Electromagnetic sensing device for communication signals

By combining data acquisition and deep neural network recognition technology in electromagnetic perception devices, the existing devices are solved by large size, high power consumption and inability to detect and identify simultaneously, and lightweight real-time signal modulation recognition is realized, suitable for portable mobile Communication equipment and micro unmanned platforms.

CN119094290APending Publication Date: 2024-12-06SHANGHAI INST OF MICROSYSTEM & INFORMATION TECH CHINESE ACAD OF SCI
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411019304.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The existing electromagnetic sensing devices are large in size and have large power consumption, which cannot meet the perception needs of portable mobile communication devices or micro unmanned platforms, and cannot realize signal detection and modulation recognition at the same time. Traditional modulation recognition technology has strong dependence on prior information. High computational complexity and low recognition accuracy.

Method used

An electromagnetic sensing device including a data acquisition unit, a signal processing unit and an AI acceleration unit is designed. I/Q sampled data is obtained through a broadband A/D converter, and signal modulation and recognition is performed using a deep neural network based on a feature set, which includes Sequence, transient, statistical, spectral and image features, and the channel attention mechanism is used to simplify feature extraction.

Benefits of technology

It realizes the lightweight detection algorithm to quickly obtain signal frequency points and bandwidth estimation, reduces the computational complexity, and achieves the purpose of real-time identification of signal modulation methods. It is suitable for the applications of portable mobile communication devices and micro unmanned platforms.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119094290A_ABST
    Figure CN119094290A_ABST
Patent Text Reader

Abstract

The invention relates to an electromagnetic sensing device for a communication signal, and the device comprises a data collection unit which is used for obtaining an I / Q sampling data signal of the communication signal in real time; the signal processing unit is used for performing signal processing on the I / Q sampling data signal to obtain a digital baseband signal, and acquiring a feature set for signal modulation identification based on the digital baseband signal; the feature set comprises a sequence feature, an instantaneous feature, a statistical feature, a spectrum feature and an image feature; and the AI acceleration unit is used for obtaining a modulation pattern of the communication signal by adopting deep neural network reasoning according to the feature set. The method can reduce the calculation complexity, and achieves the real-time recognition of the signal modulation mode.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of cognitive radio and electromagnetic sensing technology, and in particular to an electromagnetic sensing device for communication signals. Background Art

[0002] With the continuous innovation of telecommunication technology and integrated circuit technology, digitization, miniaturization and intelligence are important development directions. At present, a large number of micro radio devices for the Internet of Things and mobile communications are put into use. The electromagnetic environment applicable to the equipment is more complex and diverse, the wireless communication capabilities are richer and more stable, and the wireless signal modulation styles that meet different needs are becoming more diverse. Especially in multimedia transmission applications, the spatial electromagnetic spectrum of the signal shows a broadband trend. These changes have put forward higher requirements on the perception methods and devices of electromagnetic signals. Cognitive radio secondary users need to accurately and quickly detect and identify unknown communication signals for opportunistic communications.

[0003] At present, most electromagnetic sensing devices for unknown signals are large in size and power consumption, which cannot meet the sensing requirements of portable mobile communication devices or micro unmanned platforms. In addition, few devices can achieve both signal detection and modulation recognition capabilities on the same device. Among the commonly used signal detection technologies, matched filter detection requires prior knowledge and synchronization information of the communication signal. Cyclostationary feature detection has strong anti-interference ability, but requires a long observation time and high computational complexity. Traditional modulation recognition technology uses a likelihood-based method, but it has a strong dependence on prior information, does not meet the requirements of blind signal processing, and has the disadvantages of low recognition accuracy and high computational complexity. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide an electromagnetic sensing device for communication signals, which can reduce the computational complexity and realize real-time recognition of signal modulation modes.

[0005] The technical solution adopted by the present invention to solve the technical problem is: to provide an electromagnetic sensing device for communication signals, comprising:

[0006] A data acquisition unit, used for acquiring I / Q sampling data signals of communication signals in real time;

[0007] A signal processing unit, configured to perform signal processing on the I / Q sampled data signal to obtain a digital baseband signal, and to obtain a feature set for signal modulation identification based on the digital baseband signal; the feature set includes sequence features, instantaneous features, statistical features, spectrum features, and image features;

[0008] The AI ​​acceleration unit is used to obtain the modulation pattern of the communication signal using deep neural network inference according to the feature set.

[0009] The SDR radio frequency front end in the data acquisition unit performs orthogonal sampling on the radio frequency signal in the electromagnetic environment using a broadband A / D converter to achieve sampling of the communication signal within a wide frequency spectrum to obtain I / Q sampling data.

[0010] The signal processing unit comprises:

[0011] A signal preprocessing module, used to perform power spectrum estimation on the I / Q sampled data signal and obtain the total signal strength within the perception bandwidth by power integration;

[0012] A signal detection module is used to detect the unknown signal boundary on the estimated power spectrum and calculate the frequency point estimation and bandwidth estimation of the unknown signal;

[0013] A signal post-processing module, used to perform digital down-conversion processing and digital low-pass filtering processing on the I / Q sampled data signal in sequence according to the frequency and bandwidth of the unknown signal to obtain a digital baseband signal;

[0014] The feature generation module is used to generate a feature set for signal modulation identification based on the digital baseband signal.

[0015] The data preprocessing module comprises:

[0016] A framing part, used for dividing the I / Q sampled data signal into L segments of data, each segment of data including M data;

[0017] The power integration part is used to integrate the power of each segment of data to obtain the total signal strength within the sensing bandwidth;

[0018] The windowing transformation part is used to add the window function to each segment of data and use the fast Fourier transform to obtain the periodogram of each segment of data;

[0019] The averaging part is used to calculate the average value according to the periodogram of each segment of data to perform power spectrum estimation and obtain the estimated power spectrum.

[0020] The signal detection module comprises:

[0021] The difference part is used to perform piecewise difference on the estimated power spectrum to obtain a number of difference result vectors;

[0022] A sorting part, used for sorting the plurality of differential result vectors;

[0023] The estimation part is used to perform estimation calculation according to the index corresponding to the maximum value and the index corresponding to the second largest value in the sorting result, and the sampling frequency of the data acquisition unit to obtain the frequency point estimation and bandwidth estimation of the unknown signal.

[0024] The estimation part is Calculate the frequency estimate by Compute the bandwidth estimate where, is the frequency estimation, is the bandwidth estimation, d1 is the index corresponding to the maximum value in the sorting result, d2 is the index corresponding to the second largest value in the sorting result, and f s is the sampling frequency of the data acquisition unit, and M is the number of data in each segment.

[0025] The feature generation module comprises:

[0026] A sequence feature generating part, used for converting the digital baseband signal into a matrix form to obtain a sequence feature;

[0027] An instantaneous feature generating part, used for determining the instantaneous amplitude and instantaneous phase according to the digital baseband signal, and obtaining the instantaneous frequency based on the instantaneous phase, and obtaining the instantaneous feature according to the instantaneous amplitude and instantaneous frequency, wherein the instantaneous feature includes the instantaneous amplitude standard deviation, the standard deviation of the instantaneous amplitude absolute value, the instantaneous amplitude kurtosis, the standard deviation of the instantaneous frequency absolute value and the instantaneous frequency kurtosis;

[0028] A statistical feature generating section, used for performing high-order moment calculation on the digital baseband signal to obtain high-order cumulants, and selecting a preset number of the high-order cumulants as statistical features;

[0029] A spectrum feature generating part, used for performing Fourier transform on the digital baseband signal to obtain a spectrum diagram, and obtaining spectrum features according to the spectrum diagram, wherein the spectrum features include the maximum value of instantaneous amplitude power spectrum density, spectrum symmetry, and local maximum value of amplitude spectrum;

[0030] The image feature generating part is used to map the digital baseband signal into scattering points on a complex plane to obtain a constellation diagram, perform time-frequency transformation on the digital baseband signal to obtain a time-frequency diagram, and use the constellation diagram and the time-frequency diagram as image features.

[0031] The deep neural network comprises:

[0032] The RNN part is used to extract the sequence features to obtain the first feature;

[0033] A DNN part is used to extract the instantaneous features, statistical features and spectral features to obtain a second feature;

[0034] The CNN part is used to extract the image features to obtain the third feature;

[0035] A splicing part, used for splicing the first feature, the second feature and the third feature according to the channel to obtain a fusion feature;

[0036] The attention layer part is used to perform channel attention mechanism on the fusion features to simplify and select features to obtain the final features;

[0037] The classification part is used to obtain the modulation pattern of the communication signal according to the final characteristics.

[0038] The electromagnetic sensing device for communication signals also includes a central control unit, which includes a memory and a microcontroller. The memory is used to store the I / Q sampling data signal, the processing result of the signal processing unit and the modulation style of the communication signal; the microcontroller is respectively connected to the data acquisition unit, the signal processing unit and the AI ​​acceleration unit, and is used to coordinate the normal operation of the data acquisition unit, the signal processing unit and the AI ​​acceleration unit.

[0039] Beneficial Effects

[0040] Due to the adoption of the above-mentioned technical scheme, the present invention has the following advantages and positive effects compared with the prior art: the present invention uses a lightweight detection algorithm to quickly obtain the boundary of the target signal in the frequency domain, so that the frequency point estimation and bandwidth estimation of the signal can be obtained in real time, and at the same time, the channel attention mechanism is applied to simplify feature extraction and selection, and its computational complexity is reduced to achieve the purpose of real-time identification of signal modulation mode. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the structure of an electromagnetic sensing device for communication signals according to an embodiment of the present invention;

[0042] Figure 2 is a schematic diagram of the structure of a central control unit in an embodiment of the present invention;

[0043] Figure 3 is a structural schematic diagram of a data acquisition unit in an embodiment of the present invention;

[0044] Figure 4 is a schematic structural diagram of a signal processing unit in an embodiment of the present invention;

[0045] Figure 5 is a schematic diagram of the structure of a signal preprocessing module in an embodiment of the present invention;

[0046] Figure 6 is a schematic diagram of the structure of a signal detection module in an embodiment of the present invention;

[0047] Figure 7 is a schematic diagram of the structure of a signal post-processing module in an embodiment of the present invention;

[0048] Figure 8 is a schematic diagram of the structure of a feature generation module in an embodiment of the present invention;

[0049] Fig. 9 Schematic diagram of the structure of the AI ​​acceleration unit in an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.

[0051] The embodiments of the present invention relate to an electromagnetic sensing device for communication signals, which can detect the strength, frequency, and bandwidth of communication signals and identify the modulation pattern of communication signals, such as Figure 1 As shown, it includes a central control unit, a data acquisition unit, a signal processing unit and an AI acceleration unit.

[0052] As the core component of the electromagnetic sensing device, the central control unit is responsible for controlling and coordinating the task execution of the data acquisition unit, signal processing unit and AI acceleration unit. Figure 2 As shown, the central control unit includes a memory and a microcontroller, with the microcontroller as the core, and its functions include parsing and reading the signal detection program, modulation identification program and basic configuration information in the memory, issuing tasks to the data acquisition unit, signal processing unit and AI acceleration unit, and coordinating the normal operation of the acquisition unit, signal processing unit and AI acceleration unit. The memory is responsible for storing the I / Q sampling data obtained by the data acquisition unit, the processing results obtained by the signal processing unit and the modulation pattern of the communication signal obtained by the AI ​​acceleration unit.

[0053] like Figure 3 As shown, the data acquisition unit in this embodiment uses a broadband A / D converter to perform orthogonal sampling on the radio frequency signal in the electromagnetic environment through its own SDR radio frequency front end, so as to achieve sampling of the communication signal in a wide frequency spectrum range and obtain I / Q sampling data. The obtained I / Q sampling data is sent to the signal processing unit to complete the detection and processing of the communication signal in the wide frequency spectrum range, and is also sent to the memory of the central controller for storage.

[0054] like Figure 4 As shown, the signal processing unit in this embodiment mainly includes a signal preprocessing module, a signal detection module, a signal post-processing module and a feature generation module, which can be implemented on a general digital signal processing chip or a dedicated integrated circuit.

[0055] The signal preprocessing module is used to preprocess the I / Q sampling data signal obtained by the aforementioned data acquisition unit. The preprocessing can use periodic spectrum estimation or indirect power spectrum estimation, and obtain the total signal strength within the perception bandwidth through analog or digital power integration; the signal detection module is used to detect the unknown signal boundary of the estimated power spectrum, and calculate the frequency point estimation and bandwidth estimation of the unknown signal, wherein the detection method can use differential detection; the signal post-processing module is used to perform digital down-conversion processing and digital low-pass filtering processing on the I / Q sampling data signal according to the frequency point estimation and bandwidth estimation of the signal obtained by the aforementioned signal detection module to obtain a digital baseband signal; the feature generation module is used to transform and represent the digital baseband signal obtained by the post-processing signal module to generate a feature set for signal modulation identification, which includes sequence features, instantaneous features, statistical features, spectrum features and image features, and sending it to the AI ​​acceleration unit can use deep neural networks for modulation identification. The intensity estimation, frequency point estimation, bandwidth estimation and modulation style results obtained by the aforementioned modules are all sent to the memory of the central controller for storage.

[0056] like Figure 5 As shown, the data preprocessing module in this embodiment includes: a framing part, which is used to divide the I / Q sampling data signal into L segments of data, each segment of data includes M data; a power integration part, which is used to perform power integration on each segment of data to obtain the total signal strength within the perception bandwidth; a windowing transformation part, which is used to add a window function to each segment of data, and use fast Fourier transform to obtain the periodogram of each segment of data; an averaging part, which is used to calculate the average value according to the periodogram of each segment of data to perform power spectrum estimation and obtain an estimated power spectrum. The specific processing process is as follows:

[0057] The signal preprocessing module takes the I / Q sampled data signal as input, and its purpose is to obtain the signal strength RSSI. The I / Q sampled data signal is represented by a complex signal x(n)=i(n)+jq(n), n=0,1,...,N-1, where n is the discrete sampling time and N is the total number of sampling points. First, the signal is framed by the framing part, that is, it is divided into L segments, each segment has M data, and the i-th segment data is represented as:

[0058] x i (n)=x(n+iM-M),0≤n≤M-1,1≤i≤L

[0059] The power integration part is used to integrate the power of each segment of data to obtain the total signal strength within the perception bandwidth. The calculation method is as follows:

[0060]

[0061] Then the windowing and transformation part performs windowing and transformation, that is, adding the window function ω(n) to each data segment and using Fast Fourier Transformation (FFT) to find the periodogram of each segment. The periodogram of the i-th segment is:

[0062]

[0063] In the formula, U is called the normalization factor, which is expressed as:

[0064]

[0065] Finally, the periodograms of each segment are approximately regarded as unrelated, and the average value is calculated from the average part to obtain the power spectrum estimate:

[0066]

[0067] Assuming that the commonly used FFT point number is M′, in this embodiment, M<<M′, so that the calculation complexity of the periodogram in power spectrum estimation can be reduced with fewer transformation points (ie, sampling points), effectively improving the real-time performance of signal processing.

[0068] like Figure 6 As shown, the signal detection module in this embodiment includes: a differential part, which is used to perform segmented differentiation on the estimated power spectrum to obtain a number of differential result vectors; a sorting part, which is used to sort the several differential result vectors; an estimation part, which is used to perform estimation calculation based on the index corresponding to the maximum value and the index corresponding to the second largest value in the sorting result, and the sampling frequency of the data acquisition unit, to obtain the frequency point estimation and bandwidth estimation of the unknown signal.

[0069] The signal detection module in this embodiment is based on the estimated power spectrum Use lightweight detection algorithms to quickly obtain the boundary of the target signal in the frequency domain, thereby obtaining the frequency point estimation of the signal in real time and bandwidth estimation The estimated power spectrum Perform segmented differentiation and estimate the center frequency and bandwidth of the position signal based on the differential vector. The specific process is as follows, the segment length is α, and the differential result vector obtained by the differential part is expressed as:

[0070]

[0071] The differential result vector is sorted by the sorting part, and the maximum value is S1, the corresponding index is d1 (1≤d1≤M), the second largest value is S2, the corresponding index is d2 (1≤d2≤M), and the sampling frequency of the A / D converter in the SDR RF front end is f s, the estimation part performs the following frequency estimation calculation and bandwidth estimation calculation:

[0072]

[0073]

[0074] Since M<<M′, then at the same sampling frequency f s Next, we have: That is, the spectrum resolution is low, which will affect the detection accuracy to a certain extent, but the smaller M will reduce the computational complexity of differentiation and sorting, and improve the real-time performance of signal detection.

[0075] like Figure 7 As shown, the signal post-processing module in this embodiment takes the I / Q sampled data signal and the frequency point estimation and bandwidth estimation obtained by the signal detection module as input, and aims to obtain the baseband I / Q sampled data signal (i.e., baseband digital signal) after down-conversion and low-pass filtering. Assume that the I / Q sampled data signal can be expressed as Frequency estimation based on signal detection After digital down-conversion, we get Bandwidth estimation based on signal detection The cut-off frequency of the adjustable digital low-pass filter is changed, and finally the baseband I / Q sampling data signal required by the feature generation module is obtained.

[0076] like Figure 8 As shown, the feature generation module in this embodiment includes: a sequence feature generation part, which is used to convert the digital baseband signal into a matrix form to obtain a sequence feature; an instantaneous feature generation part, which is used to determine the instantaneous amplitude and instantaneous phase according to the digital baseband signal, and obtain the instantaneous frequency based on the instantaneous phase, and obtain the instantaneous feature according to the instantaneous amplitude and instantaneous frequency, wherein the instantaneous feature includes the instantaneous amplitude standard deviation, the standard deviation of the instantaneous amplitude absolute value, the instantaneous amplitude kurtosis, the standard deviation of the instantaneous frequency absolute value and the instantaneous frequency kurtosis; a statistical feature generation part, which is used to perform a statistical analysis on the digital baseband signal. A high-order moment calculation is performed to obtain a high-order cumulant, and a preset number of the high-order cumulants are selected as statistical features; a spectrum feature generating part is used to perform Fourier transform on the digital baseband signal to obtain a spectrum diagram, and obtain spectrum features according to the spectrum diagram, wherein the spectrum features include the maximum value of the instantaneous amplitude power spectrum density, the spectrum symmetry, and the local maximum value of the amplitude spectrum; an image feature generating part is used to map the digital baseband signal to scattering points on a complex plane to obtain a constellation diagram, perform time-frequency transform on the digital baseband signal to obtain a time-frequency diagram, and use the constellation diagram and the time-frequency diagram as image features.

[0077] The feature generation module in this embodiment uses the baseband I / Q sampling data signal obtained by signal post-processing as input, with the purpose of generating a feature set for modulation recognition by the AI ​​acceleration unit. The selected feature set includes sequence features, instantaneous features, statistical features, spectral features and image features. The generation of various features is further explained below.

[0078] The sequence feature is the baseband I / Q sampling data signal, which is written in matrix form:

[0079]

[0080] The generation of instantaneous features requires the calculation of the normalized instantaneous amplitude:

[0081]

[0082] The instantaneous phase is:

[0083]

[0084] Centralized instantaneous frequency:

[0085]

[0086] The following instantaneous features are designed based on the instantaneous amplitude and instantaneous frequency:

[0087] The standard deviation of the instantaneous amplitude is: The standard deviation of the absolute value of the instantaneous amplitude is: The standard deviation of the absolute value of the instantaneous frequency is: The instantaneous amplitude kurtosis is: The instantaneous frequency kurtosis is: Where E(x) represents the mean of the random variable x.

[0088] The statistical characteristics are the high-order cumulants of the signal calculated by the high-order moments of the signal:

[0089]

[0090]

[0091] A sample calculation is as follows:

[0092]

[0093] The high-order cumulants used in this embodiment include C 20 ,C 21 ,C 40 ,C 41 ,C 42 ,C 60 ,C 61 ,C62 ,C 63 .

[0094] The spectrum characteristics are obtained based on the Fourier transform of the signal. The maximum value of the instantaneous amplitude power spectrum density is: max =max(|DFT(A(n))| 2 ); To measure whether the spectrum is symmetrical or asymmetrical, the spectrum symmetry is: In order to make full use of the amplitude spectrum, an operation Find(x) is defined to find the three local maxima of the amplitude spectrum: Here, DFT(x) represents the discrete Fourier transform of x.

[0095] The image features include constellation diagram and time-frequency diagram. The baseband I / Q sampling data signal is mapped to scattering points on the complex plane to obtain the constellation diagram CSD (Constellation Diagram), and the baseband I / Q sampling data signal is subjected to time-frequency transformation to obtain the time-frequency image TFI (Time-Frequency Image). Candidate time-frequency analysis methods can be short-time Fourier transform (STFT), Wigner-Vieri distribution (WVD) and Cui-Williams distribution (CWD).

[0096] like Fig. 9 As shown, the modulation recognition module in the AI ​​acceleration unit in this embodiment uses the parallel computing accelerator to recognize the modulation pattern according to the multiple features generated by the feature generation module, that is, a function f is used to map the signal x(n) to the corresponding modulation pattern y, y∈{y1,…,y k The deep neural network used by the AI ​​acceleration unit includes: an RNN part, which is used to extract features from the sequence features to obtain a first feature; a DNN part, which is used to extract features from the instantaneous features, statistical features, and spectral features to obtain a second feature; a CNN part, which is used to extract features from the image features to obtain a third feature; a splicing part, which is used to splice the first feature, the second feature, and the third feature according to the channel to obtain a fusion feature; an attention layer part, which is used to perform feature simplification and selection on the fusion feature using a channel attention mechanism to obtain a final feature; and a classification part, which is used to obtain a modulation pattern of a communication signal according to the final feature.

[0097] The deep neural network uses multi-dimensional features to fully characterize the original signal, and uses the channel attention mechanism to simplify feature extraction and selection, thereby simplifying the recognition network and reducing its computational complexity to achieve real-time recognition. First, the sequence features are sent to the RNN part to obtain the first feature F1, the instantaneous features, statistical features, and spectral features are sent to the DNN part to obtain the second feature F2, and the image features are sent to the CNN part to obtain the third feature F3. Then, the first feature F1, the second feature F2, and the third feature F3 are concatenated by channel to obtain the fusion feature F, that is, F = concatenate (F1, F2, F3), and then the attention layer part executes the attention mechanism to ignore redundant features and emphasize important features to obtain the final feature F A , and finally the final feature F A The signal is sent to the classification part to obtain the modulation pattern of the signal.

[0098] It is not difficult to find that the present invention uses a lightweight detection algorithm to quickly obtain the boundary of the target signal in the frequency domain, so that the frequency point estimation and bandwidth estimation of the signal can be obtained in real time. At the same time, the channel attention mechanism is applied to simplify feature extraction and selection, and its computational complexity is reduced to achieve the purpose of real-time identification of signal modulation mode.

Claims

1. An electromagnetic sensing device for communication signals, It is characterized in that include: A data acquisition unit, used for acquiring I / Q sampling data signals of communication signals in real time; A signal processing unit, configured to perform signal processing on the I / Q sampled data signal to obtain a digital baseband signal, and to obtain a feature set for signal modulation identification based on the digital baseband signal; the feature set includes sequence features, instantaneous features, statistical features, spectrum features, and image features; The AI ​​acceleration unit is used to obtain the modulation pattern of the communication signal using deep neural network inference according to the feature set.

2. The electromagnetic sensing device for communication signals according to claim 1, It is characterized in that The SDR radio frequency front end in the data acquisition unit performs orthogonal sampling on the radio frequency signal in the electromagnetic environment using a broadband A / D converter to achieve sampling of the communication signal within a wide frequency spectrum to obtain I / Q sampling data.

3. The electromagnetic sensing device for communication signals according to claim 1, It is characterized in that The signal processing unit comprises: A signal preprocessing module, used to perform power spectrum estimation on the I / Q sampled data signal and obtain the total signal strength within the perception bandwidth by power integration; A signal detection module is used to detect the unknown signal boundary on the estimated power spectrum and calculate the frequency point estimation and bandwidth estimation of the unknown signal; A signal post-processing module, used for performing digital down-conversion processing and digital low-pass filtering processing on the I / Q sampled data signal in sequence according to the frequency point estimation and bandwidth estimation of the unknown signal to obtain a digital baseband signal; The feature generation module is used to generate a feature set for signal modulation identification based on the digital baseband signal.

4. The electromagnetic sensing device for communication signals according to claim 3, It is characterized in that The data preprocessing module comprises: A framing part, used for dividing the I / Q sampled data signal into L segments of data, each segment of data including M data; The power integration part is used to integrate the power of each segment of data to obtain the total signal strength within the sensing bandwidth; The windowing transformation part is used to add the window function to each segment of data and use the fast Fourier transform to obtain the periodogram of each segment of data; The averaging part is used to calculate the average value according to the periodogram of each segment of data to perform power spectrum estimation and obtain the estimated power spectrum.

5. The electromagnetic sensing device for communication signals according to claim 3, It is characterized in that The signal detection module comprises: The difference part is used to perform piecewise difference on the estimated power spectrum to obtain a number of difference result vectors; A sorting part, used for sorting the plurality of differential result vectors; The estimation part is used to perform estimation calculation according to the index corresponding to the maximum value and the index corresponding to the second largest value in the sorting result, and the sampling frequency of the data acquisition unit to obtain the frequency point estimation and bandwidth estimation of the unknown signal.

6. The electromagnetic sensing device for communication signals according to claim 5, It is characterized in that The estimation part is Calculate the frequency estimate by Compute the bandwidth estimate where, is the frequency estimation, is the bandwidth estimation, d 1 is the index corresponding to the maximum value in the sorted result, d 2 is the index corresponding to the second largest value in the sorted result, f s is the sampling frequency of the data acquisition unit, and M is the number of data in each segment.

7. The electromagnetic sensing device for communication signals according to claim 3, It is characterized in that The feature generation module comprises: A sequence feature generating part, used for converting the digital baseband signal into a matrix form to obtain a sequence feature; An instantaneous feature generating part, used for determining the instantaneous amplitude and instantaneous phase according to the digital baseband signal, and obtaining the instantaneous frequency based on the instantaneous phase, and obtaining the instantaneous feature according to the instantaneous amplitude and instantaneous frequency, wherein the instantaneous feature includes the instantaneous amplitude standard deviation, the standard deviation of the instantaneous amplitude absolute value, the instantaneous amplitude kurtosis, the standard deviation of the instantaneous frequency absolute value and the instantaneous frequency kurtosis; A statistical feature generating section, used for performing high-order moment calculation on the digital baseband signal to obtain high-order cumulants, and selecting a preset number of the high-order cumulants as statistical features; A spectrum feature generating part, used for performing Fourier transform on the digital baseband signal to obtain a spectrum diagram, and obtaining spectrum features according to the spectrum diagram, wherein the spectrum features include the maximum value of instantaneous amplitude power spectrum density, spectrum symmetry, and local maximum value of amplitude spectrum; The image feature generating part is used to map the digital baseband signal into scattering points on a complex plane to obtain a constellation diagram, perform time-frequency transformation on the digital baseband signal to obtain a time-frequency diagram, and use the constellation diagram and the time-frequency diagram as image features.

8. The electromagnetic sensing device for communication signals according to claim 1, It is characterized in that The deep neural network comprises: The RNN part is used to extract the sequence features to obtain the first feature; The DNN part is used to extract the instantaneous features, statistical features and spectral features to obtain the second features; the CNN part is used to extract the image features to obtain the third features; A splicing part, used for splicing the first feature, the second feature and the third feature according to the channel to obtain a fusion feature; The attention layer part is used to perform channel attention mechanism on the fusion features to simplify and select features to obtain the final features; The classification part is used to obtain the modulation pattern of the communication signal according to the final characteristics.

9. The electromagnetic sensing device for communication signals according to claim 1, It is characterized in that It also includes a central control unit, which includes a memory and a microcontroller. The memory is used to store the I / Q sampling data signal, the processing result of the signal processing unit and the modulation style of the communication signal; the microcontroller is respectively connected to the data acquisition unit, the signal processing unit and the AI ​​acceleration unit, and is used to coordinate the normal operation of the data acquisition unit, the signal processing unit and the AI ​​acceleration unit.

Citation Information

Cited By

  • Specific system OFDM signal detection and identification method based on multi-domain feature fusion

    CN121056285A

  • A specific type of OFDM signal detection and identification method based on multi-domain feature fusion

    CN121056285B