CN109814716A - A kind of motion intention coding/decoding method based on dynamic surface electromyography signal - Google Patents
A kind of motion intention coding/decoding method based on dynamic surface electromyography signal Download PDFInfo
- Publication number
- CN109814716A CN109814716A CN201910084116.0A CN201910084116A CN109814716A CN 109814716 A CN109814716 A CN 109814716A CN 201910084116 A CN201910084116 A CN 201910084116A CN 109814716 A CN109814716 A CN 109814716A
- Authority
- CN
- China
- Prior art keywords
- electromyography signal
- classifier
- dynamic
- decoding
- signal
- 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
Abstract
The present invention relates to a kind of motion intention coding/decoding methods based on dynamic surface electromyography signal, the electromyography signal for constructing training set is chosen first, secondly the electromyography signal of selection is pre-processed, generate the input sample collection that can be used for inputting classifier, and with the training set off-line training classifier constructed, then the electromyography signal for being used for off-line test classifier is chosen, optimum training collection is selected according to classifier decoding accuracy rate, and best features, the optimal classification device that off-line test is gone out is as classifier used in online recognition, using the resulting recognition result of online recognition as the control signal of man-machine interface, realize the real-time control of myoelectric control system.
Description
Technical field
The present invention relates to electromyography signal process field, especially a kind of motion intention solution based on dynamic surface electromyography signal
Code method.
Background technique
It is impaired that the positions function such as manpower is frequently accompanied by after neurotrosis, such as apoplexy and spinal cord injury, robot assisted instruction
Practice and provide effective ways for neurotrosis rehabilitation, has developed various robots and ectoskeleton for rehabilitation training, ground
Studying carefully the human-computer interactive control system for showing to have user's independent desire to participate in can be improved the safety and treatment effect of rehabilitation training
Fruit.And electromyography signal can represent nervimuscular activity as a kind of bioelectrical signals, react independent desire, characterize people's
Motion intention has been widely used for providing motion control information abundant for medical aid etc..
In the open technology used, electromyography signal is used to first have to detect when control system the starting point of electromyography signal, cuts
Input signal of the latter section of time window of starting point as disaggregated model is taken, original electromyography signal inputs classifier after feature extraction
Intention decoding is carried out, decoded result signal as input to the controller will be intended to.Surface electromyogram signal has dynamic part and steady
Polymorphic segment, in these techniques, the electromyography signal for real-time grading are the dynamic electromyography signal after starting-tool point, but are not described
Classifier how is trained to carry out off-line modeling, and that has delivered at present be intended in decoded technology to surface electromyogram signal,
For training the data set of classifier to be all based on the steady-state portion of electromyography signal.With the training of stable state electromyography signal and with stable state flesh
The technology of electric signal test can obtain high discrimination, but in real-time control system, the signal for inputting classifier is myoelectricity
The dynamic part of signal, what this classifier based on stable state electromyography signal can not obtain under the test of dynamic electromyography signal
Classification results, in addition, if being carried out being intended to decoding online with stable state electromyography signal, it means that having ignored static to stable state
Transient process, cause system delay.
Existing technology of having delivered also indicates that containing in dynamic electromyography signal has action message, deposits for having delivered in technology
Problem above, this patent propose it is a kind of be suitable for the decoded method of dynamic surface electromyography signal motion intention, to improve flesh
The real-time of electric control system.
Summary of the invention
In view of this, the purpose of the present invention is to propose to a kind of motion intention decoding sides based on dynamic surface electromyography signal
Method can be improved the real-time of myoelectric control system.
The present invention is realized using following scheme: a kind of motion intention coding/decoding method based on dynamic surface electromyography signal, packet
Include following steps:
Step S1: different training set structural schemes are designed；
Step S2: original signal is pre-processed；
Step S3: training classifier；
Step S4: the dynamic electromyography signal for being used to testing classification device is chosen；
Step S5: testing trained classifier and selects optimal characteristics and training set structural scheme；
Step S6: online to be intended to decoding.
The invention is intended to solve motion intention decoding problem.By selecting suitable classifier, training set and feature, instruction
Practice and be suitble to carry out dynamic electromyography signal to be intended to decoded classifier, realizes in myoelectric control system reference time delay to dynamic flesh
Electric signal is intended to decoded purpose.
Further, in step S1, the training set includes three classes, respectively pure dynamic electromyography signal, pure stable state myoelectricity
The binding signal of signal and stable state electromyography signal and dynamic electromyography signal.
Further, step S2 specifically includes the following steps:
Step S21: training dataset is subjected to windowing process；
Step S22: calculating the myoelectricity feature in each time window, compares different characteristic to the decoded shadow of dynamic electromyography signal
It rings, and carries out feature extraction；
Step S23: the input sample for meeting classifier requirement is generated.
Further, step S21 specifically: dynamic electromyography signal is split with time window, passes through the shape of overlapping window
Original signal is configured to multiple samples by formula；Number of windows after segmentation is calculated using following formula:
Further, in step S22, the feature that electromyography signal is used in carrying out intention decoding includes temporal signatures, frequency
Characteristic of field and time and frequency domain characteristics.
Further, step S3 specifically: classifier uses convolutional neural networks transfer learning method, with by step S1
With the processed different training set off-line training convolutional neural networks of step S2, trained classifier is obtained.
Further, step S5 specifically: using the preprocess method of step S2, be used to testing by what step S4 chose
Dynamic electromyography signal is configured to the input sample for testing classification device, then each trained classifier of off-line test, choosing
The decoding highest classifier of accuracy is taken as step S6 and is intended to decoding classifier used online, for training the classifier
Training set structural scheme is optimal case, and corresponding characteristic quantity is the feature of optimal dynamic electromyography signal.
Further, step S6 specifically: mobile according to time window splitting scheme since dynamic electromyography signal and calculate
Characteristic value inputs trained classifier after forming input picture, obtains real-time decoding result.
Secondly the present invention carries out the electromyography signal of selection pre- firstly the need of the electromyography signal chosen for constructing training set
Processing generates and can be used for inputting the input sample collection of classifier, and with the training set off-line training classifier constructed, then selects
It takes in the electromyography signal of off-line test classifier, selects optimum training collection according to classifier decoding accuracy rate, and best special
Sign, the optimal classification device that off-line test is gone out is as classifier used in online recognition, by the resulting recognition result of online recognition
As the control signal of man-machine interface, the real-time control of myoelectric control system is realized.
Compared with prior art, the invention has the following beneficial effects:
1, the decoding of the intention based on dynamic electromyography signal may be implemented in the present invention, has preferably property in terms of reducing delay
Energy.
2, the present invention can be used in human-computer interactive control system, be anticipated by dynamic surface surface electromyogram signal decoding moving
Figure uses dynamic electromyography signal decoding result as the control signal of man-machine interactive system, reduces intention assessment delay, and realization is based on
The real-time control of the man-machine interactive system of electromyography signal.
Detailed description of the invention
Fig. 1 is the principle process schematic diagram of the embodiment of the present invention.
Fig. 2 is the three phases schematic diagram of the electromyography signal of the embodiment of the present invention.
Fig. 3 is the dynamic electromyography signal time window processing mode of the embodiment of the present invention.
Fig. 4 is the electromyography signal pre-treatment step of the embodiment of the present invention.
Specific embodiment
The present invention will be further described with reference to the accompanying drawings and embodiments.
It is noted that described further below be all exemplary, it is intended to provide further instruction to the application.Unless another
It indicates, all technical and scientific terms used herein has usual with the application person of an ordinary skill in the technical field
The identical meanings of understanding.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singular
Also it is intended to include plural form, additionally, it should be understood that, when in the present specification using term "comprising" and/or " packet
Include " when, indicate existing characteristics, step, operation, device, component and/or their combination.
As shown in Figure 1, present embodiments providing a kind of motion intention coding/decoding method based on dynamic surface electromyography signal, wrap
Include following steps:
Step S1: different training set structural schemes are designed；
Step S2: original signal is pre-processed；
Step S3: training classifier；
Step S4: the dynamic electromyography signal for being used to testing classification device is chosen；
Step S5: testing trained classifier and selects optimal characteristics and training set structural scheme；
Step S6: online to be intended to decoding.
The invention is intended to solve motion intention decoding problem.By selecting suitable classifier, training set and feature, instruction
Practice and be suitble to carry out dynamic electromyography signal to be intended to decoded classifier, realizes in myoelectric control system reference time delay to dynamic flesh
Electric signal is intended to decoded purpose.
In the present embodiment, in step S1, the training set includes three classes, respectively pure dynamic electromyography signal, pure stable state
The binding signal of electromyography signal and stable state electromyography signal and dynamic electromyography signal.Wherein, the three phases of electromyography signal are as schemed
Shown in 2.
In the present embodiment, for pure dynamic electromyography signal, since electromyography signal is generated in actual motion in advance, before selection
The dynamic electromyography signal of 300ms can react motion intention, therefore the dynamic electromyography signal chosen in the present embodiment is less than or equal to
300ms.Specific Choice are as follows: choose 100ms before dynamic electromyography signal, preceding 150ms, preceding 200ms, preceding 250ms, preceding respectively
300ms constitutes different training sets.
In the present embodiment, for pure stable state electromyography signal, the preceding 100ms, preceding of electromyography signal steady-state portion is chosen respectively
200ms, preceding 300ms constitute different training sets.
In the present embodiment, the combination of dynamic electromyography signal and stable state electromyography signal.Above-mentioned two classes training set is constructed into shape
Structural scheme in formula combines two-by-two, generates 15 kinds of training set structural schemes, as shown in the table:
In the present embodiment, as shown in figure 4, step S2 specifically includes the following steps:
Step S21: training dataset is subjected to windowing process；
Step S22: calculating the myoelectricity feature in each time window, compares different characteristic to the decoded shadow of dynamic electromyography signal
It rings, and carries out feature extraction；
Step S23: the input sample for meeting classifier requirement is generated.By the flesh by time window segmentation and feature extraction
Electric signal generates the input sample for meeting classifier input form.
In the present embodiment, step S21 specifically: dynamic electromyography signal is split with time window, passes through overlapping window
Form original signal is configured to multiple samples；Number of windows after segmentation is calculated using following formula:
In the present embodiment, in step S22, the feature that electromyography signal is used in carrying out intention decoding includes time domain spy
Sign, frequency domain character and time and frequency domain characteristics.Such as choosing temporal signatures includes: wavelength, variance, root mean square and absolute average, institute
It is as follows to state feature calculation formula:
Wavelength:
Variance:
Root mean square:
Absolute average:
Wherein, N is the number of sampled point in a time window, XiFor the sampled point in time window.
In the present embodiment, step S3 specifically: the dynamic part data volume of electromyography signal is fewer, in order to realize with few
The purpose of data training classifier is measured, classifier uses convolutional neural networks transfer learning method, with by step S1 and step
The processed different training set off-line training convolutional neural networks of S2, obtain trained classifier.
In the present embodiment, in step S4, the electromyography signal for being used to testing classification device is chosen.Dynamic electromyography signal is chosen to make
For the test set of testing classification device, for example, can choose the preceding 100ms of dynamic electromyography signal, preceding 150ms, preceding 200ms, it is preceding
250ms or preceding 300ms.
In the present embodiment, step S5 specifically: using the preprocess method of step S2, be used to survey by what step S4 chose
The dynamic electromyography signal of examination is configured to the input sample for testing classification device, then each trained classification of off-line test
Device chooses the decoding highest classifier of accuracy as step S6 and is intended to decoding classifier used online, for this point of training
The training set structural scheme of class device is optimal case, and corresponding characteristic quantity is the feature of optimal dynamic electromyography signal.
In the present embodiment, step S6 specifically: since dynamic electromyography signal according in step S2 time window divide
Scheme is mobile and calculates characteristic value, inputs trained classifier after forming input picture, obtains real-time decoding result.
In the present embodiment, step S5, the decoding result in step S6 selects the mode of most ballots to be assessed, and decodes
Accuracy is the ratio of the test set number being correctly decoded and whole test set numbers.
Secondly the present embodiment carries out the electromyography signal of selection firstly the need of the electromyography signal chosen for constructing training set
Pretreatment generates and can be used for inputting the input sample collection of classifier, and with the training set off-line training classifier constructed, then
The electromyography signal for being used for off-line test classifier is chosen, selects optimum training collection according to classifier decoding accuracy rate, and best
Feature, the optimal classification device that off-line test is gone out tie the resulting identification of online recognition as classifier used in online recognition
Control signal of the fruit as man-machine interface, realizes the real-time control of myoelectric control system.
The present embodiment proposes a kind of motion intention coding/decoding method based on dynamic surface electromyography signal, it is intended that solves movement meaning
Figure decoding problem.By selecting suitable classifier, training set and feature, training is suitble to be intended to dynamic electromyography signal
Decoded classifier is realized and is intended to decoded purpose to dynamic electromyography signal in myoelectric control system reference time delay.This implementation
The advantage of example essentially consists in: the dynamic process of electromyography signal is the static transition stage for arriving steady-state process, and electromyography signal is used for
When control system, in order to improve real-time, need to carry out intention assessment, the electromyography signal delivered since dynamic electromyography signal
Intention coding/decoding method both for stable state electromyography signal, have ignored the static transient process to stable state and generate delay, can not
Meet requirement of the electromyography signal control system to real-time.The intention solution based on dynamic electromyography signal may be implemented in the present embodiment
Code has preferably performance in terms of reducing delay.The present embodiment can be used in human-computer interactive control system, pass through dynamic surface
Surface electromyogram signal decoding moving is intended to, and dynamic electromyography signal decoding result is used to subtract as the control signal of man-machine interactive system
Small intention assessment delay, realizes the real-time control of the man-machine interactive system based on electromyography signal.
It should be understood by those skilled in the art that, embodiments herein can provide as method, system or computer program
Product.Therefore, complete hardware embodiment, complete software embodiment or reality combining software and hardware aspects can be used in the application
Apply the form of example.Moreover, it wherein includes the computer of computer usable program code that the application, which can be used in one or more,
The computer program implemented in usable storage medium (including but not limited to magnetic disk storage, CD-ROM, optical memory etc.) produces
The form of product.
The application is referring to method, the process of equipment (system) and computer program product according to the embodiment of the present application
Figure and/or block diagram describe.It should be understood that every one stream in flowchart and/or the block diagram can be realized by computer program instructions
The combination of process and/or box in journey and/or box and flowchart and/or the block diagram.It can provide these computer programs
Instruct the processor of general purpose computer, special purpose computer, Embedded Processor or other programmable data processing devices to produce
A raw machine, so that being generated by the instruction that computer or the processor of other programmable data processing devices execute for real
The device for the function of being specified in present one or more flows of the flowchart and/or one or more blocks of the block diagram.
These computer program instructions, which may also be stored in, is able to guide computer or other programmable data processing devices with spy
Determine in the computer-readable memory that mode works, so that it includes referring to that instruction stored in the computer readable memory, which generates,
Enable the manufacture of device, the command device realize in one box of one or more flows of the flowchart and/or block diagram or
The function of being specified in multiple boxes.
These computer program instructions also can be loaded onto a computer or other programmable data processing device, so that counting
Series of operation steps are executed on calculation machine or other programmable devices to generate computer implemented processing, thus in computer or
The instruction executed on other programmable devices is provided for realizing in one or more flows of the flowchart and/or block diagram one
The step of function of being specified in a box or multiple boxes.
The above described is only a preferred embodiment of the present invention, being not that the invention has other forms of limitations, appoint
What those skilled in the art changed or be modified as possibly also with the technology contents of the disclosure above equivalent variations etc.
Imitate embodiment.But without departing from the technical solutions of the present invention, according to the technical essence of the invention to above embodiments institute
Any simple modification, equivalent variations and the remodeling made, still fall within the protection scope of technical solution of the present invention.
Claims (8)
1. a kind of motion intention coding/decoding method based on dynamic surface electromyography signal, it is characterised in that: the following steps are included:
Step S1: different training set structural schemes are designed；
Step S2: original signal is pre-processed；
Step S3: training classifier；
Step S4: the dynamic electromyography signal for being used to testing classification device is chosen；
Step S5: testing trained classifier and selects optimal characteristics and training set structural scheme；
Step S6: online to be intended to decoding.
2. a kind of motion intention coding/decoding method based on dynamic surface electromyography signal according to claim 1, feature exist
In: in step S1, the training set includes three classes, respectively pure dynamic electromyography signal, pure stable state electromyography signal and stable state flesh
The binding signal of electric signal and dynamic electromyography signal.
3. a kind of motion intention coding/decoding method based on dynamic surface electromyography signal according to claim 1, feature exist
In: step S2 specifically includes the following steps:
Step S21: training dataset is subjected to windowing process；
Step S22: calculating the myoelectricity feature in each time window, compares different characteristic to the decoded influence of dynamic electromyography signal,
And carry out feature extraction；
Step S23: the input sample for meeting classifier requirement is generated.
4. a kind of motion intention coding/decoding method based on dynamic surface electromyography signal according to claim 3, feature exist
In: step S21 specifically: dynamic electromyography signal is split with time window, by original signal structure by way of overlapping window
Cause multiple samples；Number of windows after segmentation is calculated using following formula:
5. a kind of motion intention coding/decoding method based on dynamic surface electromyography signal according to claim 3, feature exist
In: in step S22, the feature that electromyography signal is used in carrying out intention decoding includes temporal signatures, frequency domain character and time-frequency domain
Feature.
6. a kind of motion intention coding/decoding method based on dynamic surface electromyography signal according to claim 1, feature exist
In: step S3 specifically: classifier uses convolutional neural networks transfer learning method, is processed with by step S1 and step S2
Different training set off-line training convolutional neural networks, obtain trained classifier.
7. a kind of motion intention coding/decoding method based on dynamic surface electromyography signal according to claim 1, feature exist
In: step S5 specifically: using the preprocess method of step S2, the dynamic electromyography signal structure for being used to test that step S4 is chosen
The input sample for testing classification device is caused, then each trained classifier of off-line test, chooses decoding accuracy most
High classifier is intended to decoding classifier used as step S6 online, for training the training set structural scheme of the classifier
As optimal case, corresponding characteristic quantity are the feature of optimal dynamic electromyography signal.
8. a kind of motion intention coding/decoding method based on dynamic surface electromyography signal according to claim 1, feature exist
In step S6 specifically: it is mobile according to time window splitting scheme since the dynamic electromyography signal and calculate characteristic value, form input
Trained classifier is inputted after picture, obtains real-time decoding result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910084116.0A CN109814716B (en) | 2019-01-29 | 2019-01-29 | Movement intention decoding method based on dynamic surface electromyographic signals |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910084116.0A CN109814716B (en) | 2019-01-29 | 2019-01-29 | Movement intention decoding method based on dynamic surface electromyographic signals |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109814716A true CN109814716A (en) | 2019-05-28 |
| CN109814716B CN109814716B (en) | 2021-07-27 |
Family
ID=66605546
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910084116.0A Active CN109814716B (en) | 2019-01-29 | 2019-01-29 | Movement intention decoding method based on dynamic surface electromyographic signals |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109814716B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113139578A (en) * | 2021-03-23 | 2021-07-20 | 广东省科学院智能制造研究所 | Deep learning image classification method and system based on optimal training set |
| CN114169375A (en) * | 2021-12-11 | 2022-03-11 | 福州大学 | Myoelectric gesture recognition method based on strength-independent robust features |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014194609A1 (en) * | 2013-06-04 | 2014-12-11 | 南京神桥医疗器械有限公司 | Control method based on electromyographic signal and sensor signal for implementing fine real-time motion |
| CN106420124A (en) * | 2016-10-10 | 2017-02-22 | 上海电机学院 | Method for myoelectrically controlling artificial hand simulation system of virtual robot |
| CN106693178A (en) * | 2017-02-17 | 2017-05-24 | 福州大学 | Voluntary will-based functional electrical stimulation closed-loop control method for upper limb rehabilitation |
| CN108108654A (en) * | 2017-09-19 | 2018-06-01 | 东华大学 | A kind of Freehandhand-drawing track reconstructing method based on multi-channel surface myoelectric signal |
| CN108491077A (en) * | 2018-03-19 | 2018-09-04 | 浙江大学 | A kind of surface electromyogram signal gesture identification method for convolutional neural networks of being divided and ruled based on multithread |
| CN109199786A (en) * | 2018-07-26 | 2019-01-15 | 北京机械设备研究所 | A kind of lower limb rehabilitation robot based on two-way neural interface |
-
2019
- 2019-01-29 CN CN201910084116.0A patent/CN109814716B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2014194609A1 (en) * | 2013-06-04 | 2014-12-11 | 南京神桥医疗器械有限公司 | Control method based on electromyographic signal and sensor signal for implementing fine real-time motion |
| CN106420124A (en) * | 2016-10-10 | 2017-02-22 | 上海电机学院 | Method for myoelectrically controlling artificial hand simulation system of virtual robot |
| CN106693178A (en) * | 2017-02-17 | 2017-05-24 | 福州大学 | Voluntary will-based functional electrical stimulation closed-loop control method for upper limb rehabilitation |
| CN108108654A (en) * | 2017-09-19 | 2018-06-01 | 东华大学 | A kind of Freehandhand-drawing track reconstructing method based on multi-channel surface myoelectric signal |
| CN108491077A (en) * | 2018-03-19 | 2018-09-04 | 浙江大学 | A kind of surface electromyogram signal gesture identification method for convolutional neural networks of being divided and ruled based on multithread |
| CN109199786A (en) * | 2018-07-26 | 2019-01-15 | 北京机械设备研究所 | A kind of lower limb rehabilitation robot based on two-way neural interface |
Non-Patent Citations (1)
| Title |
|---|
| G. KANITZ 等: "Classification of Transient Myoelectric Signals for the Control of Multi-Grasp Hand Prostheses", 《IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113139578A (en) * | 2021-03-23 | 2021-07-20 | 广东省科学院智能制造研究所 | Deep learning image classification method and system based on optimal training set |
| CN114169375A (en) * | 2021-12-11 | 2022-03-11 | 福州大学 | Myoelectric gesture recognition method based on strength-independent robust features |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109814716B (en) | 2021-07-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lyon | Machine hearing: An emerging field [exploratory dsp] | |
| CN110659573B (en) | Face recognition method and device, electronic equipment and storage medium | |
| CN108986835B (en) | Based on speech de-noising method, apparatus, equipment and the medium for improving GAN network | |
| Avila et al. | Feature pooling of modulation spectrum features for improved speech emotion recognition in the wild | |
| Padi et al. | Improved speech emotion recognition using transfer learning and spectrogram augmentation | |
| CN108491808B (en) | Method and device for acquiring information | |
| CN110163181A (en) | Sign Language Recognition Method and device | |
| WO2023050650A1 (en) | Animation video generation method and apparatus, and device and storage medium | |
| CN109814716A (en) | A kind of motion intention coding/decoding method based on dynamic surface electromyography signal | |
| Zhou et al. | ICRC-HIT: A deep learning based comment sequence labeling system for answer selection challenge | |
| CN110443113A (en) | A kind of virtual reality Writing method, system and storage medium | |
| CN106227792A (en) | Method and apparatus for pushed information | |
| CN108345873A (en) | A kind of multiple degrees of freedom body motion information analytic method based on multilayer convolutional neural networks | |
| CN108831450A (en) | A kind of virtual robot man-machine interaction method based on user emotion identification | |
| CN113555032A (en) | Multi-speaker scene recognition and network training method and device | |
| CN111128240B (en) | Voice emotion recognition method based on anti-semantic-erasure | |
| CN110232928B (en) | Text-independent speaker verification method and device | |
| CN110032987A (en) | A kind of surface electromyogram signal classification method based on CMAC Neural Network model | |
| CN114141237A (en) | Speech recognition method, speech recognition device, computer equipment and storage medium | |
| CN103839047A (en) | Human motion track recognition method and device | |
| CN111062022B (en) | Slider verification method and device based on disturbance visual feedback and electronic equipment | |
| Jin et al. | Speech separation and emotion recognition for multi-speaker scenarios | |
| CN112634413A (en) | Method, apparatus, device and storage medium for generating model and generating 3D animation | |
| Iqbal et al. | Deepfake Audio Detection Via Feature Engineering And Machine Learning. | |
| CN110688470B (en) | Method and apparatus for transmitting information |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |