CN112699777A - Blind signal modulation type identification method based on convolutional neural network - Google Patents
Blind signal modulation type identification method based on convolutional neural network Download PDFInfo
- Publication number
- CN112699777A CN112699777A CN202011585710.7A CN202011585710A CN112699777A CN 112699777 A CN112699777 A CN 112699777A CN 202011585710 A CN202011585710 A CN 202011585710A CN 112699777 A CN112699777 A CN 112699777A
- Authority
- CN
- China
- Prior art keywords
- signal
- neural network
- convolutional neural
- histogram
- phase difference
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration by the use of histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
- G06T7/44—Analysis of texture based on statistical description of texture using image operators, e.g. filters, edge density metrics or local histograms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Abstract
The invention provides a blind signal modulation type identification method based on a convolutional neural network, which uses a phase difference histogram and an amplitude histogram calculated by three different sampling rates as signal characteristic parameters, classifies the signal characteristic parameters through a convolutional neural network classifier so as to obtain a signal modulation type, and has strong immunity to frequency offset and inaccurate bandwidth due to the fact that the texture of the phase difference histogram is insensitive to the frequency offset and the inaccurate bandwidth, so that the blind signal modulation type identification method is suitable for blind signal modulation type identification in a complex electromagnetic environment.
Description
Technical Field
The invention belongs to the field of radio communication, and particularly relates to a blind signal modulation type identification method based on a convolutional neural network.
Background
In the field of radio signal monitoring, modulation identification technology is more faced with blind signal identification, i.e. the conditions of unknown signal carrier frequency, unknown signal bandwidth and unknown signal symbol rate. In a complex electromagnetic environment, it is difficult to accurately measure the carrier frequency and bandwidth of the blind signal. From the viewpoint of engineering practicability, the modulation identification technology is required to have the immunity capability of frequency deviation and inaccurate bandwidth. When using modulated identification products provided by some well-known manufacturers, in the face of a blind signal, the operator often has to use several sets of carrier frequencies and bandwidths to identify each product separately to confirm the authenticity of the identification.
In the traditional modulation type identification method based on instantaneous parameters and high-order cumulant, the characteristic parameters are sensitive to the frequency deviation of a receiver and the bandwidth of a filter, and the performance in practical engineering is difficult to meet the application requirement. In the modulation type identification method based on the constellation diagram, an accurate signal symbol rate needs to be obtained and sampling rate conversion is carried out on sampling data, and besides high calculation cost, the accurate symbol rate cannot be obtained under the condition of low carrier-to-noise ratio, so that the application of the modulation type identification method is limited. In a modulation type recognition algorithm using a convolutional neural network, training data sets are often generated by a simulation means, the inherent characteristics of the data sets are frequency offset-free, and a trained model is difficult to adapt to an actual engineering application environment; some modulation identification methods directly using IQ data as input data of a neural network have the problem of being sensitive to frequency offset, interference and carrier-to-noise ratio.
Disclosure of Invention
In order to solve the problems, the invention provides a blind signal modulation type identification method based on a convolutional neural network, which takes three groups of signal phase difference histograms with different sampling rates and one group of amplitude histograms as input characteristic parameters of the convolutional neural network, and uses a radio frequency receiver to collect signal characteristic parameters under the conditions of different modulation types, carrier-to-noise ratios, symbol frequencies and frequency offsets as a training data set to train the convolutional neural network, so that the immunity of an identification model to frequency offsets and inaccurate bandwidths is improved.
In order to achieve the purpose, the invention adopts the following technical scheme:
a blind signal modulation type identification method based on a convolutional neural network comprises the following steps:
s1, the radio frequency receiver uses the sampling rate FSCollecting time domain 0 intermediate frequency complex signal IQ stream, dividing IQ stream into M sampling numberBLKIQ block S ofBLKCalculating the IQ block SBLKAnd estimating the signal bandwidthUsing a cut-off frequency ofTo IQ block SBLKPerforming filtering to obtain filtered IQ block;
S2 according to the sampling rate FSSum signal bandwidth estimationCalculating three IQ data extraction factors E4、E2And E1;
S3, according to the extraction factor E4、E2And E1Calculating a signal phase difference histogram and an amplitude histogram;
s4, in the training stage of the convolutional neural network classifier model, changing the modulation type, the carrier-to-noise ratio, the symbol rate and the carrier frequency offset of a signal source, acquiring an IQ block, and calculating and recording the phase difference histogram and the amplitude histogram of S3 as a training and verification data set for training the convolutional neural network classifier model;
s5, in the application stage of the convolutional neural network classifier model, aiming at the IQ block of the signal to be identified, calculating the phase difference histogram and the amplitude histogram of S3, and identifying the modulation type of the signal by using the convolutional neural network classifier model.
Further, the decimation factor E of S24、E2And E1The specific calculation steps are as follows:
further, the step of specifically calculating the phase difference histogram and the amplitude histogram in S3 is as follows:
s31, using the decimation factor E4、E2And E1To IQ blocks respectivelyInteger extraction is carried out to obtain three down-sampling rate IQ blocks SBLK4、SBLK2And SBLK1;
S32 for IQ block SBLK4、SBLK2And SBLK1Respectively calculate its phase array Sp4、Sp2And Sp1Calculating a phase difference array Spd4、Spd2And Spd1Wherein, in the step (A),
Spd4(k)= Sp4(k + 1) - Sp4 (k),
Spd2(k)= Sp2(k + 1) - Sp2(k),
Spd1(k)= Sp1(k + 1) - Sp1(k),
k belongs to (0, …, N-2) as a data point index in the phase array, and N is the actual sampling point number of each IQ block;
s33, phase difference array Spd4、Spd2And Spd1The value is translated to 0-359 interval from +/-180 interval and rounded, and distribution histogram H of three phase difference arrays is calculatedpd4、Hpd2And Hpd1And at the same time,
put Hpd4(k4) Is (H)pd4(178)+Hpd4(182))/2,k4∈(179,…,181),
Put Hpd2(k2) Is (H)pd2(177)+Hpd2(183))/2,k2∈(178,…,182),
Put Hpd1(k1) Is (H)pd1(176)+Hpd1(184))/2,k1∈(177,…,183),
H is to bepd4、Hpd2And Hpd1As a phase profile characteristic of the signal;
s34, IQ block SBLK2The amplitude distribution interval of (1) is normalized to 0-359, rounded, and an amplitude histogram H is calculatedmagAs an amplitude distribution characteristic of the signal.
The invention has the beneficial effects that:
1) in a sampling rate interval with the bandwidth of 1-4 times, phase difference histograms of different types of modulation signals are remarkably changed, and special texture characteristics exist, and the amplitude histogram characteristics are combined, so that the method can identify various common types of analog and digital modulation signals such as AM, FM, 2/4FSK, MSK/GMSK, 2/4/8PSK, 16/64QAM and the like;
2) when the carrier wave has frequency deviation, the texture of the phase difference histogram also deviates on a phase axis along with the frequency deviation, but the texture characteristics are not changed, the influence of the bandwidth of a filter on the texture characteristics is small, the sensitivity of the amplitude histogram to the frequency deviation and the bandwidth of the filter is low, and the characteristic parameter calculation does not depend on the symbol rate, so that the method has strong immunity to the frequency deviation and the inaccurate bandwidth and can still work in a complex electromagnetic environment with low carrier-to-noise ratio and strong channel interference;
3) the signal feature calculation process of the method is low in complexity, and the method has the characteristic of being easy to realize in engineering by combining with a mature pre-trained or modified convolutional neural network classifier.
Drawings
FIG. 1 is a flow chart of a blind signal modulation type identification method based on a convolutional neural network;
Detailed Description
The following describes in detail embodiments of the system according to the present invention with reference to examples.
In an embodiment, the radio frequency receiver uses the USRP B210, the modulation signal source uses the E4432B, and the IQ block sampling number M is setBLK=16384, receiver and signal source center frequency 1 GHz.
In the training phase of the convolutional neural network model, automatic training data acquisition software controls the modulation type, the symbol rate and the output level of a signal source through a control interface of E4432B, wherein the modulation type is respectively set to be 2FSK, 4FSK, MSK, GMSK, BPSK, QPSK, 8PSK, 16QAM and 64QAM, the symbol rate uses Bd = (6.0, 8.4, 10.8, 13.2, 15.6, 46.8) ksps, and the output level of the signal source is set to be (-60, -69, -75, -81, -87, -90, -93, -96) dBm;
acquisition software controlled receiver sampling rate FSGreater than 4 times the modulation signal bit rate BbSetting B210 gain G =50dB, setting receiver frequency offset dF = (± 0.2B)b,±0.1Bb,0)×BbTwo IQ blocks are collected for each parameter combination, and 5280 IQ blocks are collected for 9 kinds of digital modulation;
for analog modulation signals FM and AM, 240 narrowband FM IQ blocks are collected by using an analog interphone as a signal source, 240 wideband FM IQ blocks are recorded by using FM broadcasting stations with different frequency points as the signal source, 240 IQ data blocks are recorded by using an airport tower with a voice signal section and a navigation station as the signal source, and 240 IQ data blocks are recorded for the signal source by using AM modulation output of E4432B.
Through the signal sample collection process, the data set of 11 modulation types comprises 6240 IQ blocks, a phase difference histogram and an amplitude histogram of each IQ block are calculated and recorded to serve as a signal characteristic parameter data set, and the specific steps of characteristic parameter calculation and convolutional neural network classifier training are as follows:
t1, calculating fast Fourier transform magnitude spectra and estimating signal bandwidth for each IQ data blockUsing a cut-off frequency ofTo IQ block SBLKPerforming filtering to obtain filtered IQ block;
T2 according to the sampling rate FSSum signal bandwidth estimationCalculating three IQ data extraction factors E4、E2And E1The specific calculation steps are as follows:
t3, according to the decimation factor E4、E2And E1Calculating a phase difference histogram and an amplitude histogram of the signal, the specific calculation steps including:
r31, using decimation factor E4、E2And E1To IQ blocks respectivelyInteger extraction is carried out to obtain three down-sampling rate IQ blocks SBLK4、SBLK2And SBLK1;
T32 for IQ block SBLK4、SBLK2And SBLK1Respectively calculate its phase array Sp4、Sp2And Sp1Calculating a phase difference array Spd4、Spd2And Spd1Wherein, in the step (A),
Spd4(k)= Sp4(k + 1) - Sp4 (k),
Spd2(k)= Sp2(k + 1) - Sp2(k) ,
Spd1(k)= Sp1(k + 1) - Sp1(k),
k belongs to (0, N-2) and is a data point subscript in the phase array, and N is the actual sampling point number of each IQ block;
t33 phase difference array Spd4、Spd2And Spd1The values are translated to 0-359 from +/-180 intervals and rounded, and distribution histograms H of the three phase difference arrays are calculatedpd4、Hpd2And Hpd1And at the same time,
put Hpd4(k4) Is (H)pd4(178)+Hpd4(182))/2,k4∈(179,…,181),
Put Hpd2(k2) Is (H)pd2(177)+Hpd2(183))/2,k2∈(178,…,182),
Put Hpd1(k1) Is (H)pd1(176)+Hpd1(184))/2,k1∈(177,…,183),
H is to bepd4、Hpd2And Hpd1As a phase profile characteristic of the signal;
t34, IQ block S to reduce sampling rateBLK2The amplitude distribution interval is normalized to 0-359 interval and rounded, and an amplitude distribution histogram H is calculatedmagAs an amplitude distribution characteristic of the signal.
T4, in this embodiment, the convolutional neural network classifier uses a modified inclusion v3 model, and inputs a grayscale image of 150 × 180 specification, and the specific implementation steps of the training are as follows:
t41, drawing the three phase difference histograms and the three amplitude histograms into the same single-channel image matrix by using OpenCV according to Hpd4、Hpd2、Hpd1And HmagThe image files are sequentially arranged from top to bottom in equal intervals, each parameter area in the image matrix is separated by 5 pixels, the image matrix is stored as image files in 11 folders, and 70% of the image files are randomly selected as a training set, 20% of the image files are selected as a verification set, and 10% of the image files are selected as a test set;
t42, setting the classifier class list as AM, FM, 2FSK, 4FSK, MSK, GMSK, BPSK, QPSK, 8PSK, 16QAM and 64QAM, performing transfer training on the pre-training model, and recording the trained model for modulation identification.
After the convolutional neural network classifier model training is completed, the blind signal modulation recognition method specifically comprises the following steps:
r1, frequency point CF and rough bandwidth B designated according to identification task commandWThe RF receiver sets the center frequency to CF at a sampling rate FS>4×BWCollecting time domain 0 intermediate frequency complex signal IQ stream, dividing IQ stream into M sampling numberBLKIQ block S ofBLKCalculating the IQ block SBLKAnd estimating the signal bandwidthUsing a cut-off frequency ofTo IQ block SBLKPerforming filtering to obtain filtered IQ block;
R2 according to sampling rate FSSum signal bandwidth estimationCalculating three IQ data extraction factors E4、E2And E1;
R3 according to the decimation factor E4、E2And E1Calculating a phase difference histogram H of a signalpd4、Hpd2、Hpd1Sum amplitude histogram Hmag;
R4, drawing the three phase difference histograms and the three amplitude histograms into the same single-channel image matrix by using OpenCV according to Hpd4、Hpd2、Hpd1And HmagThe image matrix is sequentially arranged from top to bottom in an equal way, each parameter area in the image matrix is separated by 5 pixels, RGB three channels are set to have the same pixel value, the image matrix is transmitted to a classifier to be used as input, and the recognition result and the reliability are obtained after the prediction of the classifier.
The above description is only a preferred embodiment of the present application and is not intended to limit the present application, and various modifications and changes may be made by those skilled in the art without departing from the spirit and principle of the present application, and any modifications, equivalents, improvements, etc. made therein are intended to be included within the scope of the present application. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (3)
1. A blind signal modulation type identification method based on a convolutional neural network is characterized by comprising the following steps:
s1, the radio frequency receiver uses the sampling rate FSCollecting time domain 0 intermediate frequency complex signal IQ stream, dividing IQ stream into M sampling numberBLKIQ block S ofBLKCalculating the IQ block SBLKBy fast Fourier transforming the magnitude spectrum, estimating the signal bandwidth from the magnitude spectrumUsing a cut-off frequency ofTo IQ block SBLKPerforming filtering to obtain filtered IQ block;
S2 according to the sampling rate FSSum signal bandwidth estimationCalculating three IQ data extraction factors E4、E2And E1;
S3, according to the extraction factor E4、E2And E1Calculating a signal phase difference histogram and an amplitude histogram;
s4, in the training stage of the convolutional neural network classifier model, changing the modulation type, the carrier-to-noise ratio, the symbol rate and the carrier frequency offset of a signal source, and calculating and recording the phase difference histogram and the amplitude histogram of S3 as a training and verification data set for training the convolutional neural network classifier model;
and S5, in the application stage of the convolutional neural network classifier model, aiming at the signal to be identified, calculating the phase difference histogram and the amplitude histogram of S3, and identifying the modulation type of the signal by using the convolutional neural network classifier model.
3. the blind signal modulation type identification method based on the convolutional neural network as claimed in claim 1, wherein the step of calculating the phase difference histogram and the amplitude histogram specifically in S3 is as follows:
s31, using the decimation factor E4、E2And E1To IQ block S respectivelyBLKInteger extraction is carried out to obtain three down-sampling rate IQ blocks SBLK4、SBLK2And SBLK1;
S32 for IQ block S with reduced sampling rateBLK4、SBLK2And SBLK1Respectively calculate its phase array Sp4、Sp2And Sp1Calculating a phase difference array Spd4、Spd2And Spd1Wherein, in the step (A),
Spd4(k)= Sp4(k + 1) - Sp4 (k),
Spd2(k)= Sp2(k + 1) - Sp2(k),
Spd1(k)= Sp1(k + 1) - Sp1(k),
k belongs to (0, …, N-2) and is a data point subscript in the phase array, and N is the actual sampling point number of each data block;
s33, phase difference array Spd4、Spd2And Spd1The value is translated to 0-359 interval from +/-180 interval and rounded, and distribution histogram H of three phase difference arrays is calculatedpd4、Hpd2And Hpd1And at the same time,
put Hpd4(k4) Is (H)pd4(178)+Hpd4(182))/2,k4∈(179,…,181),
Put Hpd2(k2) Is (H)pd2(177)+Hpd2(183))/2,k2∈(178,…,182),
Put Hpd1(k1) Is (H)pd1(176)+Hpd1(184))/2,k1∈(177,…,183),
H is to bepd4、Hpd2And Hpd1Phase distribution as a signalPerforming sign;
s34, IQ block S for reducing sampling rateBLK2The amplitude distribution interval of (1) is normalized to 0-359, rounded, and an amplitude histogram H is calculatedmagAs an amplitude distribution characteristic of the signal.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011585710.7A CN112699777A (en) | 2020-12-29 | 2020-12-29 | Blind signal modulation type identification method based on convolutional neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011585710.7A CN112699777A (en) | 2020-12-29 | 2020-12-29 | Blind signal modulation type identification method based on convolutional neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112699777A true CN112699777A (en) | 2021-04-23 |
Family
ID=75511343
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011585710.7A Pending CN112699777A (en) | 2020-12-29 | 2020-12-29 | Blind signal modulation type identification method based on convolutional neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112699777A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN115277324A (en) * | 2022-07-25 | 2022-11-01 | 电信科学技术第五研究所有限公司 | FSK signal identification method based on convolutional neural network |
CN117155748A (en) * | 2023-10-27 | 2023-12-01 | 四川轻化工大学 | Modulation format identification method based on multidimensional amplitude distribution characteristics |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105119857A (en) * | 2015-07-17 | 2015-12-02 | 电子科技大学 | Radar inter-station low-jitter anti-interference signal communication link technology |
CN108566260A (en) * | 2018-02-01 | 2018-09-21 | 西安电子科技大学 | It is a kind of based on the concealed communication method for disturbing point multiple access |
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 |
WO2018205676A1 (en) * | 2017-05-08 | 2018-11-15 | 京东方科技集团股份有限公司 | Processing method and system for convolutional neural network, and storage medium |
CN110224956A (en) * | 2019-05-06 | 2019-09-10 | 安徽继远软件有限公司 | Modulation Identification method based on interference cleaning and two stages training convolutional neural networks model |
CN110557209A (en) * | 2019-07-19 | 2019-12-10 | 中国科学院微电子研究所 | Broadband signal interference monitoring method |
CN111083077A (en) * | 2019-12-06 | 2020-04-28 | 成都华日通讯技术有限公司 | Method for realizing modulation recognition of 2ASK signal and AM signal by combining neural network |
CN111259798A (en) * | 2020-01-16 | 2020-06-09 | 西安电子科技大学 | Modulation signal identification method based on deep learning |
CN111585922A (en) * | 2020-03-23 | 2020-08-25 | 成都奥特为科技有限公司 | Modulation mode identification method based on convolutional neural network |
US20200342328A1 (en) * | 2019-04-26 | 2020-10-29 | Naver Corporation | Training a convolutional neural network for image retrieval with a listwise ranking loss function |
CN111884962A (en) * | 2020-06-01 | 2020-11-03 | 山东师范大学 | Signal modulation type classification method and system based on convolutional neural network |
-
2020
- 2020-12-29 CN CN202011585710.7A patent/CN112699777A/en active Pending
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105119857A (en) * | 2015-07-17 | 2015-12-02 | 电子科技大学 | Radar inter-station low-jitter anti-interference signal communication link technology |
WO2018205676A1 (en) * | 2017-05-08 | 2018-11-15 | 京东方科技集团股份有限公司 | Processing method and system for convolutional neural network, and storage medium |
CN108566260A (en) * | 2018-02-01 | 2018-09-21 | 西安电子科技大学 | It is a kind of based on the concealed communication method for disturbing point multiple access |
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 |
US20200342328A1 (en) * | 2019-04-26 | 2020-10-29 | Naver Corporation | Training a convolutional neural network for image retrieval with a listwise ranking loss function |
CN110224956A (en) * | 2019-05-06 | 2019-09-10 | 安徽继远软件有限公司 | Modulation Identification method based on interference cleaning and two stages training convolutional neural networks model |
CN110557209A (en) * | 2019-07-19 | 2019-12-10 | 中国科学院微电子研究所 | Broadband signal interference monitoring method |
CN111083077A (en) * | 2019-12-06 | 2020-04-28 | 成都华日通讯技术有限公司 | Method for realizing modulation recognition of 2ASK signal and AM signal by combining neural network |
CN111259798A (en) * | 2020-01-16 | 2020-06-09 | 西安电子科技大学 | Modulation signal identification method based on deep learning |
CN111585922A (en) * | 2020-03-23 | 2020-08-25 | 成都奥特为科技有限公司 | Modulation mode identification method based on convolutional neural network |
CN111884962A (en) * | 2020-06-01 | 2020-11-03 | 山东师范大学 | Signal modulation type classification method and system based on convolutional neural network |
Non-Patent Citations (2)
Title |
---|
JUHONG WANG等: "Semantic segmentation of high-resolution images", 《SCIENCE CHINA(INFORMATION SCIENCES)》 * |
董睿杰等: "基于深度学习的通信辐射源调制样式识别方法", 《空军预警学院学报》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
CN113869227B (en) * | 2021-09-29 | 2023-06-09 | 西南交通大学 | Signal modulation mode identification method, device, equipment and readable storage medium |
CN115277324A (en) * | 2022-07-25 | 2022-11-01 | 电信科学技术第五研究所有限公司 | FSK signal identification method based on convolutional neural network |
CN115277324B (en) * | 2022-07-25 | 2023-11-10 | 电信科学技术第五研究所有限公司 | FSK signal identification method based on convolutional neural network |
CN117155748A (en) * | 2023-10-27 | 2023-12-01 | 四川轻化工大学 | Modulation format identification method based on multidimensional amplitude distribution characteristics |
CN117155748B (en) * | 2023-10-27 | 2024-01-23 | 四川轻化工大学 | Modulation format identification method based on multidimensional amplitude distribution characteristics |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112699777A (en) | Blind signal modulation type identification method based on convolutional neural network | |
CN107124381B (en) | Automatic identification method for digital communication signal modulation mode | |
CN107948107B (en) | Digital modulation signal classification method based on joint features | |
CN112511477A (en) | Hybrid satellite communication modulation identification method and system based on constellation diagram and deep learning | |
CN105388500B (en) | The raising method of CPFSK signal carrier tracking accuracies | |
CN109547373A (en) | Frequency deviation estimating method and frequency deviation estimation system for ofdm system frequency domain strong interference environment | |
CN102497343A (en) | Combined modulation recognition method based on clustering and support vector machine | |
CN111327554B (en) | Feature extraction method for digital modulation signal identification | |
CN108718288A (en) | Recognition of digital modulation schemes method based on convolutional neural networks | |
CN108052956A (en) | Wireless light communication subcarrier modulation constellation recognition methods under a kind of atmospheric turbulance | |
CN102571230A (en) | Distributed collaborative signal identification method based on blind estimation of higher order statistics and signal to noise ratio | |
CN110191073B (en) | Modulation mode identification method based on deep learning and suitable for changing scene | |
CN106169070A (en) | The communication specific emitter identification method and system represented based on cooperation | |
CN104756456A (en) | Method and apparatus for recognizing modulation format of quadrature modulation signal | |
CN112737992A (en) | Underwater sound signal modulation mode self-adaptive in-class identification method | |
US8155244B2 (en) | Demodulation using blind constellation identification for communication signals | |
CN109587091B (en) | Modulation format identification method of coherent optical communication system based on logistic regression algorithm | |
CN111903240B (en) | Analysis and identification method based on signal fine feature extraction | |
CN111079347A (en) | Signal-to-noise ratio estimation method based on deep learning by using constellation diagram | |
CN116506273A (en) | Novel MPSK modulation signal identification and classification method | |
CN111814703B (en) | HB-based signal joint feature extraction method under non-reconstruction condition | |
CN115913849A (en) | Electromagnetic signal identification method based on one-dimensional complex value residual error network | |
Zhang et al. | Modulation format recognition based on CNN in satellite communication system | |
CN113472483A (en) | Blind estimation method for digital modulation signal code element rate and code element conversion time | |
CN113518049B (en) | Modulation identification method based on fractional low-order polar coordinates and deep learning |
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 | ||
AD01 | Patent right deemed abandoned | ||
AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20230228 |