CN114584227A - Automatic burst signal detection method - Google Patents

Automatic burst signal detection method Download PDF

Info

Publication number
CN114584227A
CN114584227A CN202210033937.3A CN202210033937A CN114584227A CN 114584227 A CN114584227 A CN 114584227A CN 202210033937 A CN202210033937 A CN 202210033937A CN 114584227 A CN114584227 A CN 114584227A
Authority
CN
China
Prior art keywords
signal
frequency
time
neural network
burst
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.)
Granted
Application number
CN202210033937.3A
Other languages
Chinese (zh)
Other versions
CN114584227B (en
Inventor
张宇阳
解韦桐
欧阳玫丹
甘翼
丛迅超
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
CETC 10 Research Institute
Original Assignee
CETC 10 Research Institute
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by CETC 10 Research Institute filed Critical CETC 10 Research Institute
Priority to CN202210033937.3A priority Critical patent/CN114584227B/en
Publication of CN114584227A publication Critical patent/CN114584227A/en
Application granted granted Critical
Publication of CN114584227B publication Critical patent/CN114584227B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/0082—Monitoring; Testing using service channels; using auxiliary channels
    • H04B17/0087—Monitoring; Testing using service channels; using auxiliary channels using auxiliary channels or channel simulators
    • 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
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B1/00—Details of transmission systems, not covered by a single one of groups H04B3/00 - H04B13/00; Details of transmission systems not characterised by the medium used for transmission
    • H04B1/0003—Software-defined radio [SDR] systems, i.e. systems wherein components typically implemented in hardware, e.g. filters or modulators/demodulators, are implented using software, e.g. by involving an AD or DA conversion stage such that at least part of the signal processing is performed in the digital domain
    • H04B1/0007—Software-defined radio [SDR] systems, i.e. systems wherein components typically implemented in hardware, e.g. filters or modulators/demodulators, are implented using software, e.g. by involving an AD or DA conversion stage such that at least part of the signal processing is performed in the digital domain wherein the AD/DA conversion occurs at radiofrequency or intermediate frequency stage
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/20—Monitoring; Testing of receivers
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/309—Measuring or estimating channel quality parameters
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/30—Monitoring; Testing of propagation channels
    • H04B17/373—Predicting channel quality or other radio frequency [RF] parameters
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Electromagnetism (AREA)
  • Quality & Reliability (AREA)
  • Theoretical Computer Science (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Monitoring And Testing Of Transmission In General (AREA)

Abstract

The automatic burst signal detection method disclosed by the invention has the advantages of high detection speed, high detection stability and high precision. The invention is realized by the following technical scheme: in a scene of space electromagnetic signal detection discovery, a monitoring device captures a space electromagnetic radiation signal by an antenna and converts the space electromagnetic radiation signal into an electric signal, the electric signal is output to an acquisition module through radio frequency channel analog frequency conversion, filtering and amplification, a single-channel analog signal output by a receiving channel is subjected to AD sampling and digital down conversion to generate a section of signal sampling data and STFT, broadband time-frequency matrix data is output to a detection processing module, a training deep neural network model is loaded, and after the reasoning result is post-processed, the confidence probability of whether the burst signal contained in the signal sampling data of the section corresponding to the time-frequency matrix has the electromagnetic signal, the signal starting time, the signal duration, the signal center frequency and the signal bandwidth parameter are formed, so that the detection discovery of the electromagnetic space burst signal and the extraction of the parameters such as the signal occupation time and the signal bandwidth are completed.

Description

Automatic burst signal detection method
Technical Field
The invention relates to a signal monitoring technology in a complex electromagnetic environment, in particular to an automatic burst signal detection method of a broadband electromagnetic spectrum monitoring receiver.
Background
In modern communication systems, such as wireless communication systems of satellite communication, mobile communication, etc., data transmission is performed based on wireless channels, and because the wireless channel environment is poor and channel parameters tend to change relatively sharply with time during communication, burst mode is usually adopted during signal demodulation, that is, the channel parameters are considered to be constant in a short time. Before the transmitting end transmits data, a preamble for estimating channel parameters is usually added before the data to be transmitted. The core task of the receiving end is to detect the starting point of the received data, i.e. burst signal detection. The physical meaning of the method is to find out the burst starting point of a burst signal in a received signal and find out the burst starting point of the signal through a burst signal detection technology. Generally speaking, several different signal patterns, 1) continuous fixed frequency signals, often exist in an electromagnetic environment. The continuous fixed-frequency signal means that the signal continuously appears near a specific frequency point, the frequency is rarely changed along with the time, the signal is approximately represented as a vertical line on a time-frequency diagram, and the direction of the signal is stable. In an actual electromagnetic environment, there are many fixed frequency signals in the channel. The general fixed frequency signal has limited occupied bandwidth, longer duration and short class radius, and a class, namely a class cross part, is constructed. 2) Intermittent fixed frequency signals. The intermittent fixed-frequency signal refers to a signal which appears intermittently at a specific frequency point, and is intermittent at the specific frequency point, and the direction of the signal is relatively stable. 3) Random noise signals. The random noise signal refers to various channel noises and spatial interference signals. The signals are represented as randomly distributed scattered points on a time-frequency diagram. According to practical experience, the amplitude of random noise signals is generally relatively small, but the number of the random noise signals is very large, and the random noise signals have a great relationship with a frequency range and a time range, some random noise signals are very strong in frequency band, and some random noise signals are weak in frequency band. 4) The burst signal is a signal which appears randomly and stays for a period of time after the occurrence and then disappears, and is also represented as a short straight line on a time-frequency diagram. The burst signal has the characteristics of burstiness and instantaneity, and a certain method is needed to judge whether a received signal is noise or a signal carrying data before demodulation, otherwise, the signal is lost. For burst signals, if a demodulator for continuous signals is adopted to demodulate the burst signals, the following problems are inevitably caused that when the signals exist, the demodulator is accurately locked, and the signals can be correctly demodulated; when the signal disappears (only noise is present), the demodulator will drift. 5) A frequency hopping signal. The transmission power of the frequency hopping signal on different channels is consistent, and the automatic detection model characteristics of the frequency hopping signal of the receiver receiving signals on the whole frequency band are also basically consistent. To locate and interfere with a particular frequency hopping communication signal, the interferer first has to detect the frequency hopping signal and attempt to sort out the frequency sets of each hopping station. Due to the fact that the electromagnetic environment is very complex, signals are distributed densely, the data size to be processed is large, and various signals are mixed together frequently. The main frequency hopping signal detection and analysis methods at present comprise a self-correlation detection algorithm, a self-adaptive threshold denoising algorithm, an algorithm combining a self-adaptive threshold and a fixed threshold, and a time-frequency analysis method. The time-frequency analysis method has great advantages in analyzing frequency hopping signals, however, modern communication environments are increasingly complex, the quantity of intercepted data is increasingly huge due to the fact that electromagnetic environments of short wave bands are increasingly complex, the frequency hopping signals also have the trends of frequency band broadening, hopping speed increasing and the like, and it is difficult to detect the frequency hopping signals from mass data by directly using a time-frequency amplitude three-dimensional correlation method. In all of the above methods, incoming wave azimuth information of a signal is not emphasized, and the calculation amount is large, so that it is difficult to meet the requirement of rapid detection.
The detection technology of the burst signal comprises a sliding window method, a double sliding window method, a self-adaptive threshold energy adjusting detection method, a short-time autocorrelation method and a frequency domain detection algorithm. The common main methods are an energy method, an autocorrelation method and the like. The sliding window method is the most commonly used method in burst signal detection algorithms based on the energy of the received signal. However, the above algorithm has poor performance or large calculation amount, and it is difficult to meet the requirement of high-speed burst signal detection anyway. Among them, the sliding window method has a disadvantage that the threshold Th value is difficult to determine.
The development of radio communication technology is gradually expanding, and the demand of various industries for radio spectrum resources is rapidly increasing, and the radio communication environment is gradually complicated. Under the condition, monitoring management of radio spectrum resources is imperative, and the monitoring receiver serving as core equipment for spectrum monitoring management has irreplaceable effects on purifying electromagnetic environment and improving radio spectrum utilization rate. Based on the wide application of the burst communication mechanism, the electromagnetic signal monitoring device is a device for capturing electromagnetic signals in the environment, analyzing and processing the concerned signals and acquiring target signal parameters. Aiming at the situation that burst communication is more and more, the electromagnetic signal monitoring equipment needs to extract parameters of a target signal of interest, firstly, the signal needs to be detected in a complex multi-signal electromagnetic environment, the time occupation range and the frequency occupation range of each signal are determined, and single signals are filtered out one by one based on the occupation ranges and used for subsequent detailed analysis and parameter extraction. In the traditional broadband electromagnetic spectrum monitoring equipment, the bandwidth and the center frequency of a signal are detected in a signal frequency domain, namely a signal one-dimensional frequency spectrum, filtering processing is carried out based on the frequency detection result, and then the signal starting time and the burst duration are detected in a time domain. The method has the defects of low detection and discovery probability of the burst electromagnetic radiation signal, inaccurate time-frequency parameter extraction and the like.
In recent years, with the development of technologies in power control, access method, and the like in wireless communication, signal detection is performed by a frequency and time step detection method, and the discovery probability of electromagnetic signals is becoming lower. It is therefore proposed to perform a time-frequency transformation of the signal, i.e. a self-detection of the signal in a two-dimensional time-frequency diagramA method for dynamically acquiring signal time-frequency parameters. The current method for detecting signals based on time-frequency diagrams mainly adopts the steps of setting a threshold, and realizing the detection of signal time-frequency parameters by passing the threshold through the signal amplitude and fusing the spectral lines passing the threshold. In the above-mentioned detection scheme, one of the links is to adopt an adaptive threshold correlation detection algorithm, compare the matched and correlated peak value with a set threshold, and construct a decision statistic by using the ratio of signal power estimates before and after correlation, that is, to detect whether a burst signal is coming by judging whether a decision variable exceeds a predetermined threshold. The implementation complexity of the correlation detection algorithm is high, and the absolute threshold is sensitive to the signal level or the noise intensity. The false alarm probability P of the system due to the decision thresholdfAnd long decisions of local unique code sequences. More than one of the sample point correlation values may be greater than the decision threshold, so that a maximum sample point correlation value is selected to determine the starting boundary of the unique code in the burst signal. In the detection of the unique code by the correlation method, if a fixed threshold is adopted, the correct detection is difficult. The method for setting the hard threshold has the problems of high requirement on the signal-to-noise ratio, low detection and discovery probability, high false alarm rate and the like.
Disclosure of Invention
The invention aims to provide an automatic burst signal detection method which can simplify the operation of monitoring equipment, has high burst signal detection speed, small detection error, high detection and discovery probability and high time-frequency parameter estimation precision and aims at overcoming the defects of low detection and discovery probability, inaccurate time-frequency parameter extraction and the like of the broadband electromagnetic spectrum monitoring equipment for burst electromagnetic radiation signals in the prior art. The invention has small calculated amount, high speed and small detection error, and adopts the technical scheme for solving the technical problems that: an automatic burst signal detection method is characterized by comprising the following technical steps: in a scene of space electromagnetic signal detection discovery, a broadband electromagnetic spectrum monitoring device receives signals by adopting a receiving antenna, captures space electromagnetic radiation signals and converts the space electromagnetic radiation signals into electric signals, and the converted electric signals are subjected to analog frequency conversion, matched filtering and amplification by a radio frequency channel module and then output to an acquisition module for time sampling and frequency sampling; the acquisition module carries out automatic AD sampling, digital channel preprocessing and digital down-conversion according to a single-channel analog signal output by the radio frequency channel module to generate a section of signal sampling data, and carries out Short Time Fourier Transform (STFT) on the signal sampling data, the STFT windows the data in sections, selects a time-frequency localized window function, calculates power spectrums at different moments to obtain frequency information of the function near the moment tau and a broadband time-frequency matrix of a short time Fourier transform result, extracts a frame of broadband time-frequency matrix data and outputs the frame of broadband time-frequency matrix data to the detection processing module; the detection processing module adopts a high-precision floating point calculation operator to construct a deep neural network model, trains the deep neural network model by using high-precision marking data, loads the trained deep neural network model, detects and inputs burst multi-signal by using broadband time-frequency matrix data into the deep neural network model, automatically generates parameter information such as confidence probability of whether burst signals exist in an electromagnetic space, signal starting time, signal duration, signal center frequency, signal bandwidth and the like by using the deep neural network model for reasoning and post-processing a reasoning result, and forms the confidence probability of whether the burst signals exist in signal sampling data of a section corresponding to the time-frequency matrix; the deep neural network model adopts a basic network and combines a multilayer feature extraction network, utilizes characteristic information output by different feature network layers of a time-frequency boundary estimation network under different resolution scales to estimate confidence probability and time-frequency parameters of whether signals exist, fuses estimation results of detection results under different scales, combines detection results corresponding to the same target burst signal, sets a threshold, detects the burst signal according to the confidence probability of whether the burst signal exists, and detects the signal starting time, the signal duration, the signal center frequency and the signal bandwidth parameter of the probability over-threshold, thereby completing detection discovery of the electromagnetic space burst signal and extraction of the signal occupation time and frequency parameters.
Compared with the prior art, the invention has the following beneficial effects.
The operation of the monitoring device is simplified. Aiming at the automatic detection and discovery of burst electromagnetic radiation signals by broadband electromagnetic spectrum monitoring equipment, in a scene of space electromagnetic signal detection and discovery, the broadband electromagnetic spectrum monitoring equipment adopts a receiving antenna to receive signals and automatically adopts the antenna to capture the space electromagnetic radiation signals and convert the space electromagnetic radiation signals into electric signals; the signals are automatically subjected to analog frequency conversion, filtering and amplification through a radio frequency channel and output to an acquisition module; the acquisition module automatically carries out AD sampling and digital down-conversion on the single-path analog signal output by the receiving channel to generate a section of signal sampling data, carries out short-time Fourier transform on the signal sampling data, and outputs broadband time-frequency matrix data to the detection processing module; the detection processing model utilizes the deep neural network model to automatically extract existing signals, time and frequency parameters, does not need to manually observe a time-frequency graph visualized by a time-frequency matrix, does not need to manually select and burst signals according to a time-frequency picture frame to determine the signal parameters, and automatically extracts parameters such as a signal time occupation range, a signal frequency occupation range and the like by a machine, so that the design of monitoring equipment is greatly simplified, the manual operation is reduced, and the reliability of engineering realization is improved.
The burst signal detection speed is high. Aiming at space electromagnetic radiation observation, through automatic comprehensive processing, burst electromagnetic signals are found, whether confidence probability, signal starting time, signal duration, signal center frequency and signal bandwidth parameters exist or not is extracted, automatic AD sampling, digital channel preprocessing and digital down-conversion are carried out according to a single-path analog signal output by a radio frequency channel module, a section of signal sampling data is generated, short-time Fourier transform (STFT) is carried out on the signal sampling data, the STFT carries out windowing on the data in sections, a time-frequency localized window function is selected, power spectrums at different moments are calculated, frequency information of the function near a moment tau and a broadband time-frequency matrix of a short-time Fourier transform result are obtained, and a frame of broadband time-frequency matrix data is extracted and output to a detection processing module; the method forms broadband time-frequency matrix data through down-conversion and short-time Fourier transform, is efficient, convenient, small in calculation amount, easy to calculate in parallel and high in speed, can construct a deep neural network model through a basic calculation unit, and utilizes the deep neural network model for automatic reasoning. Meanwhile, the deep neural network model can be deployed in a high-performance GPU or a special AI processing chip to carry out rapid parallel reasoning, so that the rapid calculation of the deep neural network model and the rapid processing of broadband time-frequency matrix data are realized, and the information of parameters such as confidence probability, signal starting time, signal duration, signal center frequency, signal bandwidth and the like are output. The detection speed of the broadband frequency spectrum monitoring equipment on the burst electromagnetic radiation signals is greatly improved.
The detection error is small, and the detection finding probability is high. The method is characterized in that a deep neural network model is constructed by adopting a high-precision floating point calculation operator according to various radiation electromagnetic signals in the space, the deep neural network model is trained by using high-precision marking data, the deep neural network model is trained by adopting a supervised training mode based on the deep neural network model, the broadband time-frequency matrix data is processed by utilizing the trained deep neural network model, and parameter information such as confidence probability, signal starting time, signal duration, signal center frequency, signal bandwidth and the like of whether burst signals exist in the electromagnetic space is automatically generated. The method adopts floating point calculation and high-precision reasoning, takes broadband time-frequency data as input, and utilizes a neural network to automatically estimate the confidence probability of whether a plurality of burst signals exist in the time-frequency data, the signal starting time, the signal duration, the signal center frequency and the signal bandwidth parameters. The broadband electromagnetic spectrum monitoring equipment can be used for detecting electromagnetic radiation signals with high stability and high precision. The method has the advantages of high detection stability, high detection precision and high time-frequency parameter estimation precision. And can utilize the simulation of computational method, verify from aspects such as frequency deviation catches the scope, signal-to-noise ratio working threshold and level receiving range separately, the verification result shows, the invention can work reliably in lower signal-to-noise ratio, the channel adaptability is moderate and have the probability characteristic of the constant false alarm, it is a effective detection method suitable for sudden signal detection of TDMA system.
Drawings
To further illustrate, but not limit, the above-described implementations of the invention, the following description of preferred embodiments is given in conjunction with the accompanying drawings, so that the details and advantages of the invention will become more apparent.
Fig. 1 is a schematic diagram of the detection principle of the automatic burst signal detection based on the multilayer deep neural network.
Fig. 2 is a schematic diagram of the reference bounding box for each scale of the signal detection setup of the present invention.
FIG. 3 is a schematic diagram of a residual error network structure in the deep neural network model of the present invention.
Detailed Description
See fig. 1. According to the invention, in a scene of space electromagnetic signal detection discovery, a broadband electromagnetic spectrum monitoring device receives signals by adopting a receiving antenna, captures space electromagnetic radiation signals and converts the space electromagnetic radiation signals into electric signals, and the converted electric signals are subjected to analog frequency conversion, matched filtering and amplification by a radio frequency channel module and then output to an acquisition module for time sampling and frequency sampling; the acquisition module carries out automatic AD sampling, digital channel preprocessing and digital down-conversion according to a single-channel analog signal output by the radio frequency channel module to generate a section of signal sampling data, and carries out Short Time Fourier Transform (STFT) on the signal sampling data, the STFT windows the data in sections, selects a time-frequency localized window function, calculates power spectrums at different moments to obtain frequency information of the function near the moment tau and a broadband time-frequency matrix of a short time Fourier transform result, extracts a frame of broadband time-frequency matrix data and outputs the frame of broadband time-frequency matrix data to the detection processing module; the detection processing module adopts a high-precision floating point calculation operator to construct a deep neural network model, trains the deep neural network model by using high-precision marking data, loads the trained deep neural network model, inputs broadband time-frequency matrix data into the deep neural network model for burst multi-signal detection, automatically generates parameter information such as confidence probability of whether burst signals exist in an electromagnetic space, signal starting time, signal duration, signal center frequency, signal bandwidth and the like by reasoning of the deep neural network model and post-processing of a reasoning result, and forms the confidence probability of whether the burst signals exist in signal sampling data of a section corresponding to the time-frequency matrix; the deep neural network model adopts a basic network and combines a multilayer feature extraction network, utilizes characteristic information output by different feature network layers of a time-frequency boundary estimation network under different resolution scales to estimate confidence probability and time-frequency parameters of whether signals exist, fuses estimation results of detection results under different scales, combines detection results corresponding to the same target burst signal, sets a threshold, detects the burst signal according to the confidence probability of whether the burst signal exists, and detects the signal starting time, the signal duration, the signal center frequency and the signal bandwidth parameter of the probability over-threshold, thereby completing detection discovery of the electromagnetic space burst signal and extraction of the signal occupation time and frequency parameters.
When the deep neural network model is trained, the combination of the confidence error of the target signal and the time-frequency parameter estimation error is adopted as a loss function for training the deep neural network model, and the calculation mode of the cost function is as follows:
Figure BDA0003467581030000051
Figure BDA0003467581030000052
wherein N is the number of effective signals in time frequency data, λ coord is the weight of estimation error of time frequency parameter, LlocThe error is estimated for the time-frequency parameters.
The deep neural network model adopts cross entropy as a loss measure and focalloss mode to carry out weighting, calculates confidence error,
Figure BDA0003467581030000061
the time-frequency boundary estimation network calculates the time-frequency parameter estimation error by adopting the following calculation method,
Figure BDA0003467581030000062
wherein, yiRepresenting the true result of the classification,
Figure BDA0003467581030000063
Representing the prediction result of the classification, fi、ti、bi、liAnd
Figure BDA0003467581030000064
respectively representing the real result and the prediction result of the time-frequency parameter. The loss function reduces the weight occupied by a large number of simple negative samples in training, and solves the problem of positive samples in one-stage target detectionThe proportion of negative samples is seriously unbalanced.
In an alternative embodiment, the input 1-channel amplitude data of the neural network is 1 × 128 × 1024, 128-dimensional time data and 1024-dimensional frequency data, and the input data is processed by the feature extraction network to form feature information of at least 4 different scales, where the dimension of the scale 1 feature information is 32 channels 16 × 128, the dimension of the scale 2 feature information is 48 channels 8 × 64, the dimension of the scale 3 feature information is 64 channels 4 × 32, and the dimension of the scale 4 feature information is 128 channels 2 × 16.
See fig. 2. The deep neural network model adopts 8 default bounding boxes with different proportions as reference for signal detection of each Cell in the characteristics aiming at the characteristic information of each scale, and sets the length-width proportions of the 8 default bounding boxes to be 1:1, 1.5:1.5, 1:2, 2:1, 1:3, 3:1, 1:5 and 5:1 according to the frequency characteristics such as the bandwidth, the duration and the like of electromagnetic signals. The time-frequency boundary estimation network forms 8 estimation results for each Cell in the feature information under each scale, namely each default boundary box corresponds to one estimation result, and the information contained in each estimation result is marked as [ P, f ]c，tc，bc，lc]Where P represents the confidence of the presence and absence of a target. f. ofcRepresenting the relative centre frequency, t, of the signalcRepresenting the relative center time of the signal (start time + half of the signal duration), bcRepresenting the relative signal bandwidth,/cIndicating the relative duration of the signal.
The time-frequency boundary estimation network adopts a relative estimation result calculation mode to calculate:
relative center frequency of target signal
Figure BDA0003467581030000065
Relative center time of target signal
Figure BDA0003467581030000066
Relative signal bandwidth
Figure BDA0003467581030000067
Relative time length of signal
Figure BDA0003467581030000068
The relative estimation result output by the time-frequency boundary estimation network can obtain the original center frequency f of the signal of the target after the inverse transformation of the formulacCenter time tgBandwidth bgAnd duration lg，
Wherein f isg,tg,bg,lgIs the true center frequency, center time, bandwidth, duration, f, of the target signalb,tb,bb,lbThe center frequency, center time, bandwidth, duration represented by the reference bounding box.
Aiming at a burst target signal detection task, the deep neural network adopts a multilayer convolutional neural network to extract input time-frequency data characteristics, predicts a target signal information result, and utilizes a residual error network as a main network model component to better complete training and prediction. The multilayer feature extraction network adopts 26 layers of neural networks for feature extraction
Figure BDA0003467581030000071
Two layers of neural networks are connected in sequence to form a residual network block, and 3 blocks are continuously superposed, wherein [16, 3 x 3 ]]Representing a convolutional neural network outputting 16 channels 3 x 3.
The table of the design parameters of the feature extraction network in the deep neural network model is as follows:
Figure BDA0003467581030000072
see fig. 3. The residual block network block inputs multichannel characteristic data by using layer normalization (LayerNorm) and an activation function (LeakyReLu), firstly, a 3 x 3 convolutional neural network is carried out, then, a normalized LayerNorm layer, a LeakyReLu activation layer, a 3 x 3 convolutional neural network and a LayerNorm layer are sequentially overlapped, then, the input 1 x 1 convolutional neural network layer is added, and finally, a LeakyReLu activation layer is overlapped.
The present invention has been described in detail with reference to the accompanying drawings, but it should be noted that the above examples are only preferred examples of the present invention, and are not intended to limit the present invention, and those skilled in the art will be able to make various modifications and changes, for example, the number of layers of different neural networks, the number of channels of each layer of neural network, the size parameter of convolution kernel, etc. may be selected and used in combination with specific engineering projects. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

Claims (10)

1. An automatic burst signal detection method is characterized by comprising the following technical steps: in a scene of space electromagnetic signal detection discovery, a broadband electromagnetic spectrum monitoring device receives signals by adopting a receiving antenna, captures space electromagnetic radiation signals and converts the space electromagnetic radiation signals into electric signals, and the converted electric signals are subjected to analog frequency conversion, matched filtering and amplification by a radio frequency channel module and then output to an acquisition module for time sampling and frequency sampling; the acquisition module carries out automatic AD sampling, digital channel preprocessing and digital down-conversion according to a single-channel analog signal output by the radio frequency channel module to generate a section of signal sampling data, and carries out Short Time Fourier Transform (STFT) on the signal sampling data, the STFT windows the data in sections, selects a time-frequency localized window function, calculates power spectrums at different moments to obtain frequency information of the function near the moment tau and a broadband time-frequency matrix of a short time Fourier transform result, extracts a frame of broadband time-frequency matrix data and outputs the frame of broadband time-frequency matrix data to the detection processing module; the detection processing module adopts a high-precision floating point calculation operator to construct a deep neural network model, trains the deep neural network model by using high-precision marking data, loads the trained deep neural network model, detects and inputs burst multi-signal by using broadband time-frequency matrix data into the deep neural network model, automatically generates the confidence probability of whether the burst signal contained in the electromagnetic space exists, the signal starting time, the signal duration, the signal center frequency and the parameter information of the signal bandwidth through the reasoning of the deep neural network model and the post-processing of the reasoning result, and forms the confidence probability of whether the burst signal contained in the signal sampling data of the corresponding section of the time-frequency matrix exists; the deep neural network model adopts a basic network and combines a multilayer feature extraction network, utilizes characteristic information output by different feature network layers of a time-frequency boundary estimation network under different resolution scales to estimate confidence probability and time-frequency parameters of whether signals exist, fuses estimation results of detection results under different scales, combines detection results corresponding to the same target burst signal, sets a threshold, detects the burst signal according to the confidence probability of whether the burst signal exists, and detects the signal starting time, the signal duration, the signal center frequency and the signal bandwidth parameter of the probability over-threshold, thereby completing detection discovery of the electromagnetic space burst signal and extraction of the signal occupation time and frequency parameters.
2. The automated burst signal detection method of claim 1, wherein: when the deep neural network model is trained, the combination of the confidence error of the target signal and the time-frequency parameter estimation error is adopted as a loss function for training the deep neural network model, and the calculation mode of the cost function is as follows:
Figure FDA0003467581020000011
wherein N is the number of effective signals in time frequency data, λ coord is the weight of estimation error of time frequency parameter, LlocThe error is estimated for the time-frequency parameters.
3. The automated burst signal detection method of claim 1, wherein: the deep neural network model adopts cross entropy as loss measurement and loss function Focal loss to carry out weighting on the basis of a cross entropy loss function, calculates confidence error,
Figure FDA0003467581020000012
the time-frequency boundary estimation network calculates the time-frequency parameter estimation error by adopting the following calculation method,
Figure FDA0003467581020000013
wherein, yiRepresenting the true result of the classification,
Figure FDA0003467581020000014
Representing the prediction result of the classification, fi、ti、bi、liAnd
Figure FDA0003467581020000015
respectively representing the real result and the prediction result of the time-frequency parameter.
4. The automated burst signal detection method of claim 1, wherein: the input 1-channel amplitude data of the neural network are 1 × 128 × 1024, 128-dimensional time data and 1024-dimensional frequency data, the input data form feature information of at least 4 different scales through the feature extraction network, wherein the dimension of the scale 1 feature information is 32 channels 16 × 128, the dimension of the scale 2 feature information is 48 channels 8 × 64, the dimension of the scale 3 feature information is 64 channels 4 × 32, and the dimension of the scale 4 feature information is 128 channels 2 × 16.
5. The automated burst signal detection method of claim 1, wherein: the deep neural network model adopts 8 default bounding boxes with different proportions as reference for signal detection of each Cell in the characteristics aiming at the characteristic information of each scale, and sets the length-width proportions of the 8 default bounding boxes to be 1:1, 1.5:1.5, 1:2, 2:1, 1:3, 3:1, 1:5 and 5:1 according to the frequency characteristics such as the bandwidth, the duration and the like of electromagnetic signals.
6. The automated burst signal detection method of claim 1, wherein: the time-frequency boundary estimation network forms 8 estimation results for each Cell in the feature information under each scale, namely each default boundary box corresponds to one estimation result, and each estimation result corresponds to one Cell in the feature information under each scaleThe estimation result contains information recorded as [ P, fc,tc,bc,lc]Where P denotes the confidence of the presence and absence of a target, fcRepresenting the relative centre frequency, t, of the signalcRepresenting the relative centre time of the signal, bcRepresenting the relative signal bandwidth,/cIndicating the relative duration of the signal.
7. The automated burst signal detection method of claim 1, wherein: the time-frequency boundary estimation network adopts a relative estimation result calculation mode to calculate:
relative center frequency of target signal
Figure FDA0003467581020000021
Relative center time of target signal
Figure FDA0003467581020000022
Relative signal bandwidth
Figure FDA0003467581020000023
Relative time length of signal
Figure FDA0003467581020000024
The relative estimation result output by the time-frequency boundary estimation network obtains the original center frequency f of the signal of the target after the inverse transformation of the formulacCenter time tgBandwidth bgAnd duration lg，
Wherein f isg,tg,bg,lgRespectively the true center frequency, center time, bandwidth, duration, f of the target signalb,tb,bb,lbThe center frequency, center time, bandwidth, duration represented by the reference bounding box.
8. The automated burst signal detection method of claim 1, wherein: aiming at a burst target signal detection task, the deep neural network model adopts a multilayer convolutional neural network to extract input time-frequency data characteristics, predicts a target signal information result, and utilizes a residual error network as a main network model component to better complete training and prediction.
9. The automated burst signal detection method of claim 1, wherein: the multilayer feature extraction network adopts 26 layers of neural networks for feature extraction
Figure FDA0003467581020000025
Two layers of neural networks are connected in sequence to form a residual network block, and 3 residual network blocks are continuously superposed, wherein [16, 3 x 3 ]]Representing a convolutional neural network outputting 16 channels 3 x 3.
10. The automated burst signal detection method of claim 9, wherein: the residual block network block inputs multichannel characteristic data by using layer normalization (LayerNorm) and an activation function (LeakyReLu), firstly, a 3 x 3 convolutional neural network is carried out, then, a normalized LayerNorm layer, a LeakyReLu activation layer, a 3 x 3 convolutional neural network and a LayerNorm layer are sequentially overlapped, then, the input 1 x 1 convolutional neural network layer is added, and finally, a LeakyReLu activation layer is overlapped.
CN202210033937.3A 2022-01-12 2022-01-12 Automatic burst signal detection method Active CN114584227B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210033937.3A CN114584227B (en) 2022-01-12 2022-01-12 Automatic burst signal detection method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210033937.3A CN114584227B (en) 2022-01-12 2022-01-12 Automatic burst signal detection method

Publications (2)

Publication Number Publication Date
CN114584227A true CN114584227A (en) 2022-06-03
CN114584227B CN114584227B (en) 2023-08-22

Family

ID=81769094

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210033937.3A Active CN114584227B (en) 2022-01-12 2022-01-12 Automatic burst signal detection method

Country Status (1)

Country Link
CN (1) CN114584227B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116244637A (en) * 2023-05-12 2023-06-09 中星联华科技（北京）有限公司 Burst signal acquisition method and device

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH1139495A (en) * 1997-07-24 1999-02-12 Nec Corp Image supervisory device
WO2001048952A1 (en) * 1999-12-24 2001-07-05 Ntt Docomo, Inc. Method and device for transmitting burst signal in mobile communication system, information distribution method, and information distribution controller
CN105207965A (en) * 2015-08-14 2015-12-30 成都中安频谱科技有限公司 Automatic VHF/UHF frequency range modulation identification method
CN107122738A (en) * 2017-04-26 2017-09-01 成都蓝色起源科技有限公司 Automatic Communication Signals Recognition based on deep learning model and its realize system
CN109271975A (en) * 2018-11-19 2019-01-25 燕山大学 A kind of electrical energy power quality disturbance recognition methods based on big data multi-feature extraction synergetic classification
CN112735435A (en) * 2020-12-25 2021-04-30 西南电子技术研究所（中国电子科技集团公司第十研究所） Voiceprint open set identification method with unknown class internal division capability
CN112751633A (en) * 2020-10-26 2021-05-04 中国人民解放军63891部队 Broadband spectrum detection method based on multi-scale window sliding

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH1139495A (en) * 1997-07-24 1999-02-12 Nec Corp Image supervisory device
WO2001048952A1 (en) * 1999-12-24 2001-07-05 Ntt Docomo, Inc. Method and device for transmitting burst signal in mobile communication system, information distribution method, and information distribution controller
CN105207965A (en) * 2015-08-14 2015-12-30 成都中安频谱科技有限公司 Automatic VHF/UHF frequency range modulation identification method
CN107122738A (en) * 2017-04-26 2017-09-01 成都蓝色起源科技有限公司 Automatic Communication Signals Recognition based on deep learning model and its realize system
CN109271975A (en) * 2018-11-19 2019-01-25 燕山大学 A kind of electrical energy power quality disturbance recognition methods based on big data multi-feature extraction synergetic classification
CN112751633A (en) * 2020-10-26 2021-05-04 中国人民解放军63891部队 Broadband spectrum detection method based on multi-scale window sliding
CN112735435A (en) * 2020-12-25 2021-04-30 西南电子技术研究所（中国电子科技集团公司第十研究所） Voiceprint open set identification method with unknown class internal division capability

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
D. SEO等: "CNN-based Burst Signal Detection with Covariance Matrix", pages 470 - 473 *
冉雨: "自适应突发通信系统的决策技术研究" *
胡亚等: "复杂信号环境下非合作突发通信信号的实时检测", vol. 34, no. 07, pages 1463 - 1468 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116244637A (en) * 2023-05-12 2023-06-09 中星联华科技（北京）有限公司 Burst signal acquisition method and device

Also Published As

Publication number Publication date
CN114584227B (en) 2023-08-22

Similar Documents

Publication Publication Date Title
US10879946B1 (en) Weak signal processing systems and methods
CN108957399B (en) Method for realizing motion detection by using LTE signal channel state information
CN107911183B (en) Unmanned aerial vehicle identification method and system based on real-time signal characteristic analysis
CN110099019B (en) LoRa modulation signal detection method based on deep learning
CN111366950A (en) Comprehensive detection method and system for satellite navigation suppression type interference and deception interference
KR102162284B1 (en) Cloud data processing gnss jamming monitoring method and system
US7613231B2 (en) Apparatus and method for detecting code of direct sequence spread spectrum signal
CN109655794B (en) Detection and identification method for suppressing interference by narrow-band self-defense noise
Kaplan et al. Measurement based FHSS–type drone controller detection at 2.4 GHz: An STFT approach
CN116400387B (en) Weak interference detection method and system based on double-array element cross correlation
CN111046025B (en) Unmanned aerial vehicle signal detection method and device
CN114584227B (en) Automatic burst signal detection method
CN113938228B (en) Network domain cooperation interference position detection method and system
CN108155959B (en) Unmanned aerial vehicle measurement and control signal detection method under complex network environment
CN116647311B (en) Unmanned aerial vehicle communication signal processing method, system and equipment based on blind source separation
CN112929141A (en) Unmanned aerial vehicle detection and identification method and system based on graph-borne signal matching
CN117318853A (en) Burst signal detection method under low signal-to-noise ratio
CN108718223B (en) Blind spectrum sensing method for non-cooperative signals
CN116961799A (en) Signal interference detection method based on time-frequency domain distribution characteristics
Deng et al. Spectrogram-based frequency hopping signal detection in a complex electromagnetic environment
KR101154166B1 (en) Method for analyzing performance of spectrum sensing methods for cognitive radio systems
KR101755240B1 (en) Signal detection apparatus using compressive sensing and method thereof
CN114401526A (en) Narrow-band interference position detection method and system based on double-threshold judgment
Stasionis et al. Burst signal detector based on signal energy and standard deviation
CN115099289B (en) System and method for distinguishing identity of radio radiation source based on weak fingerprint characteristics

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
GR01 Patent grant
GR01 Patent grant