CN109784318A - The recognition methods of Link16 data-link signal neural network based - Google Patents
The recognition methods of Link16 data-link signal neural network based Download PDFInfo
- Publication number
- CN109784318A CN109784318A CN201910187110.6A CN201910187110A CN109784318A CN 109784318 A CN109784318 A CN 109784318A CN 201910187110 A CN201910187110 A CN 201910187110A CN 109784318 A CN109784318 A CN 109784318A
- Authority
- CN
- China
- Prior art keywords
- neural network
- layer
- input
- signal
- recognition methods
- 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
Landscapes
- Character Discrimination (AREA)
Abstract
The invention discloses a kind of recognition methods of Link16 data-link signal neural network based, first input measured signal;Then the measured signal of input is pre-processed, the pretreatment includes noise reduction process and PCA dimension-reduction treatment;It will be identified in the neural network determined by pretreated measured signal input network structure and parameter again;Recognition result judgement is finally carried out according to the signal classification of neural network output.Robustness of the present invention is good, good for the Link16 data-link signal identification effect under the complex situations such as complex background, noise jamming;Time complexity of the present invention is low simultaneously, and mainly time flower in building neural network, the identification used time of last signal is few.
Description
Technical field
The present invention relates to a kind of recognition methods of Link16 data-link signal, and in particular to neural network based to one kind
The recognition methods of Link16 data-link signal.
Background technique
The characteristics of modern war be information-based combined operation and it is accurate fight, unit of entering a war is numerous, be distributed in land, sea,
Empty, day broad area, war rhythm is fast, is interleaved together between ourselves and the enemy, both sides at war's sharp fight information superiority.Link16
Data-link is to be currently being widely used a kind of tactical data link, and major function is the real-time exchange tactics letter between belligerent unit
Breath makes each belligerent unit grasp Real-time Battlefield situation and transmission Command Control Order.Link16 data-link is communicated with time division multiple acess
Mode work, each member of system works asynchronously all in accordance with unified system time-base, and frequency hopping, straight is used on signal
When expanding, jumping, many interference protection measures such as channel coding and pulsed operation, there is very strong counterreconnaissance and interference free performance.
It is to extract detection signal Higher Order Cumulants, square spectrum and frequency range etc. to the existing recognition methods of Link16 signal
Feature simultaneously identified, however in practical applications, the factors such as complex background, noise jamming, scale seriously affect the property of identification
Can, the traditional recognition method based on feature can not reach ideal recognition effect, on the one hand show under low signal-to-noise ratio
The recognition effect of signal is general, and on the other hand these methods will often take a substantial amount of time in the feature extraction of signal.
The machine learning method to grow up in recent years especially neural network is grinding for Link16 data-link signal identification
Study carefully and provides new thinking.Machine learning method has been successfully applied to the multiple fields such as images match, recognition of face, and
And achieve preferable experiment effect.Therefore, a kind of identification of Link16 data-link signal neural network based how is obtained
Method allows to effectively make up the deficiency of existing recognition methods, this is with important research significance and practical value.
Summary of the invention
It is an object of the invention to overcome above-mentioned the deficiencies in the prior art, provide that a kind of recognition effect is good, time complexity
The recognition methods of low Link16 data-link signal neural network based.
The present invention specifically uses following technical scheme:
A kind of recognition methods of Link16 data-link signal neural network based comprising following steps:
S1, input measured signal;
S2, the measured signal of step S1 input is pre-processed, the pretreatment includes at noise reduction process and PCA dimensionality reduction
Reason;
S3, will by neural network that step S2 pretreated measured signal input network structure and parameter determine into
Row identification;
S4, recognition result judgement is carried out according to the signal classification of neural network output.
Further, in step S3, the neural network of input is BP neural network, RBF neural and PNN nerve net
Network.
Further, the determination process of the structure of the neural network and parameter are as follows: BP and RBF neural include three
Layer is successively input layer, hidden layer and output layer, and PNN neural network includes four layers, is input layer, mode layer, summation layer respectively
And output layer;BP neural network hidden layer neuron number is set as 20, and training algorithm is LM algorithm, maximum frequency of training regulation
10000 times, learning rate 0.1, training precision 0.00001;RBF neural mean square error is 0.0001, radial base
Expansion rate is 2, and maximum neuron number is training sample number;The expansion rate of the radial basis function of PNN network is 1;Record
Obtain BP, RBF, PNN network structure and parameter when best identified performance.
After adopting the above technical scheme, compared with the background technology, the present invention, having the advantages that
1, the present invention is based on the identification that Link16 data-link signal is carried out under the Larger Dynamic signal-to-noise ratio of neural network, robusts
Property is good, good for the Link16 data-link signal identification effect under the complex situations such as complex background, noise jamming.
2, time complexity of the present invention is low, mainly time flower in the building of neural network, the identification of last signal
Used time is few.
Detailed description of the invention
Fig. 1 is the flow chart that the present invention is implemented;
Fig. 2 is the structural schematic diagram of BP neural network in the embodiment of the present invention;
Fig. 3 is the structural schematic diagram of PNN neural network in the embodiment of the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Embodiment
Fig. 1 shows implementing procedure of the invention, refering to what is shown in Fig. 1, the invention discloses a kind of neural network based
The recognition methods of Link16 data-link signal comprising following steps:
S1, input measured signal;
S2, the measured signal of step S1 input is pre-processed, the pretreatment include noise reduction process and PCA (it is main at
Analysis) dimension-reduction treatment;
S3, will by neural network that step S2 pretreated measured signal input network structure and parameter determine into
Row identification;The neural network inputted herein includes three kinds, respectively BP neural network (back propagation), RBF nerve
Network (Radical Basis Function, also referred to as radial basis function neural network) and PNN neural network
(Probabilistic Neural Network, also referred to as probabilistic neural network).
The determination process of network structure and parameter are as follows: BP neural network and RBF neural have three layers, are successively inputs
Layer, hidden layer and output layer;PNN neural network has four layers, is input layer, mode layer, summation layer and output layer respectively.BP nerve
Network concealed layer neuron number is set as 20, and training algorithm is LM algorithm, and maximum frequency of training provides 10000 times, learning rate
It is 0.1, training precision 0.00001;RBF neural mean square error is 0.0001, and the expansion rate of radial base is 2, maximum
Neuron number is training sample number;The expansion rate of the radial basis function of PNN network is 1, and record obtains best identified performance
When BP, RBF, PNN network structure and parameter.
Specifically, with reference to the structure chart of BP neural network shown in Fig. 2, the algorithm of BP neural network can first random initializtion
Every connecting line weight and biasing, then in training set each input x and output y, BP algorithm can all first carry out before to
Transmission, sample convert by hidden layer from input layer, are transmitted to output layer and obtain predicted value, then according to true value and predicted value
Between error execute weight and every layer of preference that reverse feedback updates every connecting line in neural network.Stop in no arrival
Only repeated the above process in the case where condition.Forwards algorithms are as follows:
In formula (1), wijFor the weight between node i and node j, bjThreshold value between node;
The mean square error of output with the sample that classification is marked in training set is calculated, in the way of minimization error, instead
To the weight of layer-by-layer adjustment network, until reach setting maximum number of iterations or experimental error down to setting precision until.
The mean square error formula of back-propagation algorithm is as follows:
In formula (2), djFor all results of output;
The structure of RBF neural is similar to BP neural network, the difference is that its hidden layer neuron kernel function (effect
Function) it is Gaussian function, the transformation of space reflection is carried out to input information, the action function of output layer neuron is linear letter
Number exports, the output result as entire neural network after carrying out linear weighted function to the information of hidden layer neuron output.Gauss
The form of function are as follows:
In formula (3), cjFor function center vector, σjFor width vector;
For the structure of PNN neural network referring to shown in Fig. 3, first layer is input layer, is responsible for input feature value;Mode layer,
The corresponding center of each neuron, calculates the matching degree of each mode in input feature value and training set, that is, phase
Like degree, its distance is sent into Gaussian function and obtains the output of mode layer;Summation layer, is responsible for the pattern layer units of each class of connection;
It is finally output layer, that one kind of highest scoring in output summation layer.The jth neuron of the i-th quasi-mode is determined in mode layer
Input/output relationship be defined by the formula:
In formula (4), xijFor function center vector;
The neuron number of summation layer is equal with the number of data classification, it is the nerve for belonging to same class in mode layer
The output of member is weighted and averaged according to the following formula:
In formula (5), L indicates the neuron number of the i-th class, viIndicate the output of the i-th class classification.
S4, recognition result judgement is carried out according to the signal classification of neural network output: is exported according to three kinds of neural networks
Signal classification judges whether signal is Link16 data-link signal, is finally completed entire identification process.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by anyone skilled in the art,
It should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with scope of protection of the claims
Subject to.
Claims (3)
1. a kind of recognition methods of Link16 data-link signal neural network based, it is characterised in that: the following steps are included:
S1, input measured signal;
S2, the measured signal of step S1 input is pre-processed, the pretreatment includes noise reduction process and PCA dimension-reduction treatment;
S3, will know in the neural network determined by the pretreated measured signal input network structure of step S2 and parameter
Not;
S4, recognition result judgement is carried out according to the signal classification of neural network output.
2. a kind of recognition methods of Link16 data-link signal neural network based according to claim 1, feature
Be: in step S3, the neural network of input is BP neural network, RBF neural and PNN neural network.
3. a kind of recognition methods of Link16 data-link signal neural network based according to claim 2, feature
It is: the structure of the neural network and the determination process of parameter are as follows: BP and RBF neural are successively inputs comprising three layers
Layer, hidden layer and output layer, PNN neural network include four layers, are input layer, mode layer, summation layer and output layer respectively;BP mind
20 are set as through network concealed layer neuron number, training algorithm is LM algorithm, and maximum frequency of training provides 10000 times, study speed
Rate is 0.1, training precision 0.00001;RBF neural mean square error is 0.0001, and the expansion rate of radial base is 2, most
Large neuron number is training sample number;The expansion rate of the radial basis function of PNN network is 1;Record obtains best identified
BP, RBF, PNN network structure and parameter when energy.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910187110.6A CN109784318A (en) | 2019-03-13 | 2019-03-13 | The recognition methods of Link16 data-link signal neural network based |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910187110.6A CN109784318A (en) | 2019-03-13 | 2019-03-13 | The recognition methods of Link16 data-link signal neural network based |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109784318A true CN109784318A (en) | 2019-05-21 |
Family
ID=66489102
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910187110.6A Pending CN109784318A (en) | 2019-03-13 | 2019-03-13 | The recognition methods of Link16 data-link signal neural network based |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109784318A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115065580A (en) * | 2022-07-28 | 2022-09-16 | 成都华日通讯技术股份有限公司 | Link16 data chain identification and parameter estimation method under broadband |
CN116979995A (en) * | 2023-09-20 | 2023-10-31 | 成都谐盈科技有限公司 | Open loop spectrum correction method for frequency hopping system |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101729207A (en) * | 2009-12-22 | 2010-06-09 | 中兴通讯股份有限公司 | Method and device for acquiring signaling |
CN104523264A (en) * | 2014-12-31 | 2015-04-22 | 深圳职业技术学院 | Electrocardiosignal processing method |
CN106774993A (en) * | 2016-12-16 | 2017-05-31 | 深圳市虚拟现实技术有限公司 | Feature point recognition method based on neutral net |
CN206488924U (en) * | 2016-12-30 | 2017-09-12 | 广东技术师范学院 | Automatic gearbox failure diagnostic apparatus based on virtual instrument |
CN107967917A (en) * | 2016-10-19 | 2018-04-27 | 福特全球技术公司 | The vehicle periphery audio classification learnt by neural network machine |
CN108494626A (en) * | 2018-03-23 | 2018-09-04 | 河海大学 | Physical installation improperly Profibus DP industrial field bus communication failure intelligent diagnosing method |
CN108668254A (en) * | 2018-05-15 | 2018-10-16 | 福州大学 | The area positioning method of WiFi signal feature based on improved BP |
KR20190015164A (en) * | 2018-12-20 | 2019-02-13 | 차의과학대학교 산학협력단 | System and method for improving sns dysfunction based on deep learning |
-
2019
- 2019-03-13 CN CN201910187110.6A patent/CN109784318A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101729207A (en) * | 2009-12-22 | 2010-06-09 | 中兴通讯股份有限公司 | Method and device for acquiring signaling |
CN104523264A (en) * | 2014-12-31 | 2015-04-22 | 深圳职业技术学院 | Electrocardiosignal processing method |
CN107967917A (en) * | 2016-10-19 | 2018-04-27 | 福特全球技术公司 | The vehicle periphery audio classification learnt by neural network machine |
CN106774993A (en) * | 2016-12-16 | 2017-05-31 | 深圳市虚拟现实技术有限公司 | Feature point recognition method based on neutral net |
CN206488924U (en) * | 2016-12-30 | 2017-09-12 | 广东技术师范学院 | Automatic gearbox failure diagnostic apparatus based on virtual instrument |
CN108494626A (en) * | 2018-03-23 | 2018-09-04 | 河海大学 | Physical installation improperly Profibus DP industrial field bus communication failure intelligent diagnosing method |
CN108668254A (en) * | 2018-05-15 | 2018-10-16 | 福州大学 | The area positioning method of WiFi signal feature based on improved BP |
KR20190015164A (en) * | 2018-12-20 | 2019-02-13 | 차의과학대학교 산학협력단 | System and method for improving sns dysfunction based on deep learning |
Non-Patent Citations (3)
Title |
---|
BÜLENT BOLAT 等: "A DATA SELECTION METHOD FOR PROBABILISTIC NEURAL NETWORKS", 《JOURNAL OF ELECTRICAL & ELECTRONICS ENGINEERING》 * |
唐斐: "基于神经网络的调制信号识别", 《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑》 * |
李国鑫: "Link16数据链信号的识别算法研究", 《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115065580A (en) * | 2022-07-28 | 2022-09-16 | 成都华日通讯技术股份有限公司 | Link16 data chain identification and parameter estimation method under broadband |
CN115065580B (en) * | 2022-07-28 | 2024-04-02 | 成都华日通讯技术股份有限公司 | Link16 data Link identification and parameter estimation method under broadband |
CN116979995A (en) * | 2023-09-20 | 2023-10-31 | 成都谐盈科技有限公司 | Open loop spectrum correction method for frequency hopping system |
CN116979995B (en) * | 2023-09-20 | 2023-12-26 | 成都谐盈科技有限公司 | Open loop spectrum correction method for frequency hopping system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109993280A (en) | A kind of underwater sound source localization method based on deep learning | |
CN112712557B (en) | Super-resolution CIR indoor fingerprint positioning method based on convolutional neural network | |
CN104835178A (en) | Low SNR(Signal to Noise Ratio) motion small target tracking and identification method | |
CN109766791B (en) | Communication signal modulation identification method based on self-encoder | |
CN108169708A (en) | The direct localization method of modular neural network | |
CN104881521B (en) | A kind of label random set filter distribution fusion method | |
CN104459661B (en) | Method for detecting rapid artillery type dim target | |
Chang et al. | Device-free indoor localization using Wi-Fi channel state information for Internet of Things | |
Shen et al. | Node identification in wireless network based on convolutional neural network | |
CN109784318A (en) | The recognition methods of Link16 data-link signal neural network based | |
KR102407835B1 (en) | Method and apparatus for classifying pulse radar signal properties based on machine learning | |
CN110045336B (en) | Convolutional neural network-based radar interference identification method and device | |
CN114997326A (en) | Airborne radar working mode identification method based on multi-feature fusion | |
Hou et al. | Jamming Recognition of carrier-free UWB cognitive radar based on MANet | |
KR102407834B1 (en) | Method and apparatus for property-based classification of long-pulse radar signals | |
Jin et al. | Radar emitter identification based on EPSD-DFN | |
CN115422977A (en) | Radar radiation source signal identification method based on CNN-BLS network | |
Ren et al. | Fast RCS data generation based on infoGAN | |
CN115932714A (en) | Single-snapshot direction-of-arrival estimation method based on hyperbolic tangent kernel correlation entropy | |
CN114676721A (en) | Radar active suppression interference identification method and system based on radial basis function neural network | |
CN111144316B (en) | Sea surface target detection method based on stack self-encoder | |
Ziyan et al. | The research of electronic countermeasure intelligence correlation analysis based on machine learning | |
Hanusa et al. | Contact clustering and fusion for preprocessing multistatic active sonar data | |
Liu et al. | Individual Intelligent Recognition Method Based on Fingerprint Features of Radar Emitter | |
Zhai et al. | A new sense-through-foliage target recognition method based on hybrid particle swarm optimization-based wavelet twin support vector machine |
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: 20190521 |