CN112087272B - Automatic detection method for electromagnetic spectrum monitoring receiver signal - Google Patents
Automatic detection method for electromagnetic spectrum monitoring receiver signal Download PDFInfo
- Publication number
- CN112087272B CN112087272B CN202010770061.1A CN202010770061A CN112087272B CN 112087272 B CN112087272 B CN 112087272B CN 202010770061 A CN202010770061 A CN 202010770061A CN 112087272 B CN112087272 B CN 112087272B
- Authority
- CN
- China
- Prior art keywords
- signal
- detector
- difference
- local
- branch
- 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.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/20—Monitoring; Testing of receivers
Abstract
The invention discloses an automatic detection method for electromagnetic spectrum monitoring receiver signals, which comprises signal enhancement characterization and ensemble learning discrimination, wherein the signal enhancement characterization is used for solving local mean difference, local median difference and local morphological processing difference of frequency spectrums of each signal under different scales, and then an enhancement characterization matrix is formed. The integrated learning judgment comprises the learning of a branch detector and the learning of an integrated detector, wherein a signal detection training sample set is formed by a plurality of enhanced representation matrixes of signals, a sample subset is distributed to train the branch detector, and the detection result is recorded; and then constructing an integrated learning training sample set by using the detection result to train the integrated detector. After the training of the integrated detector is finished, the branch detector and the integrated detector are combined, and the integrated detector outputs a final automatic signal detection result. The invention has stronger signal characterization capability and signal identification capability, thereby realizing good signal detection accuracy and stability.
Description
Technical Field
The invention relates to the field of electromagnetic spectrum monitoring receivers, in particular to an automatic signal detection method for an electromagnetic spectrum monitoring receiver.
Background
In the task of monitoring the electromagnetic spectrum, a signal of interest needs to be detected from a wide-range spectrum, and a foundation is provided for subsequent operations such as signal analysis, demodulation and recording. The traditional signal detection work is mainly manually finished by workers, high requirements are placed on the operation level and the signal identification level of the workers, and the detection result is easily influenced by subjective factors of the workers and is unstable. In addition, the speed of manual signal detection is slow, and the requirement of rapid detection under the condition of dense signals cannot be met. Thus, the importance of automatic signal detection functionality to electromagnetic spectrum receivers is increasing.
The existing automatic signal detection method mainly automatically detects signals through the difference of the signal and the background energy amplitude of a frequency spectrum, firstly, a discrimination threshold value is calculated through a mean value filtering method or a median filtering method, and then, local components at different frequencies in the frequency spectrum are compared with the threshold value. If the amplitude is larger than or equal to the threshold value, judging that a signal exists in the corresponding frequency; and if the amplitude is smaller than the threshold value, determining that no signal exists at the corresponding frequency. The prior art has the following defects: (1) the detection rate of weak signals is low, and low-intensity signals can be missed; (2) the adaptability to the strong fluctuating frequency spectrum background is poor, and a large number of error detection results can be generated under the strong fluctuating frequency spectrum background. Therefore, the prior art cannot effectively distinguish signals from complex strong fluctuating frequency spectrum backgrounds, so that the detection accuracy is low, and the automatic signal detection requirement under a complex electromagnetic environment cannot be met.
Disclosure of Invention
The invention provides an automatic signal detection method for an electromagnetic spectrum monitoring receiver, aiming at solving the problem that the detection accuracy is low due to the fact that signals and complex strong fluctuating spectrum backgrounds cannot be effectively distinguished in the prior art, and the effectiveness of automatic signal detection is improved.
The invention adopts the following technical scheme:
an automatic detection method for electromagnetic spectrum monitoring receiver signals comprises the following steps:
step 1: signal enhancement characterization
Step 1.1: solving local mean difference, local median difference and local morphological processing difference of the frequency spectrum of each signal under different scales;
step 1.2: the local mean difference, the local median difference and the local morphological processing difference of each signal are spliced with original frequency spectrum data to form an enhanced representation matrix;
step 2: ensemble learning discrimination
Performing branch detector learning and integrated detector learning;
step 2.1: the enhanced characterization matrixes of the signals form a signal detection training sample set, and m subsets extracted from the signal detection training sample set are used for training m branch signal detectors;
step 2.2: after the training of the m branch detectors is finished, processing the whole training samples by adopting the branch detectors to obtain m groups of detection results, combining the m groups of detection results to form an integrated learning training sample set, and training the integrated detector by utilizing the integrated learning training sample set;
Step 2.3: after the training of the integrated detector is completed, the branch detector and the integrated detector are combined, namely, the processing result of the branch detector is used as the input of the integrated detector, and the final automatic signal detection result is output by the integrated detector.
Preferably, the local mean difference dmean() Is defined as follows:
wherein f represents frequency, s represents scale parameter for solving local mean difference, f' represents current data point frequency for calculation, and si() Representing the frequency spectrum to be processed;
local median difference dmedian() Is defined as follows:
dmedian(f,s)=si(f)-median(si(f-s),...,si(f+s)) (2)
wherein, mean(s)i(f-s),...,si(f + s)) represents si(f-s),...,siThe median value of (f + s);
the definition of the local morphological processing difference exposure () is:
wherein M represents a structural element of a morphological operation, DMIs the region of action of M.
Preferably, the enhanced characterization matrix comprises mean-based preprocessing results at different scales, median-based preprocessing results at different scales, and morphology-based preprocessing results at different scales, with multi-scale analysis capability.
Preferably, the signal detection training sample set is:
T={(x1,y1),...,(xi,yi),...(xn,yn)} (6)
wherein x isiRepresenting normalized spectral data, yiRepresenting a signal distribution label, n representing the number of training samples;
the ensemble learning training sample set is as follows:
z(i,j)Representing the result of the ith signal sample processed by the jth branch detector; for the ith sample, the processing results of all m branch detectors constitute the vector z(i,1),...,z(i,j),...z(i,m)]As input signal for the integrated detector.
Preferably, the machine learning model employed by the branch detector is an artificial neural network.
The invention has the beneficial effects that:
(1) the signal characterization capability is stronger, and the difference between the signal and the complex frequency spectrum background can be distinguished by adopting an enhanced characterization method, so that the signal detection accuracy is improved.
(2) The signal identification capability is stronger, and the complementary enhancement of the detection capability is realized by integrating a plurality of detection models, so that the good signal detection accuracy and stability are achieved.
(3) By adopting a hierarchical training mode of a branch detector and an integrated detector, the method has excellent novel signal learning adaptability and performance expansion capability.
Drawings
Fig. 1 is a schematic representation of signal enhancement.
FIG. 2 is a schematic diagram of a branch detector-integrated detector training process.
Fig. 3 is a schematic view of the overall process of automatic signal detection.
Detailed Description
The following description of the embodiments of the present invention will be made with reference to the accompanying drawings:
With reference to fig. 1 to 3, an automatic signal detection method for an electromagnetic spectrum monitoring receiver includes the following steps:
one important reason for the low accuracy of signal detection is that the difference between the signal and the strong fluctuating background in the spectrum is not obvious, and the effective clue information for distinguishing the signal from the background is insufficient. Aiming at the problem, the patent provides a multi-scale signal enhancement characterization method, which displays the difference between a signal and a background under different scales, and provides more obvious difference clue information for a subsequent discriminator through an enhanced signal characterization mode, thereby reducing the difficulty in distinguishing between the signal and a frequency spectrum background and providing a basis for improving the automatic detection accuracy of the signal.
Step 1: signal enhancement characterization
Step 1.1: and solving the local mean difference, the local median difference and the local morphological processing difference of the frequency spectrum of each signal under different scales.
Local mean difference dmean() Is defined as:
wherein f represents frequency, s represents scale parameter for solving local mean difference, f' represents current data point frequency for calculation, and si() Representing the spectrum to be processed;
local median difference dmedian() Is defined as follows:
dmedian(f,s)=si(f)-median(si(f-s),...,si(f+s)) (2)
wherein, mean(s)i(f-s),...,si(f + s)) represents si(f-s),...,siThe median value of (f + s);
The definition of the local morphological processing difference encoding () is:
wherein M represents a structural element of a morphological operation, DMIs the region of action of M.
Step 1.2: and the local mean difference, the local median difference and the local morphological processing difference of each signal are spliced with the original frequency spectrum data to form an enhanced characterization matrix.
The enhanced characterization matrix comprises preprocessing results based on mean values at different scales, preprocessing results based on median values at different scales and preprocessing results based on morphology at different scales.
Each row of the enhanced representation matrix is related to the distribution of the signals, but the rows are not identical and have diversity, so that more obvious and richer difference information is provided for subsequent signal detection, and the effectiveness of signal detection is improved.
Different from a signal detection mode based on a single threshold value in the prior method, the invention adopts a multi-detector integrated learning mode to judge whether the signal exists.
And 2, step: ensemble learning discrimination
Performing branch detector learning and integrated detector learning;
step 2.1: the enhanced representation matrixes of the signals form a signal detection training sample set, and m subsets extracted from the signal detection training sample set are used for training m branch signal detectors;
The signal detection training sample set is as follows:
T={(x1,y1),...,(xi,yi),...(xn,yn)} (6)
wherein x isiRepresenting normalized spectral data, yiRepresenting the signal distribution label and n representing the number of training samples.
The machine learning model used by the branch detector is an artificial neural network. The structures of the branch detectors are different from each other, thereby ensuring the capability difference and complementarity of the detectors.
Step 2.2: after the training of the m branch detectors is finished, processing the whole training samples by adopting the branch detectors to obtain m groups of detection results, combining the m groups of detection results to form an integrated learning training sample set, and training the integrated detector by utilizing the integrated learning training sample set;
the ensemble learning training sample set is as follows:
z(i,j)representing the result of the ith signal sample after being processed by the jth branch detector; for the ith sample, the processing results of all m branch detectors constitute a vector [ z ](i,1),...,z(i,j),...z(i,m)]As input signal for the integrated detector.
The integrated detector integrates the advantages of each branch detector, and the obtained information is more comprehensive, so that the integrated detector has better signal detection accuracy and stability.
Step 2.3: after the training of the integrated detector is completed, the branch detector and the integrated detector are combined, namely, the processing result of the branch detector is used as the input of the integrated detector, and the final automatic signal detection result is output by the integrated detector.
It is to be understood that the above description is not intended to limit the present invention, and the present invention is not limited to the above examples, and those skilled in the art may make various changes, modifications, additions and substitutions within the spirit and scope of the present invention.
Claims (4)
1. An automatic signal detection method for an electromagnetic spectrum monitoring receiver is characterized by comprising the following steps:
step 1: signal enhancement characterization
Step 1.1: solving local mean difference, local median difference and local morphological processing difference of the frequency spectrum of each signal under different scales;
local mean difference dmean() Is defined as follows:
wherein f represents frequency, s represents scale parameter for solving local mean difference, f' represents current data point frequency for calculation, and si() Representing the spectrum to be processed;
local median difference dmedian() Is defined as:
dmedian(f,s)=si(f)-median(si(f-s),...,si(f+s)) (2)
wherein, mean(s)i(f-s),...,si(f + s)) represents si(f-s),...,siThe median value of (f + s);
the definition of the local morphological processing difference exposure () is:
wherein M represents a structural element of a morphological operation, DMIs the region of action of M;
step 1.2: the local mean difference, the local median difference and the local morphological processing difference of each signal are spliced with original frequency spectrum data to form an enhanced representation matrix;
Step 2: ensemble learning discrimination
Performing branch detector learning and integrated detector learning;
step 2.1: the enhanced characterization matrixes of the signals form a signal detection training sample set, and m subsets extracted from the signal detection training sample set are used for training m branch signal detectors;
step 2.2: after the training of the m branch detectors is finished, processing all training samples by adopting the branch detectors to obtain m groups of detection results, combining the m groups of detection results to form an ensemble learning training sample set, and training the ensemble detector by utilizing the ensemble learning training sample set;
step 2.3: after the training of the integrated detector is completed, the branch detector and the integrated detector are combined, namely, the processing result of the branch detector is used as the input of the integrated detector, and the final automatic signal detection result is output by the integrated detector.
2. The method according to claim 1, wherein said detecting step comprises the steps of,
the enhanced characterization matrix comprises preprocessing results based on mean values under different scales, preprocessing results based on median values under different scales and preprocessing results based on morphology under different scales, and has multi-scale analysis capability.
3. The method according to claim 1, wherein said detecting step comprises the steps of,
the signal detection training sample set is:
T={(x1,y1),...,(xi,yi),...(xn,yn)} (6)
wherein x isiRepresenting normalized spectral data, yiRepresenting a signal distribution label, n representing the number of training samples;
the ensemble learning training sample set is as follows:
z(i,j)representing the result of the ith signal sample after being processed by the jth branch detector; for the ith sample, the processing results of all m branch detectors constitute a vector [ z ](i,1),...,z(i,j),...z(i,m)]As input signal for the integrated detector.
4. The method of claim 1, wherein the machine learning model used by the finger detector is an artificial neural network.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010770061.1A CN112087272B (en) | 2020-08-04 | 2020-08-04 | Automatic detection method for electromagnetic spectrum monitoring receiver signal |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010770061.1A CN112087272B (en) | 2020-08-04 | 2020-08-04 | Automatic detection method for electromagnetic spectrum monitoring receiver signal |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112087272A CN112087272A (en) | 2020-12-15 |
CN112087272B true CN112087272B (en) | 2022-07-19 |
Family
ID=73734809
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010770061.1A Active CN112087272B (en) | 2020-08-04 | 2020-08-04 | Automatic detection method for electromagnetic spectrum monitoring receiver signal |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112087272B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016201648A1 (en) * | 2015-06-17 | 2016-12-22 | 中国科学院自动化研究所 | Steganalysis method based on local learning |
EP3466343A1 (en) * | 2016-05-26 | 2019-04-10 | Vinno Technology (Suzhou) Co., Ltd. | Pulse doppler ultrahigh spectrum resolution imaging processing method and processing system |
CN109840588A (en) * | 2019-01-04 | 2019-06-04 | 平安科技(深圳)有限公司 | Neural network model training method, device, computer equipment and storage medium |
CN110008817A (en) * | 2019-01-29 | 2019-07-12 | 北京奇艺世纪科技有限公司 | Model training, image processing method, device, electronic equipment and computer readable storage medium |
CN110009573A (en) * | 2019-01-29 | 2019-07-12 | 北京奇艺世纪科技有限公司 | Model training, image processing method, device, electronic equipment and computer readable storage medium |
KR102064301B1 (en) * | 2019-08-28 | 2020-01-09 | 세종대학교 산학협력단 | Signal detection apparatus using ensemble machine learning based on MIMO system and method thereof |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108932480B (en) * | 2018-06-08 | 2022-03-15 | 电子科技大学 | Distributed optical fiber sensing signal feature learning and classifying method based on 1D-CNN |
US11593634B2 (en) * | 2018-06-19 | 2023-02-28 | Adobe Inc. | Asynchronously training machine learning models across client devices for adaptive intelligence |
-
2020
- 2020-08-04 CN CN202010770061.1A patent/CN112087272B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016201648A1 (en) * | 2015-06-17 | 2016-12-22 | 中国科学院自动化研究所 | Steganalysis method based on local learning |
EP3466343A1 (en) * | 2016-05-26 | 2019-04-10 | Vinno Technology (Suzhou) Co., Ltd. | Pulse doppler ultrahigh spectrum resolution imaging processing method and processing system |
CN109840588A (en) * | 2019-01-04 | 2019-06-04 | 平安科技(深圳)有限公司 | Neural network model training method, device, computer equipment and storage medium |
CN110008817A (en) * | 2019-01-29 | 2019-07-12 | 北京奇艺世纪科技有限公司 | Model training, image processing method, device, electronic equipment and computer readable storage medium |
CN110009573A (en) * | 2019-01-29 | 2019-07-12 | 北京奇艺世纪科技有限公司 | Model training, image processing method, device, electronic equipment and computer readable storage medium |
KR102064301B1 (en) * | 2019-08-28 | 2020-01-09 | 세종대학교 산학협력단 | Signal detection apparatus using ensemble machine learning based on MIMO system and method thereof |
Non-Patent Citations (1)
Title |
---|
通信辐射源个体识别技术研究;丁敏;《中国优秀硕士学位论文全文数据库》;20180215;全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN112087272A (en) | 2020-12-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN114549522B (en) | Textile quality detection method based on target detection | |
CN113989279B (en) | Plastic film quality detection method based on artificial intelligence and image processing | |
CN113838054B (en) | Mechanical part surface damage detection method based on artificial intelligence | |
CN110443117B (en) | Wind turbine generator fault diagnosis method | |
CN111797887A (en) | Anti-electricity-stealing early warning method and system based on density screening and K-means clustering | |
CN109409216B (en) | Speed self-adaptive indoor human body detection method based on subcarrier dynamic selection | |
CN115100214A (en) | Textile quality detection method based on image processing | |
CN115294109A (en) | Real wood board production defect identification system based on artificial intelligence, and electronic equipment | |
CN106708009A (en) | Ship dynamic positioning measurement system multiple-fault diagnosis method based on support vector machine clustering | |
CN112087272B (en) | Automatic detection method for electromagnetic spectrum monitoring receiver signal | |
CN112147221B (en) | Steel rail screw hole crack identification method and system based on ultrasonic flaw detector data | |
CN113447455B (en) | Fiber material detection system based on terahertz technology | |
CN113408805A (en) | Lightning ground flashover identification method, device, equipment and readable storage medium | |
CN111738081A (en) | Deep neural network sonar target detection method difficult for sample retraining | |
CN112014821A (en) | Unknown vehicle target identification method based on radar broadband characteristics | |
JPH11258296A (en) | Automatic discrimination method for partial discharge | |
CN109870404A (en) | A kind of awning Structural Damage Identification, device and terminal device | |
CN113743351B (en) | Remote sensing image scene recognition method based on edge direction semantic information | |
CN115082722B (en) | Equipment defect detection method, system, terminal and medium based on forward sample | |
CN117370898B (en) | Electronic data safety control system | |
CN116486178B (en) | Defect detection method and device, electronic equipment and storage medium | |
CN110399398B (en) | Method for identifying visible alarm high-incidence area of power transmission line channel | |
CN117480461A (en) | Vibration measurement error determination method and vibration error discrimination system using the same | |
CN115468647A (en) | Acoustic linear frequency modulation identification method and system for aircraft fault | |
CN113780273A (en) | Water meter pointer reading method based on target detection and key point detection |
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 | ||
CB02 | Change of applicant information | ||
CB02 | Change of applicant information |
Address after: 266555 No. 98 Xiangjiang Road, Huangdao District, Qingdao City, Shandong Province Applicant after: CLP kesiyi Technology Co.,Ltd. Address before: 266555 No. 98 Xiangjiang Road, Huangdao District, Qingdao City, Shandong Province Applicant before: CHINA ELECTRONICS TECHNOLOGY INSTRUMENTS Co.,Ltd. |
|
GR01 | Patent grant | ||
GR01 | Patent grant |