral apoplexy patient based on motor imagery | |
| CN117290730A (en) | Optimization method of individual emotion recognition model | |
| CN114983447B (en) | Human action recognition, analysis and storage wearable device based on AI technology | |
| CN116421200A (en) | Brain electricity emotion analysis method of multi-task mixed model based on parallel training | |
| CN114004257A (en) | Myoelectric gesture recognition method based on lightweight convolutional neural network |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200818 |