CN110099020A - A kind of unmanned plane electromagnetic signal management and Modulation Mode Recognition method - Google Patents
A kind of unmanned plane electromagnetic signal management and Modulation Mode Recognition method Download PDFInfo
- Publication number
- CN110099020A CN110099020A CN201910432119.9A CN201910432119A CN110099020A CN 110099020 A CN110099020 A CN 110099020A CN 201910432119 A CN201910432119 A CN 201910432119A CN 110099020 A CN110099020 A CN 110099020A
- Authority
- CN
- China
- Prior art keywords
- unmanned plane
- electromagnetic signal
- modulation mode
- data
- mode recognition
- 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
Classifications
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/04—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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/0012—Modulated-carrier systems arrangements for identifying the type of modulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of unmanned plane electromagnetic signal management and Modulation Mode Recognition methods, belong to electromagnetic signal storage and identification technology field.The recognition methods obtains unmanned plane electromagnetic signal details first and stores, then Modulation Mode Recognition model is constructed, electromagnetic signal waveform and modulation system data repetitive exercise Modulation Mode Recognition model parameter are inputted, realizes the Modulation Mode Recognition to unknown unmanned plane electromagnetic signal waveform.The present invention stores unmanned plane acquired electromagnetic data and environmental data simultaneously, and data structure is versatile and has extendible ability;The Modulation Mode Recognition model promotes the characterization ability to multiclass modulation system using advanced part and temporal aspect.
Description
Technical field
The invention belongs to electromagnetic signal storages and identification technology field, in particular to one kind to be based on database and deep learning
The management of unmanned plane electromagnetic signal and Modulation Mode Recognition method.
Background technique
Unmanned systems data-link carries the vital task of unmanned systems command and control and information transmission, is unmanned systems
Important component.It is the unmanned systems of representative in battle reconnaissance, long-range strike, space flight measurement and control, empty day group using unmanned plane
It is studied and is applied in the fields such as net, emergency communication, logistics transportation, environmental monitoring.Gradually recently as unmanned systems
It is perfect, the application of unmanned systems gradually present explosive growth and to safety of the unmanned systems data-link under complex environment,
More stringent requirements are proposed for reliability and adaptability.However, existing unmanned systems data-link is in the case where being applied to strong Antagonistic Environment
Information transmission, the great demand task such as cooperation when, there is also anti-interference, anti-intercepting and capturings, anti-deception scarce capacity, complicated
The problems such as deficiency of the cognitive ability of electromagnetic environment, data storage is chaotic, needs a kind of unmanned plane electromagnetism letter highly effective and versatile
Number management and Modulation Mode Recognition method.
In unmanned plane electromagnetic signal management aspect, ignores complicated electromagnetic signal Wave data, only store corresponding signal
Feature is far from enough, and this storage mode is difficult to embody the complex information of electromagnetic signal, is unfavorable for subsequent analysis and reproduction work.
In terms of unmanned plane electromagnetic signal Modulation Mode Recognition, existing algorithm is based on Higher Order Cumulants and wavelet transformation etc. and artificially sets
Determine feature, and in complicated electromagnetic environment, the electromagnetic signal feature and theoretical value otherness of actual acquisition are larger, lack feedback
Electromagnetic signal Modulation Mode Recognition algorithm versatility it is poor, meanwhile, traditional artificial settings feature is to multiclass electromagnetic signal tune
The characterization scarce capacity of mode processed, it usually needs various features combination can cover multiclass modulation system, and such method is usual
The case where there are signal characteristic redundancies, and calculating process complexity is excessively high, real-time is difficult to ensure.
Summary of the invention
The present invention increases unmanned plane electromagnetic signal to improve the versatility and expandability of unmanned plane electromagnetic signal management
The characterization ability of feature proposes a kind of unmanned plane electromagnetic signal management and Modulation Mode Recognition method, can store nothing simultaneously
Man-machine electromagnetic signal waveform and characteristic, have versatility and expandability, and construct high performance electromagnetic signal modulation methods
Formula identification model realizes common unmanned plane electromagnetic signal Modulation Mode Recognition.
The unmanned plane electromagnetic signal management provided by the invention and Modulation Mode Recognition method, the specific steps are as follows:
The first step obtains unmanned plane electromagnetic signal details;
In order to store unmanned plane acquired electromagnetic data, Overall Acquisition unmanned plane electromagnetic signal details, including nothing are needed
Man-machine acquired electromagnetic data and environmental data.
Second step, the storage of unmanned plane electromagnetic signal details;
Unmanned plane electromagnetic signal details based on acquisition, are successively stored in MYSQL for signal data and environmental data
In database (being the application software of a Relational DBMS), and electromagnetic signal waveform and modulation system are stored
For the MAT file (being the data file in matlab) of reference numeral.
Third step constructs Modulation Mode Recognition model;
Based on unmanned plane electromagnetic signal modulation system otherness, deep learning convolutional neural networks and long short-term memory are merged
Network constructs high-performance and versatile unmanned plane electromagnetic signal Modulation Mode Recognition model.
4th step, training Modulation Mode Recognition model realization Modulation Mode Recognition;
Based on the Modulation Mode Recognition model constructed, electromagnetic signal waveform and modulation system data repetitive exercise tune are inputted
Mode identification model parameter processed realizes the Modulation Mode Recognition to unknown unmanned plane electromagnetic signal waveform.
Main advantages of the present invention are:
(1) unmanned plane acquired electromagnetic data and environmental data are stored simultaneously, data structure is versatile and has extendible
Ability;
(2) construct the unmanned plane electromagnetic signal Modulation Mode Recognition model based on deep learning, using advanced part and
Temporal aspect promotes the characterization ability to multiclass modulation system.
Detailed description of the invention
Fig. 1 is unmanned plane electromagnetic signal management of the invention and Modulation Mode Recognition method overall step flow chart;
Fig. 2 is unmanned plane acquired electromagnetic data query result schematic diagram of the invention;
Fig. 3 is Modulation Mode Recognition model structure schematic diagram of the invention;
Fig. 4 is length of the invention memory unit structure figure in short-term.
Fig. 5 is deep learning model Modulation Mode Recognition result figure of the invention.
Specific embodiment
With reference to the accompanying drawing, specific implementation method of the invention is described in detail.
The present invention is a kind of unmanned plane electromagnetic signal management and Modulation Mode Recognition method, overall flow figure as shown in Figure 1,
The storage of unmanned plane electromagnetic signal details is embodied in the difference of the prior art and establishes Modulation Mode Recognition model, that is, is realized
Unmanned plane acquired electromagnetic data and environmental data storage, and construct unmanned plane electromagnetic signal modulation system based on deep learning
Identification model.Specific implementation method the following steps are included:
The first step obtains unmanned plane electromagnetic signal details, including acquired electromagnetic data and environmental data.
Unmanned plane electromagnetic signal details are made of acquired electromagnetic data and environmental data two parts.Acquired electromagnetic data
Including signal waveform data (passing through analog-to-digital conversion), signal modulation mode (common digital modulation mode), signal working frequency points (nothing
Man-machine electromagnetic signal frequency range: 840.5-845MHz, 1430-1444MHz and 2408-2440MHz), sample frequency and signal-to-noise ratio (0-
20dB);Environmental data includes the phase between signal source latitude and longitude coordinates (retaining after decimal point three), amblent air temperature and bay
Potential difference information.
Second step, the storage of unmanned plane electromagnetic signal details.
Based on the unmanned plane electromagnetic signal details that the first step obtains, in order successively by signal data and environmental data
Be stored in MYSQL database, and by acquired electromagnetic data signal waveform data and signal modulation mode be saved as reference numeral
MAT data file.
After obtaining unmanned plane electromagnetic signal details, software interface is created on matlab, successively with number, modulation methods
Formula, working frequency points, sample frequency, signal-to-noise ratio, unmanned plane coordinate longitude, latitude, amblent air temperature type and bay phase difference
Sequence by electromagnetic signal details be stored in database (MYSQL) in, realize be based on signal-to-noise ratio, modulation system and working frequency points
For the sample searching function of condition, electromagnetic signal waveform is observed by visual control;And using form (each knot of structural body
Structure body first is classified as signal waveform data, second is classified as modulation system information), signal waveform data and modulation system information are deposited
Storage is the file (MAT form) of reference numeral, and can be inquired by signal characteristic, as shown in Figure 2.
Third step, the otherness of the multi-signal modulation system based on second step storage, constructs high-performance and versatile
Unmanned plane electromagnetic signal Modulation Mode Recognition model.
The present invention is based on the mode of convolutional neural networks and the long fusion of memory network in short-term, building deep learning is unmanned electromechanical
Magnetic signal Modulation Mode Recognition model (being hereinafter identification model), due to convolutional neural networks CNN and long short-term memory net
Network LSTM is each provided with unique structure, and to improve Modulation Mode Recognition accuracy rate, two are merged by the way of serial and concurrent
Kind special neural network, as shown in figure 3, the deep learning unmanned plane electromagnetic signal Modulation Mode Recognition model mainly by
Convolutional neural networks and long memory network in short-term are constituted, and in parallel model, untreated signal waveform data leads to simultaneously
Crossing two kinds of network mappings is the characteristic pattern for having respective network attribute, and signal modulation mode classification is superposed to after full articulamentum
Feature vector.In serial model, untreated signal waveform data first passes around convolutional neural networks and is mapped as part
After characteristic pattern, by the long memory network LSTM in short-term of this characteristic pattern input, signal modulation side is mapped as by long memory network in short-term
Formula characteristic of division vector.
Convolutional neural networks are a kind of multi-level neural network based on deep learning, traditional convolutional neural networks model
Generally comprise convolutional layer, active coating and pond layer.In the present invention, the input of convolutional neural networks model is original unmanned electromechanics
Magnetic signal Wave data, input of the output of each convolutional layer after nonlinear activation as next convolutional layer, by more
The form that a convolutional layer stacks carrys out the profound difference portrayed between Different Modulations classification.
Each convolutional layer can be made of the convolution kernel of multiple and different sizes, and the size of convolution kernel sets manually and in training
Preceding initiation parameter, convolution kernel be in each sample of signal it is shared, so each convolution kernel can be described as a spy
Levy extraction unit, connected by convolution kernel between the neuron of different layers, it is this it is unrelated with position come shared parameter mode energy
The quantity of learning parameter is enough effectively reduced, and convolution algorithm forms the similarly suitable extraction signal of communication office of form of local sensing open country
The feature of portion's variation, passes through a convolution karyogenesis characteristic results y having a size of kiProcess it is as follows:
yi=f (Wxi:i+k-1+b)
Wherein, W indicates weight coefficient, xi:i+k-1Indicate digital communication signal from the i moment to the signal waveform at i+k-1 moment
Data, b are offset parameters, and f () is nonlinear activation function, and the present invention uses ReLu (Rectified Linear Units)
As activation primitive.Due to digital modulation signals be single-channel data, information content is relatively fewer, the present invention in and unused pool
Layer carries out dimensionality reduction to data.
Long memory network in short-term is a kind of special recurrent neural network, and traditional recurrent neural network is in processing long enough
When the information of degree, study can be lost to the ability for being spaced information farther out, this gradient disappearance problem can be understood as forgetting for human brain
Thing, and long memory network in short-term then solves the problems, such as this by the interaction for designing progress neuron in a special way of structure,
Therefore long memory network in short-term possesses the ability of powerful learning and memory time series data, and long memory network cellular construction in short-term is as schemed
Shown in 4.
Long memory network in short-term controls the intensity that information reaches neuron by door, and door can be realized information
Selecting type pass through, wherein being operated comprising sigmod neural net layer and multiplication.Sigmod layers of output determines input information
Output par, c, 0 representative cannot pass through completely, and 1 represents without hindrance pass through.The first step of long memory network in short-term is decision from nerve
The information abandoned in member, this functionality are realized by forgetting door.The door can read the input x of current t momenttAnd t-1
The hidden layer at moment exports ht-1, the multiple of numerical value between one 0 to 1 is generated as the recall info in neuron.To unmanned plane
When electromagnetic signal modulation system is identified, forgeing door can be regarded as the data information currently remembered in neuron, and work as length
When memory network processing generate to digital modulation information special variation when, it is intended that neural network forget before information.
ft=sigmod (Wf·[ht-1,xt]+bf)
Wherein, ftIt indicates to forget door, WfIt indicates to forget door weight parameter, ht-1Indicate state output, xtIndicate t moment length
When memory network input data, bfIt indicates to forget door amount of bias.
Then the new information in neuron is determined, wherein (sigmod layers) of the input gate updated value determined in neuron, and
Tanh layers can generate a new candidate vectorIt is added in neuron.
it=sigmod (Wi·[ht-1,xt]+bi)
Wherein, itIndicate input gate, WiIndicate input gate weight parameter, biIndicate input gate amount of bias, WCIndicate this moment
Network exports weight parameter, bCIndicate that network exports amount of bias.
By the state vector C at neuron t-1 momentt-1With ftIt is multiplied, discards the nonsensical information in part, and withResults added, that is, complete neuron state by the renewal process at t-1 moment to t moment, new candidate vector CtMeeting
It is changed according to the degree for updating neuron state.
Finally determine that the output valve of long memory network in short-term, the process of output need two steps according to neuron state,
First using the output par, c in sigmod layers of determining neuron state, then neuron state is handled by tanh layers
Numerical value between to -1 to 1, sigmod layers and tanh layers of output multiplied result are output.
ot=sigmod (Wo·[ht-1,xt]+bo)
ht=ot·tanh(Ct)
Wherein, otIndicate out gate, WoAnd boRespectively indicate out gate weight coefficient and amount of bias.
4th step inputs unmanned plane electromagnetic signal waveform and modulation based on the Modulation Mode Recognition model of third step building
Mode data, repetitive exercise Modulation Mode Recognition model parameter realize unmanned plane electromagnetic signal Modulation Mode Recognition.
By calling the reference numeral data file of second step storage, by unmanned plane electromagnetic signal waveform and modulation system number
According to the Modulation Mode Recognition model for the deep learning that input third step has constructed, actual result and theoretical mark are calculated by cross entropy
The loss of label, the method iteration optimization Modulation Mode Recognition model parameter declined using optimizer based on gradient, and realize unknown
Unmanned plane electromagnetic signal Modulation Mode Recognition, as shown in figure 5, by Modulation Mode Recognition model (depth integration provided by the invention
Model) recognition result, with convolutional neural networks, long memory network and wavelet transformation+SVM identification knot in short-term
Fruit compares, the results show that serial deep learning unmanned plane electromagnetic signal Modulation Mode Recognition model proposed by the present invention
Recognition accuracy is significantly higher than other three kinds of classical ways, in the section that signal-to-noise ratio is 0-20dB, identification proposed by the present invention
The model classification accuracy that is averaged reaches 90.7%, the results showed that identification model of the invention is more applicable than classical signals processing method
Classify in unmanned plane electromagnetic signal modulation system.
Claims (3)
1. a kind of unmanned plane electromagnetic signal management and Modulation Mode Recognition method, it is characterised in that:
Specific step is as follows:
The first step obtains unmanned plane electromagnetic signal details;
The unmanned plane electromagnetic signal details, including unmanned plane acquired electromagnetic data and environmental data;
Second step, the storage of unmanned plane electromagnetic signal details;
Unmanned plane electromagnetic signal details based on acquisition, are successively stored in MYSQL data for signal data and environmental data
In library, and electromagnetic signal waveform and modulation system are stored as to the MAT file of reference numeral;
Third step constructs Modulation Mode Recognition model;
Based on unmanned plane electromagnetic signal modulation system otherness, deep learning convolutional neural networks and long short-term memory net are merged
Network constructs unmanned plane electromagnetic signal Modulation Mode Recognition model;
4th step, training Modulation Mode Recognition model realization Modulation Mode Recognition;
Based on the Modulation Mode Recognition model constructed, electromagnetic signal waveform and modulation system data repetitive exercise modulation methods are inputted
Formula identification model parameter realizes the Modulation Mode Recognition to unknown unmanned plane electromagnetic signal waveform.
2. a kind of unmanned plane electromagnetic signal management according to claim 1 and Modulation Mode Recognition method, it is characterised in that:
Acquired electromagnetic data described in the first step include signal waveform data, signal modulation mode, signal working frequency points, sample frequency and
Signal-to-noise ratio;Environmental data includes the phase information between signal source latitude and longitude coordinates, amblent air temperature and bay.
3. a kind of unmanned plane electromagnetic signal management according to claim 1 and Modulation Mode Recognition method, it is characterised in that:
Convolutional neural networks and long memory network in short-term are merged by the way of serial and concurrent in third step;In parallel model, not
It is the characteristic pattern for having respective network attribute that treated signal waveform data, which passes through two kinds of network mappings simultaneously, by connecting entirely
Signal modulation mode characteristic of division vector is superposed to after connecing layer;In serial model, untreated signal waveform data is first
First pass through after convolutional neural networks are mapped as local feature figure, by the long memory network in short-term of this characteristic pattern input, by it is long in short-term
Memory network is mapped as signal modulation mode characteristic of division vector;
The input of convolutional neural networks model is original unmanned plane electromagnetic signal Wave data, and the output of each convolutional layer is passed through
Input after nonlinear activation as next convolutional layer portrays a variety of tune by way of multiple convolutional layers stack come profound
Difference between mode classification processed;Each convolutional layer is made of the convolution kernel of multiple and different sizes, and the size of convolution kernel is by manually setting
Determine and training preceding initiation parameter, convolution kernel is shared in each sample of signal, so each convolution kernel claims
It for a feature extraction unit, is connected between the neuron of different layers by convolution kernel, it is raw to pass through the convolution kernel having a size of k
At characteristic results yiProcess it is as follows:
yi=f (Wxi:i+k-1+b)
Wherein, W indicates weight coefficient, xi:i+k-1Indicate digital communication signal from the i moment to the signal waveform data at i+k-1 moment,
B is offset parameter, and f () is nonlinear activation function, using ReLu activation primitive;
The first step of long memory network in short-term is the information that decision is abandoned from neuron, and this functionality is by forgetting Men Laishi
Existing, which can read the input x of current t momenttH is exported with the hidden layer at t-1 momentt-1, as the memory in neuron
Information generates the multiple of numerical value between one 0 to 1:
ft=sigmod (Wf·[ht-1,xt]+bf)
Wherein, ftIt indicates to forget door, WfIt indicates to forget door weight parameter, ht-1Indicate state output, xtIndicate that t moment length is remembered in short-term
Recall network inputs data, bfIt indicates to forget door amount of bias;
Then the new information in neuron is determined, wherein input gate determines the updated value in neuron, and tanh layers can generate one
A new candidate vectorIt is added in neuron:
it=sigmod (Wi·[ht-1,xt]+bi)
Wherein, itIndicate input gate, WiIndicate input gate weight parameter, biIndicate input gate amount of bias, WCIndicate this moment network
Export weight parameter, bCIndicate that network exports amount of bias;
By the state vector C at neuron t-1 momentt-1With ftIt is multiplied, discards the nonsensical information in part, and withKnot
Fruit is added, that is, completes neuron state by the renewal process at t-1 moment to t moment, new candidate vector CtIt can be according to update
The degree of neuron state is changed:
Finally determine that the output valve of long memory network in short-term, the process of output need two steps according to neuron state, first
Using the output par, c in sigmod layers of determining neuron state, then by neuron state by tanh layers processing obtain -1 to
Numerical value between 1, sigmod layers and tanh layers of output multiplied result is output:
ot=sigmod (Wo·[ht-1,xt]+bo)
ht=ot·tanh(Ct)
Wherein, otIndicate out gate, WoAnd boRespectively indicate out gate weight coefficient and amount of bias.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910432119.9A CN110099020A (en) | 2019-05-23 | 2019-05-23 | A kind of unmanned plane electromagnetic signal management and Modulation Mode Recognition method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910432119.9A CN110099020A (en) | 2019-05-23 | 2019-05-23 | A kind of unmanned plane electromagnetic signal management and Modulation Mode Recognition method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110099020A true CN110099020A (en) | 2019-08-06 |
Family
ID=67448938
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910432119.9A Pending CN110099020A (en) | 2019-05-23 | 2019-05-23 | A kind of unmanned plane electromagnetic signal management and Modulation Mode Recognition method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110099020A (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107392097A (en) * | 2017-06-15 | 2017-11-24 | 中山大学 | A kind of 3 D human body intra-articular irrigation method of monocular color video |
CN108600135A (en) * | 2018-04-27 | 2018-09-28 | 中国科学院计算技术研究所 | A kind of recognition methods of signal modulation mode |
WO2018200045A1 (en) * | 2017-04-27 | 2018-11-01 | Raytheon Company | Machine learning algorithm with binary pruning technique for automatic intrapulse modulation recognition |
CN109726524A (en) * | 2019-03-01 | 2019-05-07 | 哈尔滨理工大学 | A kind of rolling bearing remaining life prediction technique based on CNN and LSTM |
CN109765539A (en) * | 2019-01-28 | 2019-05-17 | 珠海格力电器股份有限公司 | Indoor user behavior monitoring method and apparatus, electrical equipment and home furnishing monitoring system |
CN109787929A (en) * | 2019-02-20 | 2019-05-21 | 深圳市宝链人工智能科技有限公司 | Signal modulate method, electronic device and computer readable storage medium |
-
2019
- 2019-05-23 CN CN201910432119.9A patent/CN110099020A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018200045A1 (en) * | 2017-04-27 | 2018-11-01 | Raytheon Company | Machine learning algorithm with binary pruning technique for automatic intrapulse modulation recognition |
CN107392097A (en) * | 2017-06-15 | 2017-11-24 | 中山大学 | A kind of 3 D human body intra-articular irrigation method of monocular color video |
CN108600135A (en) * | 2018-04-27 | 2018-09-28 | 中国科学院计算技术研究所 | A kind of recognition methods of signal modulation mode |
CN109765539A (en) * | 2019-01-28 | 2019-05-17 | 珠海格力电器股份有限公司 | Indoor user behavior monitoring method and apparatus, electrical equipment and home furnishing monitoring system |
CN109787929A (en) * | 2019-02-20 | 2019-05-21 | 深圳市宝链人工智能科技有限公司 | Signal modulate method, electronic device and computer readable storage medium |
CN109726524A (en) * | 2019-03-01 | 2019-05-07 | 哈尔滨理工大学 | A kind of rolling bearing remaining life prediction technique based on CNN and LSTM |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Setiono et al. | Neural-network feature selector | |
Raitoharju et al. | Training radial basis function neural networks for classification via class-specific clustering | |
CN106919980A (en) | A kind of increment type target identification system based on neuromere differentiation | |
Tung et al. | eFSM—A novel online neural-fuzzy semantic memory model | |
WO2014060001A1 (en) | Multitransmitter model of the neural network with an internal feedback | |
CN109766995A (en) | The compression method and device of deep neural network | |
Zambrano et al. | Efficient computation in adaptive artificial spiking neural networks | |
CN114429189B (en) | High-universality landslide displacement rate early warning method | |
CN111858565B (en) | Multi-dimensional high-precision track intelligent prediction method based on line segment clustering | |
Bodyanskiy et al. | Adaptive multivariate hybrid neuro-fuzzy system and its on-board fast learning | |
CN116562908A (en) | Electric price prediction method based on double-layer VMD decomposition and SSA-LSTM | |
KR20210039921A (en) | Operation method of deep neural network system configured to optimize neural network model | |
CN115392451A (en) | Method for avoiding forgetting of artificial neural network for sustainable learning | |
Lin et al. | Designing mamdani type fuzzy rule using a collaborative FCM scheme | |
Gao et al. | Road Traffic Freight Volume Forecast Using Support Vector Machine Combining Forecasting. | |
Sabes et al. | Reinforcement learning by probability matching | |
CN110188621A (en) | A kind of three-dimensional face expression recognition methods based on SSF-IL-CNN | |
CN110099020A (en) | A kind of unmanned plane electromagnetic signal management and Modulation Mode Recognition method | |
Yang et al. | Hyper-spectral Image Pixel Classification based on Golden Sine and Chaotic Spotted Hyena Optimization Algorithm | |
Kil | Function Approximation Based on a Network with Kernel Functions of Bounds and Locality: an Approach of Non‐Parametric Estimation | |
Chuang et al. | Adaptive fuzzy regression clustering algorithm for TSK fuzzy modeling | |
Fukai | Computational models of Idling brain activity for memory processing | |
He et al. | Application of neural network model based on combination of fuzzy classification and input selection in short term load forecasting | |
Yu | Optimization of Artificial Intelligence Algorithm based on Neural Network in Complex System | |
Parsa | Bayesian-based Multi-Objective Hyperparameter Optimization for Accurate, Fast, and Efficient Neuromorphic System Designs |
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: 20190806 |