CN109787927A - Modulation Identification method and apparatus based on deep learning - Google Patents
Modulation Identification method and apparatus based on deep learning Download PDFInfo
- Publication number
- CN109787927A CN109787927A CN201910003499.4A CN201910003499A CN109787927A CN 109787927 A CN109787927 A CN 109787927A CN 201910003499 A CN201910003499 A CN 201910003499A CN 109787927 A CN109787927 A CN 109787927A
- Authority
- CN
- China
- Prior art keywords
- model
- signal
- frequency characteristics
- processed
- layer
- 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.)
- Withdrawn
Links
Landscapes
- Mobile Radio Communication Systems (AREA)
Abstract
The Modulation Identification method and apparatus based on deep learning that this application involves a kind of, which comprises the filter that signal to be processed inputs in pretreated model, with pretreated model is subjected to convolution algorithm;Sampling and normalized are carried out to the result of convolution algorithm, obtain time-frequency characteristics;Time-frequency characteristics are handled, determine the modulation type of signal to be processed.Deep learning algorithm is introduced the treatment process of signal of communication by this method, pretreated model is constructed using deep learning algorithm, to extract the time-frequency characteristics of signal of communication, recognition efficiency is high, and the type of manageable modulating mode can be expanded by autonomous learning;The ability that this method makes communication equipment or machine have autonomous learning, independently update, so that preferably reply mobile communications network develops brought problem and challenge.
Description
Technical field
This application involves signal of communication processing technology fields, and in particular to a kind of Modulation Identification method based on deep learning
And device.
Background technique
5th third-generation mobile communication technology of the continuous promotion with user to mobile communication demand, higher speed more wideband is met the tendency of
And it gives birth to.With the development of the 5th third-generation mobile communication technology, the variation of communication environment is more complicated, in order to improve the utilization rate of frequency band
And guarantee transmission reliability, it needs using a variety of different modulation systems.The purpose of Modulation Identification is exactly, be in more modulation
It can be to the modulating mode of the signal of communication received under the background of signal simultaneous transmission and under the environment of priori conditions deficiency
It is correctly identified, provides foundation for next analysis signal, processing signal.
In signal of communication processing, feature extraction is a vital step.Traditional feature extraction algorithm is based on artificial
Analysis extracts cyclic cumulants, the higher-order spectrum of signal in conjunction with the methods of statistics by determining data transformation for mula and method
Feature etc. such as obtains the time-frequency characteristics of signal dependent on STFT (Short Time Fourier Transform), then therefrom statistics obtains the height of signal
Rank counts measure feature etc..
In the related technology, identification technology is the frame based on expertise and predefined mathematical model, needs human intervention
Statistical analysis process after feature extraction, and extraction.Which is larger by subjective impact, varies with each individual, and problem also compares
It is more, for example recognition efficiency is low, the modulation system that can identify is limited etc..Once there is new modulating mode, then it is original
Recognition methods would generally fail.
As it can be seen that depending on complex man's work point in the case where frequency spectrum resource efficient multiplexing demand, communication environment variation are complicated
Analysis extract feature conventional communication signals identification technology often have greatly it is limiting, can no longer meet actual use need
It asks.
Summary of the invention
To be overcome the problems, such as present in the relevant technologies at least to a certain extent, the application provides a kind of based on deep learning
Modulation Identification method and apparatus.
According to the embodiment of the present application in a first aspect, providing a kind of Modulation Identification method based on deep learning, comprising:
The filter that signal to be processed inputs in pretreated model, with pretreated model is subjected to convolution algorithm;
Sampling and normalized are carried out to the result of convolution algorithm, obtain time-frequency characteristics;
Time-frequency characteristics are handled, determine the modulation type of signal to be processed.
Further, the pretreated model be Boltzmann machine is limited by convolution trained in advance, including input layer,
Hidden layer and output layer;
The input layer includes two channels, is respectively used to input the real and imaginary parts of signal to be processed.
Further, the training method of the pretreated model includes:
Input training sample and learning rate;
Initialization model parameter;
Successively each of training sample data are sent into model and are iterated operation, and mould is updated according to learning rate
Shape parameter.
Further, the model parameter includes:
Filter parameter, the amount of bias of input layer, the amount of bias of hidden layer.
Further, the result to convolution algorithm carries out sampling and normalized, comprising:
The length of filter is obtained from filter parameter;
Sampling step length is determined according to the length of filter;
It is sampled according to result of the sampling step length to convolution algorithm;
Result after sampling is substituted into preset standardization formula, is normalized.
It is further, described that time-frequency characteristics are handled, comprising:
Time-frequency characteristics input feature vector is extracted into model and carries out operation, obtains characteristic parameter;
Characteristic parameter input disaggregated model is handled, classification results are obtained.
Further, the modulation type for determining signal to be processed, comprising:
The modulation type of signal to be processed is determined according to classification results.
Further, the Feature Selection Model is that convolution is limited Boltzmann machine, including input layer, hidden layer and output
Layer;
The disaggregated model is back propagation artificial neural network model.
Further, the Feature Selection Model is identical as the structure of the pretreated model, and model parameter is different.
According to the second aspect of the embodiment of the present application, a kind of Modulation Identification device based on deep learning is provided, comprising:
Preprocessing module, the filter for inputting signal to be processed in pretreated model, with pretreated model carry out
Convolution algorithm;
Sampling module carries out sampling and normalized for the result to convolution algorithm, obtains time-frequency characteristics;
Discrimination module determines the modulation type of signal to be processed for handling time-frequency characteristics.
The technical solution that embodiments herein provides can include the following benefits:
Deep learning algorithm is introduced the treatment process of signal of communication by this method, and pre- place is constructed using deep learning algorithm
Model is managed, to extract the time-frequency characteristics of signal of communication, recognition efficiency is high, and can expand and can locate by autonomous learning
The type of the modulating mode of reason.The ability that this method makes communication equipment or machine have autonomous learning, independently update, thus more preferably
It copes with mobile communications network and develops brought problem and challenge in ground.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
The application can be limited.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the application
Example, and together with specification it is used to explain the principle of the application.
Fig. 1 is a kind of flow chart of Modulation Identification method based on deep learning shown according to an exemplary embodiment.
Fig. 2 is that single layer convolution is limited Boltzmann machine structural schematic diagram.
Fig. 3 is mapping relations schematic diagram of the CRBM network visible layer to hidden layer.
Fig. 4 is a kind of circuit block of Modulation Identification device based on deep learning shown according to an exemplary embodiment
Figure.
Specific embodiment
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistent with the application.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects be described in detail in claims, the application.
Fig. 1 is a kind of flow chart of Modulation Identification method based on deep learning shown according to an exemplary embodiment.
Method includes the following steps:
Step 101: the filter that signal to be processed inputs in pretreated model, with pretreated model is subjected to convolution fortune
It calculates;
Step 102: sampling and normalized being carried out to the result of convolution algorithm, obtain time-frequency characteristics;
Step 103: time-frequency characteristics being handled, determine the modulation type of signal to be processed.
Deep learning algorithm is introduced the treatment process of signal of communication by this method, and pre- place is constructed using deep learning algorithm
Model is managed, to extract the time-frequency characteristics of signal of communication, recognition efficiency is high, and can expand and can locate by autonomous learning
The type of the modulating mode of reason.The ability that this method makes communication equipment or machine have autonomous learning, independently update, thus more preferably
It copes with mobile communications network and develops brought problem and challenge in ground.
In some embodiments, the pretreated model is the limited Boltzmann machine of process convolution trained in advance, including defeated
Enter layer, hidden layer and output layer;
The input layer includes two channels, is respectively used to input the real and imaginary parts of signal to be processed.
Referring to Fig. 2, convolution is limited Boltzmann machine (convolutional restricted Boltzmann
Machine, CRBM) it is a kind of extension on the basis of limited Boltzmann machine (RBM), the convolution in convolutional neural networks is grasped
It applies in RBM and just generates a kind of new model --- CRBM.Single layer CRBM structure is input layer respectively by up of three-layer
(visible layer) V, hidden layer H and output layer (pond layer) P.
Referring to Fig. 3, the detailed process that CRBM network is mapped to hidden layer from input layer is embodied, example is filter in figure
Size is 3 × 3, and the filter for having K group different carries out convolution with the data of input layer respectively, finally obtains K group hidden layer.Separately
Outside, for neural network, it is also necessary to which amount of bias is set, and a critically important characteristic of CRBM network is that biasing is shared, i.e., defeated
Enter the shared biasing c of layer, every layer of hidden layer shares a biasing bk, the parameter of trained network is greatly reduced, can be improved
The speed of network training.The last layer is pond layer, and common pondization operation includes maximum value pond, mean value pond etc., Chi Hua
Refer to and certain characteristic statistics carried out to the region of specified size, such as mean value pond is exactly to take mean value in a certain zonule the most
Output.Pond layer can reduce trained parameter, and can also prevent the appearance of over-fitting.
For embodiments herein, it is by an input layer that pretreated model essence, which is exactly a CRBM structure,
(or visible layer), a hidden layer and an output layer composition.Input layer is made of the real value unit array of Nc × Nv, wherein Nc
Represent be signal port number, the Nc=2 in this model, input be respectively modulated signal real and imaginary parts, Nv is letter
Number sampling quantity.Hidden layer is made of " Ng group ", wherein each group be a 1 × Nh real value unit array, cause to export
Corresponding Ng × Nh the binary unit of layer, wherein binary value indicates the state of activation of each unit in hidden layer.It is each implicit
The group of layer is associated with Nc × Nw filter, and filter weight is shared between all positions of signal in organizing.It is noticeable
It is relationship between parameter is Nh=Nv-Nw+1.
The filter that input signal is 1 with step-length in convolutional layer carries out convolution, but not all time point is all
It needs.It notices in STFT, sliding window is usually laminated in the 1/3 of length of window, and 1/2,2/3, this facilitates in short-term
Interior acquisition information, and ensure that no information is lost.Similarly, after convolutional layer, sample level is introduced in what we constructed
In network.Data volume can be greatly reduced in this way, to reduce the calculation amount and training complexity in subsequent network.In this model
In, the step-length stride that we set down-sampling is related with the length of filter, stride=Nw/2.
If for the average value of each input variable close to zero, covariance is approximately equal (such as 1) on training set, then speed is restrained
Degree is usually faster.Heuristic it shall apply to all levels in addition, this.If input data by sigmoid function before being activated
It is not normalized, then the value after activating will enter flat site, and output layer will restrain and lead to important reconstructed error.For
This, input data and output data with zero-mean and identical covariance are arranged using normalization layer (more precisely, assisting
1) variance both is set to.Referred to as the standardized method of " Z score " can indicate are as follows:
Wherein x is the data of input, and μ is the mean value of input data, and δ is the variance of input data.
In some embodiments, the training method of the pretreated model includes:
Input training sample and learning rate;
Initialization model parameter;
Successively each of training sample data are sent into model and are iterated operation, and mould is updated according to learning rate
Shape parameter.
In some embodiments, the model parameter includes:
Filter parameter, the amount of bias of input layer, the amount of bias of hidden layer.
Wherein, filter parameter includes filter length Nw, the amount of bias of input layer include input layer shared one partially
C is set, the amount of bias of hidden layer includes the biasing b of every layer of hidden layerk, initialization is using random initializtion.
In some embodiments, the result to convolution algorithm carries out sampling and normalized, comprising:
The length of filter is obtained from filter parameter;
Sampling step length is determined according to the length of filter;
It is sampled according to result of the sampling step length to convolution algorithm;
Result after sampling is substituted into preset standardization formula, is normalized.
It is described that time-frequency characteristics are handled in some embodiments, comprising:
Time-frequency characteristics input feature vector is extracted into model and carries out operation, obtains characteristic parameter;
Characteristic parameter input disaggregated model is handled, classification results are obtained.
In some embodiments, the modulation type for determining signal to be processed, comprising:
The modulation type of signal to be processed is determined according to classification results.
In some embodiments, the Feature Selection Model be convolution be limited Boltzmann machine, including input layer, hidden layer and
Output layer;
The disaggregated model is back propagation artificial neural network model.
In some embodiments, the Feature Selection Model is identical as the structure of the pretreated model, and model parameter is different.
Fig. 4 is a kind of circuit block of Modulation Identification device based on deep learning shown according to an exemplary embodiment
Figure.The device includes:
Preprocessing module 401, for by signal to be processed input pretreated model, with pretreated model in filter into
Row convolution algorithm;
Sampling module 402 carries out sampling and normalized for the result to convolution algorithm, obtains time-frequency characteristics;
Discrimination module 403 determines the modulation type of signal to be processed for handling time-frequency characteristics.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, no detailed explanation will be given here.
It is understood that same or similar part can mutually refer in the various embodiments described above, in some embodiments
Unspecified content may refer to the same or similar content in other embodiments.
It should be noted that term " first ", " second " etc. are used for description purposes only in the description of the present application, without
It can be interpreted as indication or suggestion relative importance.In addition, in the description of the present application, unless otherwise indicated, the meaning of " multiple "
Refer at least two.
Any process described otherwise above or method description are construed as in flow chart or herein, and expression includes
It is one or more for realizing specific logical function or process the step of executable instruction code module, segment or portion
Point, and the range of the preferred embodiment of the application includes other realization, wherein can not press shown or discussed suitable
Sequence, including according to related function by it is basic simultaneously in the way of or in the opposite order, Lai Zhihang function, this should be by the application
Embodiment person of ordinary skill in the field understood.
It should be appreciated that each section of the application can be realized with hardware, software, firmware or their combination.Above-mentioned
In embodiment, software that multiple steps or method can be executed in memory and by suitable instruction execution system with storage
Or firmware is realized.It, and in another embodiment, can be under well known in the art for example, if realized with hardware
Any one of column technology or their combination are realized: having a logic gates for realizing logic function to data-signal
Discrete logic, with suitable combinational logic gate circuit specific integrated circuit, programmable gate array (PGA), scene
Programmable gate array (FPGA) etc..
Those skilled in the art are understood that realize all or part of step that above-described embodiment method carries
It suddenly is that relevant hardware can be instructed to complete by program, the program can store in a kind of computer-readable storage medium
In matter, which when being executed, includes the steps that one or a combination set of embodiment of the method.
It, can also be in addition, can integrate in a processing module in each functional unit in each embodiment of the application
It is that each unit physically exists alone, can also be integrated in two or more units in a module.Above-mentioned integrated mould
Block both can take the form of hardware realization, can also be realized in the form of software function module.The integrated module is such as
Fruit is realized and when sold or used as an independent product in the form of software function module, also can store in a computer
In read/write memory medium.
Storage medium mentioned above can be read-only memory, disk or CD etc..
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is contained at least one embodiment or example of the application.In the present specification, schematic expression of the above terms are not
Centainly refer to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be any
One or more embodiment or examples in can be combined in any suitable manner.
Although embodiments herein has been shown and described above, it is to be understood that above-described embodiment is example
Property, it should not be understood as the limitation to the application, those skilled in the art within the scope of application can be to above-mentioned
Embodiment is changed, modifies, replacement and variant.
Claims (10)
1. a kind of Modulation Identification method based on deep learning characterized by comprising
The filter that signal to be processed inputs in pretreated model, with pretreated model is subjected to convolution algorithm;
Sampling and normalized are carried out to the result of convolution algorithm, obtain time-frequency characteristics;
Time-frequency characteristics are handled, determine the modulation type of signal to be processed.
2. according to the method described in claim 1, it is characterized by: the pretreated model be by convolution trained in advance by
Limit Boltzmann machine, including input layer, hidden layer and output layer;
The input layer includes two channels, is respectively used to input the real and imaginary parts of signal to be processed.
3. according to the method described in claim 2, it is characterized in that, the training method of the pretreated model includes:
Input training sample and learning rate;
Initialization model parameter;
Successively each of training sample data are sent into model and are iterated operation, and are joined according to learning rate more new model
Number.
4. according to the method described in claim 3, it is characterized in that, the model parameter includes:
Filter parameter, the amount of bias of input layer, the amount of bias of hidden layer.
5. according to the method described in claim 4, it is characterized in that, the result to convolution algorithm is sampled and is normalized
Processing, comprising:
The length of filter is obtained from filter parameter;
Sampling step length is determined according to the length of filter;
It is sampled according to result of the sampling step length to convolution algorithm;
Result after sampling is substituted into preset standardization formula, is normalized.
6. method according to claim 1-5, which is characterized in that described to handle time-frequency characteristics, comprising:
Time-frequency characteristics input feature vector is extracted into model and carries out operation, obtains characteristic parameter;
Characteristic parameter input disaggregated model is handled, classification results are obtained.
7. according to the method described in claim 6, it is characterized in that, the modulation type for determining signal to be processed, comprising:
The modulation type of signal to be processed is determined according to classification results.
8. according to the method described in claim 6, it is characterized by: the Feature Selection Model is that convolution is limited Boltzmann
Machine, including input layer, hidden layer and output layer;
The disaggregated model is back propagation artificial neural network model.
9. according to the method described in claim 8, it is characterized by: the knot of the Feature Selection Model and the pretreated model
Structure is identical, and model parameter is different.
10. a kind of Modulation Identification device based on deep learning characterized by comprising
Preprocessing module, the filter for inputting signal to be processed in pretreated model, with pretreated model carry out convolution
Operation;
Sampling module carries out sampling and normalized for the result to convolution algorithm, obtains time-frequency characteristics;
Discrimination module determines the modulation type of signal to be processed for handling time-frequency characteristics.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910003499.4A CN109787927A (en) | 2019-01-03 | 2019-01-03 | Modulation Identification method and apparatus based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910003499.4A CN109787927A (en) | 2019-01-03 | 2019-01-03 | Modulation Identification method and apparatus based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109787927A true CN109787927A (en) | 2019-05-21 |
Family
ID=66499888
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910003499.4A Withdrawn CN109787927A (en) | 2019-01-03 | 2019-01-03 | Modulation Identification method and apparatus based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109787927A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111988252A (en) * | 2020-08-24 | 2020-11-24 | 成都华日通讯技术股份有限公司 | Signal modulation mode identification method based on deep learning |
CN113869227A (en) * | 2021-09-29 | 2021-12-31 | 西南交通大学 | Signal modulation mode identification method, device, equipment and readable storage medium |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107634923A (en) * | 2017-09-21 | 2018-01-26 | 佛山科学技术学院 | A kind of distributed communication signal modulate method |
CN107979554A (en) * | 2017-11-17 | 2018-05-01 | 西安电子科技大学 | Radio signal Modulation Identification method based on multiple dimensioned convolutional neural networks |
US10003483B1 (en) * | 2017-05-03 | 2018-06-19 | The United States Of America, As Represented By The Secretary Of The Navy | Biologically inspired methods and systems for automatically determining the modulation types of radio signals using stacked de-noising autoencoders |
CN108234370A (en) * | 2017-12-22 | 2018-06-29 | 西安电子科技大学 | Modulation mode of communication signal recognition methods based on convolutional neural networks |
CN108449295A (en) * | 2018-02-05 | 2018-08-24 | 西安电子科技大学昆山创新研究院 | Combined modulation recognition methods based on RBM networks and BP neural network |
CN108600137A (en) * | 2018-04-28 | 2018-09-28 | 重庆邮电大学 | A kind of novel multicarrier recognition methods based on reverse transmittance nerve network |
CN108616470A (en) * | 2018-03-26 | 2018-10-02 | 天津大学 | Modulation Signals Recognition method based on convolutional neural networks |
CN108650202A (en) * | 2018-05-11 | 2018-10-12 | 大唐联诚信息系统技术有限公司 | A kind of signal modulation mode identification method and device |
CN108718288A (en) * | 2018-03-30 | 2018-10-30 | 电子科技大学 | Recognition of digital modulation schemes method based on convolutional neural networks |
CN108768907A (en) * | 2018-01-05 | 2018-11-06 | 南京邮电大学 | A kind of Modulation Identification method based on temporal characteristics statistic and BP neural network |
-
2019
- 2019-01-03 CN CN201910003499.4A patent/CN109787927A/en not_active Withdrawn
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10003483B1 (en) * | 2017-05-03 | 2018-06-19 | The United States Of America, As Represented By The Secretary Of The Navy | Biologically inspired methods and systems for automatically determining the modulation types of radio signals using stacked de-noising autoencoders |
CN107634923A (en) * | 2017-09-21 | 2018-01-26 | 佛山科学技术学院 | A kind of distributed communication signal modulate method |
CN107979554A (en) * | 2017-11-17 | 2018-05-01 | 西安电子科技大学 | Radio signal Modulation Identification method based on multiple dimensioned convolutional neural networks |
CN108234370A (en) * | 2017-12-22 | 2018-06-29 | 西安电子科技大学 | Modulation mode of communication signal recognition methods based on convolutional neural networks |
CN108768907A (en) * | 2018-01-05 | 2018-11-06 | 南京邮电大学 | A kind of Modulation Identification method based on temporal characteristics statistic and BP neural network |
CN108449295A (en) * | 2018-02-05 | 2018-08-24 | 西安电子科技大学昆山创新研究院 | Combined modulation recognition methods based on RBM networks and BP neural network |
CN108616470A (en) * | 2018-03-26 | 2018-10-02 | 天津大学 | Modulation Signals Recognition method based on convolutional neural networks |
CN108718288A (en) * | 2018-03-30 | 2018-10-30 | 电子科技大学 | Recognition of digital modulation schemes method based on convolutional neural networks |
CN108600137A (en) * | 2018-04-28 | 2018-09-28 | 重庆邮电大学 | A kind of novel multicarrier recognition methods based on reverse transmittance nerve network |
CN108650202A (en) * | 2018-05-11 | 2018-10-12 | 大唐联诚信息系统技术有限公司 | A kind of signal modulation mode identification method and device |
Non-Patent Citations (2)
Title |
---|
周东青: ""基于深度限制波尔兹曼机的辐射源信号识别"", 《国防科技大学学报》 * |
周龙梅: ""基于深度学习的通信信号识别技术研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111988252A (en) * | 2020-08-24 | 2020-11-24 | 成都华日通讯技术股份有限公司 | Signal modulation mode identification method based on deep learning |
CN113869227A (en) * | 2021-09-29 | 2021-12-31 | 西南交通大学 | Signal modulation mode identification method, device, equipment and readable storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107203999A (en) | A kind of skin lens image automatic division method based on full convolutional neural networks | |
CN112041856A (en) | Cross-modal neural network for prediction | |
CN106295591A (en) | Gender identification method based on facial image and device | |
CN109829478B (en) | Problem classification method and device based on variation self-encoder | |
CN106326857A (en) | Gender identification method and gender identification device based on face image | |
CN106951753A (en) | The authentication method and authentication device of a kind of electrocardiosignal | |
CN111785366B (en) | Patient treatment scheme determination method and device and computer equipment | |
CN107749757A (en) | A kind of data compression method and device based on stacking-type own coding and PSO algorithms | |
CN109784312A (en) | Teaching Management Method and device | |
CN112163637B (en) | Image classification model training method and device based on unbalanced data | |
CN111582396A (en) | Fault diagnosis method based on improved convolutional neural network | |
CN109787927A (en) | Modulation Identification method and apparatus based on deep learning | |
CN109787929A (en) | Signal modulate method, electronic device and computer readable storage medium | |
CN115985513B (en) | Data processing method, device and equipment based on multiple groups of chemical cancer typing | |
Kursun et al. | Flower recognition system with optimized features for deep features | |
CN111680642A (en) | Terrain classification method and device | |
WO2022236416A1 (en) | Machine learning-based surgical instrument characterization | |
CN114239657A (en) | Time sequence signal identification method based on complex value interference neural network | |
CN112801283B (en) | Neural network model, action recognition method, device and storage medium | |
CN106779062A (en) | A kind of multi-layer perception (MLP) artificial neural network based on residual error network | |
CN108122028A (en) | Training method, device and the computer readable storage medium of depth Nonlinear Principal Component Analysis network | |
CN108665001A (en) | It is a kind of based on depth confidence network across subject Idle state detection method | |
CN110490876A (en) | A kind of lightweight neural network for image segmentation | |
CN111652051B (en) | Face detection model generation method, device, equipment and storage medium | |
CN112749797B (en) | Pruning method and device for neural network model |
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 | ||
WW01 | Invention patent application withdrawn after publication | ||
WW01 | Invention patent application withdrawn after publication |
Application publication date: 20190521 |