Communication anti-interference method based on space-time-frequency domain combination

Abstract

The invention discloses a communication anti-interference method based on space-time-frequency domain combination. The method comprises the following steps: S1, preprocessing an obtained signal; s2, for the preprocessed signal, acquiring a time domain waveform on a time domain, acquiring a time-frequency diagram on a frequency domain, and acquiring a direction of arrival on a space domain; s3, feature extraction is performed on the time domain waveform and the time frequency graph based on a multi-domain feature fusion network, high-dimensional feature vectors of time domain features and frequency domain features are obtained, feature fusion is performed on the high-dimensional feature vectors of the time domain and the frequency domain, and a summed feature vector is obtained; and S4, judging the interference type according to the summed feature vector. According to the method, the limitation that an existing anti-interference technology only aims at a single dimension is solved, multi-domain cooperative anti-interference is achieved by combining feature extraction and interference recognition of the space dimension, the time dimension and the frequency dimension, and the anti-interference performance of an ad hoc network terminal is improved so as to adapt to the constantly changing interference environment.

Description

Technical Field

[0001] The present invention relates to the technical field of communication anti-jamming, and specifically to a communication anti-jamming method based on joint space-time-frequency domain. Background Art

[0002] With the rapid development of wireless communication technology, communication systems have been widely applied in military, civilian, commercial and other fields. These systems usually need to be quickly deployed and achieve communication between nodes without the support of fixed infrastructure. However, due to the openness of the wireless channel, these systems are vulnerable to various interferences during information transmission, including natural interferences (such as lightning, sunspot activities, etc.) and man-made interferences (such as hostile signals, electromagnetic compatibility problems, etc.). These interferences can lead to a decline in signal quality and even communication interruption, seriously affecting the reliability and stability of the communication system.

[0003] To improve the anti-jamming ability of communication systems, researchers have developed a variety of technologies. These technologies can be mainly divided into two categories: time-domain anti-jamming technologies and frequency-domain anti-jamming technologies. Time-domain anti-jamming technologies mainly avoid interference by adjusting the signal transmission time. For example, the time-hopping (TH) technology sends signals at different time intervals, making it difficult for interferers to predict the signal transmission moment, thereby reducing the impact of interference. However, the TH technology is not ideal when facing continuous broadband interference because it cannot be adjusted in frequency. Frequency-domain anti-jamming technologies reduce interference by changing the signal frequency. The frequency-hopping (FH) technology allows the signal to quickly switch between multiple frequencies to avoid interference at specific frequencies. This technology is very effective in combating frequency-selective interference, but it is equally powerless when facing time-based interference. Although these technologies have improved the anti-jamming ability of communication systems to a certain extent, they usually only consider interference in a single dimension, while interference in the real environment is often multi-dimensional. For example, an interference signal may affect the communication system both in time and frequency. Therefore, existing single-dimensional anti-jamming technologies often struggle to provide comprehensive protection against complex space-time-frequency domain interference.

[0004] The implementation scheme closest to the present invention is the two-dimensional anti-jamming technology that combines time and frequency domain technologies. This technology attempts to improve the anti-jamming ability by adjusting both in time and frequency simultaneously. For example, some studies have proposed an adaptive frequency-hopping technology based on OFDM (orthogonal frequency division multiplexing), which can achieve frequency hopping on OFDM subcarriers and adjust in time simultaneously. However, these technologies still have limitations. They do not handle the spatial dimension sufficiently and cannot fully utilize spatial diversity to improve the anti-jamming ability. Summary of the Invention

[0005] The object of the present invention is to propose a communication anti-jamming method based on joint space-time-frequency domain, to solve the limitation of existing anti-jamming technologies that only target a single dimension, and to achieve multi-domain collaborative anti-jamming by jointly extracting features and identifying interference in the three dimensions of space, time, and frequency, so as to improve the anti-jamming performance of self-organizing network terminals and adapt to the changing interference environment.

[0006] The technical solution adopted by the present invention is as follows: A communication anti-jamming method based on joint space-time-frequency domain, comprising the following steps:

[0007] S1. Preprocess the acquired signal;

[0008] S2. For the preprocessed signal, obtain the time-domain waveform in the time domain, the time-frequency diagram in the frequency domain, and the direction of arrival in the space domain;

[0009] S3. Based on the multi-domain feature fusion network, respectively extract features from the time-domain waveform and the time-frequency diagram to obtain high-dimensional feature vectors of time-domain features and frequency-domain features, and fuse the high-dimensional feature vectors of the time domain and the frequency domain to obtain a summed feature vector;

[0010] S4. Judge the interference type according to the summed feature vector:

[0011] If the signal is oversaturated in the time domain and the frequency domain and cannot communicate, it is determined that the interference type is global or approximately global suppression interference, then determine the direction of the incoming wave based on the signal direction of arrival estimation coefficient, and use null steering in the direction of the incoming wave to suppress the interference direction signal; if the signal is not oversaturated, the following judgment is made:

[0012] Judge whether the interference type is fixed-frequency narrowband interference. If so, use windowed overlapping frequency domain suppression or hopping frequency point elimination; or

[0013] Judge whether the interference type is frequency-domain dynamic interference. If so, use Turbo code interleaved coding for anti-jamming.

[0014] As a preferred solution, the preprocessing includes the following sub-steps:

[0015] Use the mean method to centralize the data so that the mean of the data is 0; design a filter according to the interference bandwidth of the signal and denoise the signal; perform power normalization so that the average power of the normalized data is 1.

[0016] As a preferred solution, the time-domain features include kurtosis coefficient and peak-to-average ratio coefficient.

[0017] As a preferred solution, the frequency-domain features include average spectral flatness coefficient, broadband factor, and narrowband interference detection coefficient.

[0018] As a preferred solution, the feature extraction includes the following steps:

[0019] Extract the underlying features from the received time-domain waveform or time-frequency diagram through the DenseNet network; alternately stack the dense connection modules and transition layers to deeply extract high-dimensional features; use the global average pooling layer and the fully connected layer to reduce the dimension and further integrate the features to obtain a high-dimensional feature vector.

[0020] As a preferred solution, a one-dimensional DenseNet network is used for extracting time-domain features, and a two-dimensional DenseNet network is used for extracting frequency-domain features.

[0021] As a preferred solution, feature fusion includes adding the obtained high-dimensional feature vectors element by element to obtain a new feature vector, and then passing it into the fully connected layer and the Softmax classification layer for processing.

[0022] As a preferred solution, the overlapping-add output after windowing includes the following steps: perform windowing on the time-domain data to reduce spectral leakage; perform FFT transformation and interference frequency suppression on the windowed signal; restore the time-domain data through overlapping addition.

[0023] As a preferred solution, nulling suppression includes adjusting the antenna coefficients to form a null in the direction of arrival of the interference signal to suppress the interference.

[0024] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0025] This method provides a comprehensive anti-jamming system that combines the space, time, and frequency domains. From the perspective of anti-jamming technology, this method not only considers the anti-jamming measures in time and frequency, but also introduces the processing in the spatial dimension. The kurtosis coefficient and peak-to-average ratio coefficient are extracted in the time domain; the average spectral flatness coefficient, broadband factor, and narrowband interference detection coefficient are extracted in the frequency domain; the direction-of-arrival estimation coefficient of the signal is extracted in the spatial domain. The extracted feature parameters are used for interference identification. From the perspective of the anti-jamming strategy, considering the threat degree of specific interference patterns, the priority of the multi-domain anti-jamming strategy is designed, that is, after interference identification, anti-jamming processing is performed in the time domain, frequency domain, and spatial domain respectively according to the type of interference signal, which can improve the anti-jamming performance of the ad-hoc network terminal to adapt to the changing interference environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0027] Figure 1 It is a schematic diagram of the overall process of the present invention;

[0028] Figure 2 Schematic diagram of feature extraction and feature fusion of the present invention;

[0029] Figure 3 Schematic diagram of interference type judgment of the present invention;

[0030] Figure 4 Schematic diagram of the airspace anti-interference process of the present invention;

[0031] Figure 5 Schematic diagram of the frequency domain anti-interference process of the present invention. Detailed implementation manners

[0032] Next, the present invention will be specifically described through exemplary implementation manners. However, it should be understood that, without further description, the elements, structures and features in one implementation manner can also be beneficially combined into other implementation manners.

[0033] It should be noted that: unless otherwise defined, the technical terms or scientific terms used herein should have the ordinary meanings understood by those of ordinary skill in the field to which the present invention belongs. The words such as "a", "an" or "the" used in the specification and claims of this patent application for the present invention do not express a limitation of quantity, but mean that there is at least one; the "first", "second" and "third" used herein should not be regarded as a limitation on the order of components, but are only used to distinguish different components; the words such as "comprising" or "including" indicate that the elements or objects appearing before "comprising" or "including" cover the elements or objects listed after "comprising" or "including" and their equivalents, but do not exclude other elements or objects with the same functions.

[0034] In order to more clearly describe the specific manner of the communication anti-interference method based on joint space-time-frequency domain, in combination with the attached Figures 1-5 Describe this embodiment:

[0035] As Figure 1 and Figure 2 shown, the communication anti-interference method based on joint space-time-frequency domain includes the following steps:

[0036] S1. Preprocess the acquired signal to improve the stability of feature extraction;

[0037] The preprocessing includes the following sub-steps:

[0038] First, centralize the data using the mean method to make the mean of the data 0; then design a filter according to the interference bandwidth of the signal, filter and denoise the signal to reduce the influence of noise and clutter; finally, perform power normalization to make the average power of the normalized data 1.

[0039] S2. For the preprocessed signal, obtain the time-domain waveform in the time domain, the time-frequency diagram in the frequency domain, and the direction of arrival in the spatial domain;

[0040] Among them, the direction of arrival of the signal is determined by the direction-of-arrival estimation coefficient of the signal, and this parameter can be used to identify and suppress multipath effects and directional interference.

[0041] S3. Based on the multi-domain feature fusion network, extract features from the time-domain waveform and the time-frequency diagram respectively, obtain the high-dimensional feature vectors of the time-domain features and the frequency-domain features, and fuse the high-dimensional feature vectors of the time domain and the frequency domain to obtain the summed feature vector;

[0042] Feature extraction is to extract the time-domain, frequency-domain, and transform-domain features of the data. In the time domain, the kurtosis coefficient and the peak-to-average ratio coefficient are extracted; the kurtosis coefficient is used to measure the kurtosis of the signal, that is, the sharpness of the signal peak, while the peak-to-average ratio coefficient measures the fluctuation degree of the signal. These two parameters are particularly important for identifying impulse interference and burst interference. By monitoring the changes of these parameters in real time, the system can quickly identify the interference in the time domain and take corresponding anti-interference measures. In the frequency domain, the average spectral flatness coefficient, the broadband factor, and the narrowband interference detection coefficient are extracted to identify and quantify the frequency-domain interference. The average spectral flatness coefficient reflects the uniformity of the signal spectrum, the broadband factor measures the bandwidth occupancy of the signal, and the narrowband interference detection coefficient is specifically used to identify narrowband interference in the frequency domain. The extraction and analysis of these parameters enable the system to effectively identify and suppress the frequency-domain interference.

[0043] After feature extraction, the interference signal type is determined by the interference recognition domain algorithm decision joint multi-dimensional feature discrimination, and the corresponding anti-interference processing algorithm is controlled to be selected. By analyzing the characteristic parameters in the time domain, frequency domain, and spatial domain, the presence and type of the interference signal are identified. Anti-interference processing will be carried out in the corresponding dimension according to the type and characteristics of the interference.

[0044] Refer to Figure 2, in order to improve the recognition ability of time-domain and frequency-domain features, this solution uses a Multi-Domain Feature Fusion Network (MDFFNet) to synchronously extract the time-domain and frequency-domain features of the signal. MDFFNet takes the time-domain waveform and time-frequency diagram of the received signal as input sources, extracts features respectively and fuses them, and then completes the recognition. MDFFNet consists of three modules: a time-domain waveform feature extraction module, a frequency-domain feature extraction module, and a feature fusion module. The time-domain waveform feature extraction module takes the time-domain waveform of the received signal as input and uses one-dimensional DenseNet to extract high-dimensional abstract features. The time-frequency diagram feature extraction module receives the time-frequency diagram of the signal and extracts high-dimensional abstract features with the help of two-dimensional DenseNet. Both modules use Dense Blocks (DBs) to enhance feature reuse and improve generalization performance. The feature fusion module is responsible for fusing time-frequency features and time-domain features, and further learning and integrating the feature representation through a fusion mechanism to complete the final recognition prediction.

[0045] Feature extraction includes time-domain feature extraction and frequency-domain feature extraction:

[0046] Time-domain feature extraction: It is composed of the alternating superposition of a one-dimensional convolutional layer, a one-dimensional Dense Block (1D-DB) module, and a one-dimensional transition layer. The one-dimensional convolutional layer and 1D-DB are responsible for extracting the underlying features and high-dimensional features of the time-domain data respectively. The one-dimensional convolutional layer takes the co-directional orthogonal IQ signal as input, where L in is the length of the input signal, C in is the number of channels (the number of channels of the IQ signal is 2), and its feature extraction process can be expressed as:

[0047] x′ T =F 1D-Conv (x T , ψ)

[0048] where F 1D-Conv represents the one-dimensional convolution operation process, ψ is the training parameter of the one-dimensional convolutional layer, represents the output with a length of L1 and a number of channels of C1 (i.e., the number of convolution kernels). The time-domain branch deeply extracts the high-level features of the IQ signal by alternately stacking 1D-DB modules and one-dimensional transition layers. After the alternating 1D-DB and transition layers, a global average pooling layer (GAP) and a fully connected layer are used to reduce the dimension and further integrate the features. GAP integrates global information, reduces the dimensionality of the feature map, and reduces parameters to prevent overfitting. The fully connected layer then integrates the dimensionality-reduced feature vectors to obtain the high-level feature vectors of the time-domain waveform data.

[0049] Frequency-domain feature extraction: It consists of a two-dimensional convolutional layer, a two-dimensional dense connection block (2D-DB), and a two-dimensional transition layer. The primary task of the convolutional layer is to extract the underlying features of the input time-frequency diagram. Given a time-frequency diagram as the input, where W in ×H in is the size of the input image, and C in is the number of channels. A two-dimensional convolutional layer is constructed to extract the embedded features of x TF , and its form can be expressed as:

[0050]

[0051] where F 2D-Conv represents the two-dimensional convolution operation process, are the training parameters of the convolutional layer, is the output feature map with a size of W1×H1 and the number of channels C1 (i.e., the number of convolutional kernels). The feature reuse effect of a single convolutional layer is not good, it is prone to overfitting, and the recognition accuracy is limited. Therefore, multiple 2D-DB modules are introduced after the convolutional layer to strengthen feature transfer, improve the performance of the model, and streamline the parameters. Each 2D-DB module contains multiple densely connected convolutional layers, and each layer takes the outputs of all previous convolutional layers as the input. This connection method is conducive to feature transfer and information flow, enhances feature reuse, and reduces the model parameters. The 2D-DB blocks and the transition layer are used alternately to fully extract the high-level abstract features of the time-frequency diagram. The 2D-DB module reduces the computational amount, and the transition layer controls the size of the feature map. After the last 2D-DB module, a global average pooling layer (GAP) and a fully connected layer are connected to produce a high-dimensional feature vector of the time-frequency data.

[0052] Feature fusion: After obtaining the features of the time-frequency diagram data and the time-domain data, cross-domain information interaction needs to be carried out to achieve feature fusion, and then complete the modulation recognition of the input signal and give the model prediction results, that is, the probabilities of each modulation type. According to whether to perform re-learning on the fused features and the differences in the fusion methods, the feature summation fusion method is adopted. In the figure, and represent the outputs of the time-frequency branch and the time-domain branch respectively. Feature summation fusion adds the feature vectors of the time-frequency diagram and the time-domain feature extraction branch element by element to obtain a new feature vector, and then passes it to the fully connected layer and the Softmax classification layer for processing, so as to obtain the final prediction results. The feature vector after summation is expressed as: where x a represents the feature vector after summation, and ‘+’ represents the element-by-element addition operation.

[0053] S4. Determine the interference type according to the feature vector after summation:

[0054] For general interference signals, if it is impossible to suppress the entire space, time, and frequency domains, the communicating parties can always conduct anti-interference communication through specific time and frequency domain anti-interference technologies. Therefore, the key to joint space-time-frequency anti-interference lies in how to adopt anti-interference technologies in specific domains according to different interference patterns. In anti-interference design, first, judge the interference pattern through interference recognition in the time domain and frequency domain;

[0055] Refer to Figure 3 , if the signal is oversaturated in the time domain and frequency domain and unable to communicate, determine that the interference type is global or approximately global suppression interference, then determine the direction of the incoming wave based on the direction-of-arrival estimation coefficient of the signal, and implement the spatial domain anti-interference strategy, that is, use nulling in the direction of the incoming wave to suppress the signal in the interference direction; if the signal is not oversaturated, make the following judgments:

[0056] Judge whether the interference type is fixed-frequency narrowband interference. If so, implement the frequency domain anti-interference strategy, that is, use windowed overlapping frequency domain suppression or hopping frequency point elimination, and a specific interference-free communication dimension can be found for communication; or

[0057] Judge whether the interference type is frequency domain dynamic interference. If so, implement the time domain anti-interference strategy, that is, use dynamic anti-interference such as Turbo code interleaving coding and efficient anti-random collision.

[0058] Specifically:

[0059] Spatial domain anti-interference strategy: Analyze the signal in the direction of the incoming wave in the spatial domain. If the signal energy is too large and specific data cannot be received, adjust the coefficients of the antenna to null the signal in this direction and suppress the interference. An exemplary processing method is as follows:

[0060] Refer to Figure 4 , use a 2*2 antenna, and the transmitting end performs space-time coding (Space-Time Block Coding, STBC) STBC-Alamouti coding. Encode two consecutive transmitted symbols x1 and x2 according to the space-time codeword matrix to obtain In the first symbol period, the two symbols x1 and x2 are respectively transmitted from two antennas. In the second symbol period, these two symbols are transmitted again, where the first antenna transmits The second antenna transmits The receiving end uses the maximum likelihood detection algorithm to perform diversity reception on the signals received by the two antennas respectively to obtain diversity gain.

[0061] Time domain anti-interference technology: To reduce the impact of time domain burst interference on system performance, Turbo code is used for channel coding, and interleaving is increased, such as Figure 2As shown. Through simulation verification, it is to verify whether the receiving end can receive the original information after k consecutive bits of information are lost or n bits of random interference occur during the signal transmission process. As can be seen from the following table, the anti-random interference performance of Turbo codes is much higher than the anti-consecutive interference ability.

[0062] Table 1: Anti-interference ability of Turbo codes with different code rates and a code length of 512

[0063] Coding rate 1 / 5 1 / 3 2 / 5 1 / 2 Anti-interference ability (random interference) 75％ 58％ 50％ 40％ Anti-interference ability (continuous interference) 3.9％ 3.9％ 4.2％ 4.2％

[0064] Table 2: Anti-interference ability of Turbo codes with different code rates and a code length of 2048

[0065] Coding rate 1 / 5 1 / 3 2 / 5 1 / 2 Anti-interference ability (random interference) 76％ 60％ 53％ 44％ Anti-interference ability (continuous interference) 2.3％ 2.2％ 1.9％ 1.8％

[0066] Frequency domain anti-interference technology: Refer to Figure 5 , by performing FFT transformation on the signal, the spectra of the signal and the interference are obtained to remove the interference frequency, achieving the effect of suppressing interference, and then the time-domain data output is obtained through IFFT transformation. However, the window function of the discrete Fourier transform is a rectangular window, and its sidelobe energy is relatively large, which will cause serious spectral leakage, resulting in an incorrect estimation of the signal spectrum. Therefore, windowing is performed on the signal to make the signal smoother to reduce spectral leakage, but additional signal-to-noise ratio loss will be brought after windowing. Therefore, the present invention adopts a frequency domain interference suppression method of overlapping and adding outputs after windowing. This method effectively reduces the influence of windowing on the signal-to-noise ratio.

[0067] The present invention has been proven feasible through laboratory simulation and on-site testing. The experimental results show that compared with the anti-interference technology of a single dimension, the anti-interference method adopting the present invention can significantly improve the anti-interference ability of the communication system, and the stability and reliability of the signal are significantly enhanced.

[0068] The parts not detailed in the above embodiments are prior art.

[0069] It should be noted that although the present invention has been described through the above embodiments, the present invention can also have many other embodiments. Without departing from the spirit and scope of the present invention, those skilled in the art can obviously make various corresponding changes and deformations to the present invention, but these changes and deformations should all fall within the scope protected by the appended claims of the present invention and their equivalents.

Claims

1. A communication anti-interference method based on the joint space-time-frequency domain, characterized in that: The following steps are involved: S1. Preprocess the acquired signal; S2. For the preprocessed signal, obtain the time domain waveform in the time domain, obtain the time-frequency diagram in the frequency domain, and obtain the direction of arrival in the spatial domain; S3, based on the multi-domain feature fusion network, feature extraction is performed on the time domain waveform and the time-frequency diagram respectively, high-dimensional feature vectors of the time domain features and the frequency domain features are obtained, and the high-dimensional feature vectors of the time domain and the frequency domain are feature-fused to obtain the summed feature vector; S4. Determine the interference type based on the summed eigenvector: If the signal is oversaturated in the time domain and frequency domain and cannot communicate, the interference type is determined to be global or nearly global suppression interference. The direction of arrival is determined based on the signal arrival direction estimation coefficient, and the direction of arrival nulling is used to suppress the interference direction signal. If the signal is not oversaturated, the following judgment is made: Determine whether the interference type is fixed-frequency narrowband interference. If so, use windowed overlapping frequency domain suppression or frequency hopping frequency elimination; or It is determined whether the interference type is frequency domain dynamic interference. If so, Turbo code interleaving encoding is used for anti-interference.

2. The communication anti-interference method based on space-time-frequency domain joint according to claim 1 is characterized in that: The preprocessing consists of the following sub-steps: Use the mean method to center the data so that the mean of the data is 0; Design filters based on the interference bandwidth of the signal and reduce the noise of the signal; Power normalization is performed so that the average power of the normalized data is 1.

3. The communication anti-interference method based on space-time-frequency domain combination according to claim 1 is characterized in that: The time domain features include moment kurtosis coefficient and peak-to-average ratio coefficient.

4. The communication anti-interference method based on space-time-frequency domain combination according to claim 1 is characterized in that: The frequency domain features include an average spectrum flatness coefficient, a broadband factor and a narrowband interference detection coefficient.

5. The communication anti-interference method based on space-time-frequency domain combination according to claim 1 is characterized in that: Feature extraction includes the following steps: The received time domain waveform or time-frequency graph is passed through the DenseNet network to extract the underlying features; Alternately stack densely connected modules and transition layers to deeply extract high-dimensional features; The global average pooling layer and the fully connected layer are used to reduce the dimension of the features and further integrate them to obtain a high-dimensional feature vector.

6. The communication anti-interference method based on space-time-frequency domain combination according to claim 5 is characterized in that: The one-dimensional DenseNet network is used to extract time domain features, and the two-dimensional DenseNet network is used to extract frequency domain features.

7. The communication anti-interference method based on space-time-frequency domain combination according to claim 6 is characterized in that: Feature fusion involves adding the acquired high-dimensional feature vectors element by element to obtain a new feature vector, which is then passed to the fully connected layer and the Softmax classification layer for processing.

8. The communication anti-interference method based on space-time-frequency domain combination according to claim 1 is characterized in that: The windowed overlap-add output includes the following sub-steps: Windowing is performed on the time domain data to reduce spectrum leakage; Perform FFT transformation and interference frequency suppression on the windowed signal; The time domain data is restored by overlap-add.

9. The method according to claim 1, characterized in that: Null suppression involves adjusting the antenna coefficients to form a null in the direction of arrival of the interference signal to suppress the interference.

Images