A method and apparatus for precision interference avoidance in a radio system

Abstract

The application discloses a precise interference avoidance method and device in a radio system, and relates to the field of signal processing, and comprises the following steps: constructing a sensing matrix through sensing node data, and capturing a transient interference signal through aperiodic scanning; signal data is extracted through tensor decomposition to obtain time domain, frequency domain, space domain and modulation domain features, a dual-mode spectrum analysis model is constructed, an instantaneous frequency spectrum diagram is reconstructed by using compressed sensing, and an interference mode is predicted through LSTM; after fusing the instantaneous frequency spectrum diagram and the predicted interference diagram, threat assessment is performed through a multi-level interference classification model; according to the interference category and the threat level, beamforming is optimized, adaptive nulls are generated, a power density optimization model is constructed, and the transmission power is dynamically adjusted; an anti-interference frequency hopping sequence is generated based on a chaotic mapping algorithm, spectrum camouflage and anti-tracking interference are realized. The application has the advantages that through multi-dimensional sensing, intelligent prediction and adaptive beamforming, precise identification and dynamic avoidance of interference are realized, and the interference avoidance capability of the wireless system is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of signal processing, in particular to a precise interference avoidance method and device in a radio system. BACKGROUND

[0002] In modern radio communication systems, the problem of interference is increasingly serious, especially in the environment of limited spectrum resources and frequent sharing, interference has a significant impact on the performance and reliability of the system. In order to improve the anti-interference ability of the radio system, the precise interference avoidance technology emerges as the times require. This kind of technology not only applies to military communication, satellite communication and other high security demand scenes, but also is widely used in 5G and future communication network, and is the key to improve the reliability and stability of the radio system.

[0003] Most of the current radio interference avoidance technologies on the market rely on static spectrum monitoring and simple interference suppression algorithms, which are difficult to cope with the dynamic changes of complex electromagnetic environment, resulting in poor real-time performance and adaptability of the system. Secondly, the traditional method usually ignores the prediction and adaptive adjustment of the interference mode, resulting in inaccurate identification and suppression of the interference source, especially in the case of high-intensity or burst interference, which cannot respond in time and effectively. In addition, the current spectrum allocation and power optimization method of some systems is too simple, and does not fully consider the multi-node cooperation, which may lead to waste of system resources or further aggravation of interference. SUMMARY

[0004] In order to improve the existing method and platform, a precise interference avoidance method and device in a radio system are provided, which realizes precise identification and dynamic avoidance of interference through multi-dimensional perception, intelligent prediction and adaptive beamforming, and improves the anti-interference and communication support ability of the wireless system in complex electromagnetic environment.

[0005] To achieve the above purpose, the technical scheme adopted by the present application is:

[0006] A precise interference avoidance method in a radio system, comprising:

[0007] Based on the detection data of each perception node in the radio system, a perception matrix is constructed, and the target frequency band is scanned non-periodically through asynchronous pulse capture technology to obtain transient interference pulse signal data;

[0008] Based on the obtained signal data, real-time feature decoupling is carried out, and the time domain waveform features, frequency domain spectrum features, spatial arrival angle features and modulation domain feature vectors of the interference signal are extracted based on tensor decomposition algorithm;

[0009] A dual-mode spectrum analysis model is constructed, the real-time mode reconstructs the instantaneous spectrum graph of the disturbed area through the compression sensing algorithm, identifies the spectrum hole and the boundary of the high-power interference area, and the prediction mode predicts the interference mode and intensity distribution in the future time through the long short-term memory network to obtain the predicted interference graph;

[0010] The obtained instantaneous spectrum graph and the predicted interference graph are fused, and the identified interference signals are accurately classified and threat level evaluated through the pre-trained multi-level interference classification model;

[0011] Based on the interference signal category and the threat level, the optimal beamforming weight vector is calculated, the adaptive null is generated for the direction of the interference source, and the beam gain of the receiving direction of the target signal is optimized;

[0012] Based on the spectrum hole distribution identified by the real-time mode of the dual-mode spectrum analysis model and the beamforming result, a power density optimization model is constructed to dynamically allocate the transmit power spectral density of each sensing node and communication node;

[0013] Through the prediction mode of the dual-mode spectrum analysis model, an anti-interference frequency hopping sequence is dynamically generated based on the chaos mapping algorithm for dynamic spectrum camouflage and anti-tracking interference.

[0014] Preferably, based on the detection data of each sensing node in the radio system, a sensing matrix is constructed, and the target frequency band is scanned non-periodically through the asynchronous pulse capture technology to obtain transient interference pulse signal data, specifically including:

[0015] Based on the signal data collected by each sensing node, the sensing matrix is constructed by summarizing and fusing the signal data;

[0016] The sensing matrix simultaneously covers the spatial dimension, time dimension, frequency dimension and signal modulation dimension of the target frequency band;

[0017] The target frequency band is scanned non-periodically through the asynchronous pulse capture technology, and the burst interference pulse is captured in real time based on the ability to quickly capture transient signals to obtain transient interference pulse signal data.

[0018] Preferably, the real-time feature decoupling is performed based on the obtained signal data, and the time domain waveform feature, frequency domain spectrum feature, spatial arrival angle feature and modulation domain feature vector of the interference signal are extracted based on the tensor decomposition algorithm, specifically including:

[0019] The signal data is decoupled in the time domain through filtering and difference method to identify and distinguish the target signal and noise;

[0020] The signal data is decoupled in the frequency domain through the Fourier transform method to obtain the energy distribution of each frequency band, and the low-energy noise frequency band is removed;

[0021] Based on the decoupled signal data, a high-dimensional tensor data structure is constructed, and each dimension represents a feature of the signal.

[0022] The original tensor is decomposed into multiple factor matrices by Tucker decomposition, which respectively represent the time domain waveform feature, frequency domain spectrum feature, spatial domain arrival angle feature and modulation domain feature of the signal.

[0023] Based on the extracted features, a neural network classification algorithm is used to further judge whether the signal is a target signal, noise or potential interference signal.

[0024] Preferably, the dual-mode spectrum analysis model is constructed, the real-time mode reconstructs the instantaneous spectrum graph of the interference area by the compressed sensing algorithm, identifies the spectrum hole and high-power interference area boundary, and the prediction mode predicts the interference mode and intensity distribution in the future time by the long short-term memory network, and the specific steps of obtaining the predicted interference graph include:

[0025] A dual-mode spectrum analysis model is constructed, which includes a real-time mode and a prediction mode;

[0026] The real-time mode reconstructs the sparse signal data in the sensing matrix by the compressed sensing algorithm, and based on the sparsity of the signal, the complete spectrum graph is reconstructed from a small amount of measurement data;

[0027] Based on the reconstructed spectrum data, the instantaneous spectrum graph of the interference area is generated, the energy distribution data of each frequency band at the current time point is obtained, and the spectrum hole and high-power interference area boundary are identified;

[0028] The prediction mode trains the LSTM model based on historical interference data, identifies the periodicity, volatility and intensity change law of the interference mode;

[0029] Based on the prediction result of the trained LSTM, a future predicted interference graph is generated, and the interference intensity distribution of each frequency band in the prediction time is obtained.

[0030] Preferably, the obtained instantaneous spectrum graph and the predicted interference graph are fused, and the identified interference signal is accurately classified and threat level evaluated by a pre-trained multi-level interference classification model, which specifically includes:

[0031] The obtained instantaneous spectrum graph and the predicted interference graph output by the prediction mode are fused by weighted average and splicing method;

[0032] The interference signal is accurately classified by a pre-trained multi-level interference classification model;

[0033] The multi-level interference classification model includes: the first level distinguishes between intentional interference and unintentional interference, the second level identifies specific interference types based on a transfer learning framework, and the third level combines signal arrival intensity and spatial domain information to coarsely locate the interference source.

[0034] Based on the multi-level interference classification model classification result, the threat level of the interference is comprehensively evaluated in combination with the strength, type, duration of the interference and the positioning accuracy of the interference source.

[0035] Preferably, the optimal beamforming weight vector is calculated based on the interference signal category and the threat level, the adaptive null is generated for the direction of the interference source, and the beam gain of the target signal receiving direction is optimized, which specifically includes:

[0036] Based on the direction of the interference source determined based on the coarse positioning data of the interference source, the optimization focus of the beamforming is determined according to the angle between the interference source and the target signal receiving direction. If the angle is greater than a threshold, the focus is to enhance the gain of the target signal receiving direction. If the angle is less than the threshold, the focus is to suppress the interference source.

[0037] Based on the positioning of the interference source and the direction of the target signal, the preliminary beamforming weight vector is calculated through the array factor of the array antenna, and the beamforming weight vector is dynamically adjusted through the least mean square error method.

[0038] An adaptive null is generated for the direction of the intentional interference source, and the depth and width of the null are dynamically adjusted based on the strength of the interference signal.

[0039] Under the premise of interference suppression, the gain weight of each unit of the antenna array is dynamically adjusted to maximize the beam gain.

[0040] Preferably, the power density optimization model is constructed based on the distribution of the spectral hole identified by the real-time pattern recognition of the dual-mode spectrum analysis model and the beamforming result, and the transmit power spectral density of each sensing node and communication node is dynamically allocated, which specifically includes:

[0041] The power density optimization model is constructed based on the distribution of the spectral hole identified by the real-time pattern recognition of the dual-mode spectrum analysis model and the beamforming result, and the target is to maximize the communication performance and minimize the total transmit power.

[0042] The available spectral hole is matched with the corresponding node for spectral resource, and priority is allocated based on the environment of each node, the distance from the interference source, and the beam gain factor.

[0043] A power spectral density allocation matrix is established, each row representing a node and each column representing a frequency band, and the elements in the matrix represent the transmit power density of the node on the frequency band.

[0044] Preferably, the power density optimization model is constructed based on the distribution of the spectral hole identified by the real-time pattern recognition of the dual-mode spectrum analysis model and the beamforming result, and the transmit power spectral density of each sensing node and communication node is dynamically allocated, which specifically includes:

[0045] A power density optimization model is constructed based on the distribution of spectrum holes and the beamforming results of real-time mode recognition of the dual-mode spectrum analysis model, with the goal of maximizing communication performance and minimizing total transmit power;

[0046] The available spectrum holes are matched with the corresponding nodes for spectrum resource, and priorities are assigned based on the environment of each node, the distance from the interference source, and the beam gain factor.

[0047] A power spectrum density allocation matrix is established, with each row representing a node and each column representing a frequency band. The elements in the matrix represent the transmit power density of the node on that frequency band.

[0048] Preferably, the chaotic mapping algorithm uses an improved Logistic-Tent composite chaotic system to generate a frequency hopping key sequence; the synchronization frequency hopping rate adjustment range is 10 hops per second to 10,000 hops per second.

[0049] Further, a precise interference avoidance device in a wireless radio system is proposed, comprising:

[0050] The perception matrix module: through the perception of the detection data of the nodes, the module constructs a perception matrix covering the dimensions of space, time, frequency and signal modulation, providing data support for subsequent interference analysis;

[0051] The transient interference signal acquisition module: the module uses asynchronous pulse acquisition technology to perform non-periodic scanning on the target frequency band, quickly captures and acquires transient interference pulse signal data;

[0052] The feature decoupling module: based on the acquired signal data, the module uses filtering, Fourier transform and tensor decomposition algorithms to extract the time domain, frequency domain, spatial domain and modulation domain features of the interference signal;

[0053] The dual-mode spectrum analysis model module: the module is used to construct a dual-mode spectrum analysis model of real-time mode and prediction mode, to reconstruct the instantaneous spectrum graph through compressed sensing algorithm, to identify the spectrum holes and interference areas, and to predict the interference mode and intensity distribution based on the LSTM model;

[0054] The interference source suppression module: based on the interference signal category and threat level, the module calculates the optimal beamforming weight vector, suppresses the interference source through adaptive nulling, and optimizes the reception direction gain of the target signal;

[0055] The power density optimization and anti-interference module: the module dynamically allocates the transmit power spectrum density of each perception node and communication node based on the identification results of the dual-mode spectrum analysis model, dynamically generates an anti-interference frequency hopping sequence based on the chaotic mapping algorithm, and resists tracking interference through spectrum camouflage and frequency hopping to ensure the anti-interference ability of the system;

[0056] Processor: the processor is used for processing the calculation process of each formula and the construction calculation process of each model.

[0057] Compared with the prior art, the advantages of the present application are that:

[0058] By fusing multi-dimensional perception data, tensor feature decoupling, dual-mode spectrum analysis, intelligent interference classification and adaptive beamforming, etc. Advanced technologies, the whole process of intelligent closed loop from interference perception, identification, prediction to avoidance is realized. Its greatest advantage is that it has very high real-time and accuracy, can quickly identify and distinguish intentional and unintentional interference in complex electromagnetic environment, dynamically generate null and frequency hopping strategy, realize high-fidelity reception of target signal and effective suppression of interference source. At the same time, the spectrum camouflage and frequency hopping mechanism based on chaos mapping greatly improves the anti-tracking interference ability of the system, combined with the power density optimization strategy, guarantees the communication performance while reducing the system energy consumption. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 The method proposed by the present application is shown in the schematic diagram;

[0060] Figure 2 The data acquisition of transient interference pulse signal proposed by the present application is shown in the schematic diagram;

[0061] Figure 3 The real-time feature decoupling proposed by the present application is shown in the schematic diagram;

[0062] Figure 4 The acquisition of predicted interference map proposed by the present application is shown in the schematic diagram;

[0063] Figure 5 The classification and threat level evaluation proposed by the present application is shown in the schematic diagram;

[0064] Figure 6 The calculation of optimal beamforming weight vector proposed by the present application is shown in the schematic diagram;

[0065] Figure 7 The dynamic allocation of transmit power spectral density proposed by the present application is shown in the schematic diagram;

[0066] Figure 8 The dynamic spectrum camouflage and anti-tracking interference proposed by the present application is shown in the schematic diagram. DETAILED DESCRIPTION

[0067] The following description is used to disclose the present application so that those skilled in the art can implement the present application. The preferred embodiments in the following description are only as examples, and other obvious modifications can be thought by those skilled in the art.

[0068] A precise interference avoidance device in a radio system, comprising:

[0069] The perception matrix module: the module summarizes and fuses the detection data of the perception nodes, constructs a perception matrix covering the dimensions of space, time, frequency and signal modulation, and provides data support for subsequent interference analysis;

[0070] The transient interference signal acquisition module: the module uses asynchronous pulse acquisition technology to perform non-periodic scanning on the target frequency band, quickly captures and acquires transient interference pulse signal data;

[0071] The feature decoupling module: based on the acquired signal data, the module uses filtering, Fourier transform and tensor decomposition algorithm to extract the time domain, frequency domain, spatial domain and modulation domain features of the interference signal;

[0072] The dual-mode spectrum analysis model module: the module is used to construct a dual-mode spectrum analysis model of real-time mode and prediction mode, reconstruct the instantaneous spectrum graph through the compressed sensing algorithm, identify the spectrum holes and interference areas, and predict the interference mode and intensity distribution based on the LSTM model;

[0073] The interference source suppression module: based on the interference signal category and threat level, the module calculates the optimal beamforming weight vector, suppresses the interference source through adaptive nulling, and optimizes the receiving direction gain of the target signal;

[0074] The power density optimization and anti-interference module: the module combines the identification results of the dual-mode spectrum analysis model, dynamically allocates the transmit power spectral density of each perception node and communication node, dynamically generates an anti-interference frequency hopping sequence based on the chaos mapping algorithm, and resists tracking interference through spectrum camouflage and frequency hopping to ensure the anti-interference ability of the system;

[0075] The processor: the processor is used to process the calculation process of each formula and the construction calculation process of each model.

[0076] Referring to Figure 1 The figure shows a precise interference avoidance method in a radio system, which includes:

[0077] Step one: based on the detection data of each perception node in the radio system, a perception matrix is constructed, and non-periodic scanning is performed on the target frequency band through asynchronous pulse acquisition technology to acquire transient interference pulse signal data;

[0078] Step two: real-time feature decoupling based on the acquired signal data, extracting the time domain waveform feature, frequency domain spectrum feature, spatial domain arrival angle feature and modulation domain feature vector of the interference signal based on the tensor decomposition algorithm;

[0079] Step three: build a dual-mode spectrum analysis model, the real-time mode reconstructs the instantaneous spectrum graph of the disturbed area through the compressed sensing algorithm, identifies the spectrum hole and the boundary of the high-power interference area, and the prediction mode predicts the interference mode and intensity distribution in the future time through the long short-term memory network to obtain the predicted interference graph;

[0080] Step four: fuse the obtained instantaneous spectrum graph and the predicted interference graph, and accurately classify and threat level assess the identified interference signals through the pre-trained multi-level interference classification model;

[0081] Step five: calculate the optimal beamforming weight vector based on the interference signal category and threat level, generate adaptive nulls for the interference source direction, and optimize the beam gain of the target signal receiving direction;

[0082] Step six: based on the distribution of the spectrum hole identified by the real-time mode of the dual-mode spectrum analysis model and the beamforming result, build a power density optimization model to dynamically allocate the transmit power spectral density of each sensing node and communication node;

[0083] Step seven: based on the prediction mode of the dual-mode spectrum analysis model, dynamically generate an anti-interference frequency hopping sequence based on the chaos mapping algorithm for dynamic spectrum camouflage and anti-tracking interference.

[0084] Referring to Figure 2 As shown, based on the detection data of each sensing node in the radio system, a sensing matrix is constructed, and the target frequency band is scanned non-periodically through an asynchronous pulse capture technology to obtain transient interference pulse signal data, which specifically includes:

[0085] Based on the signal data collected by each sensing node, the sensing matrix is constructed by summarizing and fusing the signal data;

[0086] The sensing matrix simultaneously covers the spatial dimension, time dimension, frequency dimension and signal modulation dimension of the target frequency band;

[0087] Through the asynchronous pulse capture technology, the target frequency band is scanned non-periodically, and based on the ability to quickly capture transient signals, the burst interference pulse is captured in real time to obtain the transient interference pulse signal data.

[0088] Specifically, the target frequency band is detected by the distributed sensing nodes, and each sensing node is responsible for capturing the signal data in its coverage area. Based on the constructed sensing matrix, the asynchronous pulse capture technology is used to scan the target frequency band non-periodically. The key of this technology lies in its ability to quickly capture transient interference pulse signals, adapt to the burstness and non-periodicity of signals in time and frequency, and mainly rely on high time resolution detection of signal changes and fast analysis ability of signals;

[0089] When a sudden interference pulse appears in the sensing matrix, the system can react quickly and capture these transient signals. The change in signal intensity can be expressed as:

[0090]

[0091] in, The signal data collected for each node, It is the time dimension. It is the frequency dimension. The amplitude of the transient signal. and It refers to the sudden timing and frequency of the interference signal. This indicates the concentration of a signal at a certain point;

[0092] Through asynchronous pulse acquisition technology, the system can efficiently capture transient interference pulse signals within the target frequency band.

[0093] See Figure 3 As shown, real-time feature decoupling is performed based on the acquired signal data. The extraction of time-domain waveform features, frequency-domain spectrum features, spatial angle-of-arrival features, and modulation-domain feature vectors of the interference signal based on the tensor decomposition algorithm specifically includes:

[0094] The acquired signal data is decoupled from time-domain features using filtering and differential methods to identify and distinguish target signals from noise.

[0095] The signal data is decoupled in the frequency domain by using the Fourier transform method to obtain the energy distribution of each frequency band and remove low-energy noise frequency bands.

[0096] Based on the decoupled signal data, a high-dimensional tensor data structure is constructed, where each dimension represents a feature of the signal;

[0097] The original tensor is decomposed into multiple factor matrices by Tucker decomposition, which represent the time-domain waveform characteristics, frequency-domain spectrum characteristics, spatial-domain angle of arrival characteristics, and modulation-domain characteristics of the signal, respectively.

[0098] Based on the extracted features, a neural network classification algorithm is used to further determine whether the signal is a target signal, noise, or potential interference signal.

[0099] Specifically, the main task of time-domain feature decoupling is to separate the target signal and noise from the original signal. The differential method is used to further remove high-frequency noise from the signal and enhance the smoothness of the target signal.

[0100] The time domain signal can be converted to the frequency domain by Fourier transform, so as to decouple the frequency domain features. The core task of frequency domain feature decoupling is to remove the low-energy noise frequency band. The noise frequency band is usually represented by a lower amplitude, while the target signal is concentrated in a certain frequency range. By setting a threshold, the frequency band smaller than the threshold in the frequency spectrum is regarded as noise, and is removed from the signal;

[0101] After decoupling the time domain and the frequency domain, a high-dimensional tensor data structure is constructed to store the feature information of each dimension. Each tensor dimension represents a feature of the signal. After constructing the tensor data structure, a tensor decomposition algorithm is used to decompose the tensor to extract the time domain, frequency domain, spatial domain and modulation domain features of the signal. Tucker decomposition is a tensor decomposition method that can decompose a high-dimensional tensor into multiple factor matrices. After decomposition, each factor matrix contains the features of the signal in each dimension.

[0102] After feature extraction, classification and discrimination of the signal are performed based on the extracted feature vectors.

[0103] Referring to Figure 4 A dual-mode spectrum analysis model is constructed. The real-time mode reconstructs the instantaneous spectrum graph of the interference area by a compressive sensing algorithm, identifies the spectrum hole and the high-power interference area boundary, and the prediction mode predicts the interference mode and intensity distribution in the future time by a long short-term memory network. The predicted interference graph specifically includes:

[0104] A dual-mode spectrum analysis model is constructed, which includes a real-time mode and a prediction mode;

[0105] The real-time mode reconstructs the sparse signal data in the sensing matrix by a compressive sensing algorithm. Based on the sparsity of the signal, the complete spectrum graph is reconstructed from a small amount of measurement data;

[0106] Based on the reconstructed spectrum data, the instantaneous spectrum graph of the interference area is generated, the energy distribution data of each frequency band at the current time point is obtained, and the spectrum hole and the high-power interference area boundary are identified;

[0107] The prediction mode trains the LSTM model based on historical interference data to identify the periodicity, volatility and intensity change law of the interference mode;

[0108] Based on the prediction result of the trained LSTM, a future predicted interference graph is generated to obtain the interference intensity distribution of each frequency band in the prediction time.

[0109] Specifically, the real-time mode reconstructs the sparse spectrum signal by a compressive sensing algorithm to realize the spectrum graph recovery of the interference area. The spectrum data obtained by the sensing node has a sparse characteristic, which is represented as an observation vector Furthermore, the observation dimension is much smaller than the original signal dimension; let the spectral signal be a sparse vector. The observation matrix is ,satisfy:

[0110]

[0111] because It is sparse, and can be reconstructed using compressed sensing algorithms. Recovery ;

[0112] The sparse solution is obtained by minimizing the L1 norm, and the spectral signal is then reconstructed. That is, the energy distribution at each frequency point, which is remapped into a two-dimensional spectrum.

[0113] Spectral holes are regions where the spectral energy is less than a set threshold, and high-power interference regions are regions where the spectral energy exceeds the upper threshold.

[0114] The prediction model constructs the interference intensity of each frequency point in the spectrum within a historical time period as time series data, trains an LSTM model, captures the time dependence and nonlinearity in the signal, outputs the predicted value for the future time period, and synthesizes the interference intensity of each frequency point within the predicted time period into a predicted interference map.

[0115] See Figure 5 As shown, the acquired instantaneous spectrum and predicted interference map are fused together, and a pre-trained multi-level interference classification model is used to accurately classify and assess the threat level of the identified interference signals. Specifically, this includes:

[0116] The instantaneous spectrogram obtained is fused with the predicted interference map output by the prediction mode by weighted averaging and splicing methods.

[0117] The interference signal is accurately classified using a pre-trained multi-level interference classification model;

[0118] The multi-level interference classification model includes: a first level to distinguish between intentional and unintentional interference; a second level to identify specific interference types based on a transfer learning framework; and a third level to coarsely locate the interference source by combining signal arrival strength and spatial information.

[0119] Based on the classification results of the multi-level interference classification model, and combined with the intensity, type, duration of the interference and the positioning accuracy of the interference source, the threat level of the interference is comprehensively assessed.

[0120] Specifically, in the multi-stage interference classification model, the first stage is used to distinguish between intentional interference and unintentional interference. The classifier performs preliminary interference classification based on the features in the instantaneous spectrum and the predicted interference map. Specifically, intentional interference is usually caused by an enemy or malicious attack, and is characterized by specific modulation methods, spectral features, and intensity distribution. Unintentional interference may be caused by environmental noise, electromagnetic interference, or equipment failure, and is usually characterized by random spectral disturbances or small amplitude fluctuations.

[0121] In the second stage, the identified interference signal types are classified in more detail through a transfer learning framework. Transfer learning uses knowledge learned from other fields or similar tasks to apply existing models to new interference type identification tasks, improving classification effectiveness, especially in cases where there are insufficient training samples. In this stage, detailed features of the interference signal, such as frequency domain spectrum, time domain waveform, modulation method, and spatial domain distribution, are used to further determine the interference type. By comparing the features of various interference signals, the classifier can identify the interference type based on the transfer learning model.

[0122] In the third stage, the signal arrival intensity and spatial information are combined to perform coarse positioning of the interference source. In this stage, the system combines the signal intensity data received by each perception node to infer the possible location of the interference source based on the intensity distribution. The spatial distribution of multiple sensors is used to perform coarse positioning of the interference source using spatial arrival angle technology. Each perception node can estimate the direction of the interference source based on the arrival angle of the signal, and the measurement data from multiple sensors are combined to perform positioning.

[0123] Combining the results of the above three stages, the system can comprehensively evaluate the threat of the interference. The threat level evaluation usually considers the intensity, type, duration of the interference, and the positioning accuracy of the interference source.

[0124] Referring to Figure 6 The optimal beamforming weight vector is calculated based on the interference signal category and threat level, adaptive nulls are generated for the direction of the interference source, and the beam gain of the target signal reception direction is optimized. Specifically, it includes:

[0125] Based on the coarse positioning data of the interference source, the direction of the interference source is determined. According to the angle between the interference source and the target signal reception direction, the optimization focus of beamforming is determined. If the angle is greater than a threshold, the focus is to enhance the gain of the target signal reception direction. If the angle is less than a threshold, the focus is to suppress the interference source.

[0126] Based on the positioning of the interference source and the direction of the target signal, a preliminary beamforming weight vector is calculated through the array factor of the array antenna, and the beamforming weight vector is dynamically adjusted through the least mean square error method.

[0127] The adaptive null is generated for the direction of the intentional interference source, and the depth and width of the null are dynamically adjusted based on the intensity of the interference signal;

[0128] Under the premise of interference suppression, the gain weight of each unit of the antenna array is dynamically adjusted to maximize the beam gain.

[0129] Specifically, according to the known interference source positioning data, the rough direction of the interference source is determined, and the included angle between the direction of the interference source and the direction of the target signal is calculated. The included angle will determine the strategy of optimizing the beamforming:

[0130] If the included angle is greater than the preset threshold, the optimization focus is to enhance the gain of the target signal receiving direction;

[0131] If the included angle is less than the preset threshold, the optimization focus is to suppress the interference source;

[0132] According to the directions of the interference source and the target signal, the array factor of the array antenna is used for preliminary beamforming design. The array factor can help determine the gain in the target direction and suppress the direction of the interference source. Based on the interference source positioning data and the direction of the target signal, the minimum mean square error method is used to dynamically adjust the preliminary beamforming weight vector.

[0133] For the known direction of the intentional interference source, the signal intensity of the interference source is suppressed to the lowest by designing an adaptive null. The generation of the null is realized by adjusting the phase and amplitude of the array antenna, ensuring that the received power in the direction of the interference source is close to zero. The depth and width of the null are dynamically adjusted according to the signal intensity of the interference source and its influence range. By monitoring the intensity change of the interference source, the position, depth and width of the null are adjusted in time to achieve the optimal interference suppression effect.

[0134] Under the premise of interference suppression, the gain weight of each unit of the antenna array is dynamically adjusted to optimize the beam gain. According to the geometric structure of the antenna array and the relative position between the direction of the target signal and the direction of the interference source, the gain weight of each unit is reasonably allocated to make the beam achieve optimal gain in the direction of the target signal receiving direction and minimum gain in the direction of the interference source.

[0135] Referring to Figure 7 The power density optimization model is constructed based on the distribution of the spectrum hole identified by the dual-mode spectrum analysis model in real time and the beamforming result, and the transmit power spectral density of each sensing node and communication node is dynamically allocated, which specifically includes:

[0136] The power density optimization model is constructed based on the distribution of the spectrum hole identified by the dual-mode spectrum analysis model in real time and the beamforming result, and the target is to maximize the communication performance and minimize the total transmit power;

[0137] The available spectrum hole is matched with the corresponding node for spectrum resource, and priority is allocated based on the environment of each node, distance from the interference source, and beam gain factor;

[0138] A power spectral density allocation matrix is established, each row representing a node and each column representing a frequency band, and the element in the matrix representing the transmission power density of the node on the frequency band.

[0139] Specifically, the objective of the optimization model is to maximize the communication performance while minimizing the total transmission power. According to the real-time identified spectrum hole distribution, the hole region is matched with the spectrum resource of the sensing node or communication node, and each spectrum hole should be allocated to one or more communication nodes to ensure the maximum utilization of the frequency band. When allocating spectrum resources, the environmental factors of the nodes and the distance from the interference source are considered. Nodes far from the interference source are allocated higher power density, while nodes close to the interference source should be allocated lower power density to reduce interference.

[0140] Based on the beamforming result, the beam gain of each node is considered. If the beam gain of a node is high, its transmission power can be appropriately reduced to avoid interference, while nodes with low beam gain may need more power support.

[0141] According to the communication needs of the nodes, environmental factors, distance from the interference source, and idle condition of the spectrum resource, each node is allocated priority, and a power spectral density allocation matrix is constructed. Each row of the matrix represents a sensing node or communication node, each column represents a frequency band, and each element in the matrix represents the transmission power density of the node on the specific frequency band. According to the above priority allocation, spectrum hole matching and beam gain factor, the transmission power density of each node on each frequency band is calculated.

[0142] Referring to Figure 8 As shown, through the prediction mode of the dual-mode spectrum analysis model, an anti-interference frequency hopping sequence is dynamically generated based on the chaotic mapping algorithm, and dynamic spectrum camouflage and anti-tracking interference specifically includes:

[0143] The interference intensity, frequency distribution and change rate data in the future time are obtained through the prediction mode of the dual-mode spectrum analysis model;

[0144] The initial value is obtained based on the current system state as the starting condition of the chaotic mapping, and the key sequence is generated through iterative operation according to the chaotic mapping algorithm, and is used as the control signal of the subsequent frequency hopping sequence;

[0145] The frequency band of the hollow region in the spectrum is taken as the target frequency band of the frequency hopping sequence, the generated chaotic key is combined with the spectrum hollow information, and based on the value of the key sequence and the distribution of the spectrum hollow, the system hops between the hollow frequency bands, each value of the key sequence corresponds to a new frequency or frequency band, forming a frequency hopping pattern;

[0146] The rate of frequency hopping is adaptively adjusted based on the real-time interference situation and the predicted change rate, and the chaotic key and the frequency hopping pattern are adjusted according to the real-time frequency hopping effect.

[0147] Specifically, based on the selected initial value, a series of key sequences are generated through a chaotic mapping algorithm. The key generated by the chaotic mapping algorithm through iterative operation has good randomness and unpredictability, and is suitable for encryption and spectrum camouflage tasks. Through spectrum hollow analysis, the idle frequency bands in the spectrum are identified, the key sequence generated by the chaotic mapping algorithm is combined with the spectrum hollow information to form a frequency hopping sequence. Specifically, each value of the key sequence will correspond to a new frequency or frequency band, ensuring that the system hops between the spectrum hollows;

[0148] During the frequency hopping process, the system needs to monitor the situation of the interference source in real time. If the interference source suddenly becomes stronger or changes position, the system needs to respond quickly, dynamically adjust the frequency hopping rate based on the predicted change rate of the interference signal, for example, when it is predicted that the interference signal will change strongly in a short time, the frequency hopping rate should be accelerated to increase the spectrum camouflage effect; when the interference signal is relatively stable, the frequency hopping rate can be appropriately slowed down to reduce the frequency of spectrum hopping of the system;

[0149] According to the real-time frequency hopping effect, the parameters of the chaotic key sequence and the frequency hopping pattern are adjusted, and through the feedback mechanism, the system can evaluate the current frequency hopping effect after each frequency hopping period, and adjust the frequency hopping strategy according to the actual interference situation, further optimizing the anti-interference ability.

[0150] It should be noted that the above-mentioned embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments. Moreover, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or may be advantageous.

[0151] Each embodiment in the specification is described in a progressive manner, and the same or similar parts of each embodiment can be referred to each other, and each embodiment focuses on the difference from other embodiments.

[0152] The above only describes the preferred embodiments of the present application, and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A precise interference avoidance method in a radio system, characterized in that, include: Based on the detection data of each sensing node in the radio system, a sensing matrix is ​​constructed, and the target frequency band is scanned non-periodically using asynchronous pulse acquisition technology to obtain transient interference pulse signal data. Real-time feature decoupling is performed based on the acquired signal data, and time-domain waveform features, frequency-domain spectrum features, spatial-domain angle of arrival features, and modulation-domain feature vectors of the interference signal are extracted based on the tensor decomposition algorithm. A dual-mode spectrum analysis model is constructed. The real-time mode reconstructs the instantaneous spectrum map of the interfered area through the compressed sensing algorithm to identify spectrum holes and the boundary of high-power interference area. The prediction mode predicts the interference mode and intensity distribution in the future time through the long short-term memory network to obtain the predicted interference map. The acquired instantaneous spectrum map and the predicted interference map are fused together, and the identified interference signals are accurately classified and their threat level is assessed by a pre-trained multi-level interference classification model. The optimal beamforming weight vector is calculated based on the type and threat level of the interference signal, an adaptive null is generated for the direction of the interference source, and the beam gain in the direction of the target signal reception is optimized. Based on the distribution of spectral holes and beamforming results of real-time pattern recognition using a dual-mode spectrum analysis model, a power density optimization model is constructed to dynamically allocate the transmit power spectral density of each sensing node and communication node. By using the prediction mode of the dual-mode spectrum analysis model, and based on the chaotic mapping algorithm, an anti-interference frequency hopping sequence is dynamically generated to perform dynamic spectrum camouflage and anti-tracking interference.

2. The precise interference avoidance method in a radio system according to claim 1, characterized in that, The process of constructing a sensing matrix based on detection data from each sensing node in the radio system, and acquiring transient interference pulse signal data by performing a non-periodic scanning of the target frequency band using asynchronous pulse acquisition technology specifically includes: The signal data collected by each sensing node is summarized and fused to construct a sensing matrix; The perception matrix simultaneously covers the spatial dimension, temporal dimension, frequency dimension, and signal modulation dimension of the target frequency band; The target frequency band is scanned non-periodically using asynchronous pulse acquisition technology. Based on the ability to quickly capture transient signals, sudden interference pulses are captured in real time to obtain transient interference pulse signal data.

3. The precise interference avoidance method in a radio system according to claim 1, characterized in that, The real-time feature decoupling based on the acquired signal data, and the extraction of time-domain waveform features, frequency-domain spectrum features, spatial angle-of-arrival features, and modulation-domain feature vectors of the interference signal based on the tensor decomposition algorithm, specifically include: The acquired signal data is decoupled from time-domain features using filtering and differential methods to identify and distinguish target signals from noise. The signal data is decoupled in the frequency domain by using the Fourier transform method to obtain the energy distribution of each frequency band and remove low-energy noise frequency bands. Based on the decoupled signal data, a high-dimensional tensor data structure is constructed, where each dimension represents a feature of the signal; The original tensor is decomposed into multiple factor matrices by Tucker decomposition, which represent the time-domain waveform characteristics, frequency-domain spectrum characteristics, spatial-domain angle of arrival characteristics, and modulation-domain characteristics of the signal, respectively. Based on the extracted features, a neural network classification algorithm is used to further determine whether the signal is a target signal, noise, or potential interference signal.

4. The precise interference avoidance method in a radio system according to claim 1, characterized in that, The construction of the dual-mode spectrum analysis model involves the real-time mode reconstructing the instantaneous spectrum of the interfered region using a compressed sensing algorithm to identify spectral holes and the boundaries of high-power interference regions, and the predictive mode using a long short-term memory network to predict the interference patterns and intensity distributions in the future, thereby obtaining a predicted interference map. Specifically, this includes: A dual-mode spectrum analysis model is constructed, which includes a real-time mode and a prediction mode; The real-time mode reconstructs the sparse signal data in the sensing matrix using a compressed sensing algorithm. Based on the sparsity of the signal, a complete spectrum is reconstructed using a small amount of measurement data. Based on the reconstructed spectrum data, an instantaneous spectrum map of the interfered area is generated, the energy distribution data of each frequency band at the current time point is obtained, and the spectrum holes and high-power interference area boundaries are identified. The prediction model is trained using LSTM model based on historical disturbance data to identify the periodicity, volatility and intensity variation of the disturbance pattern. Based on the prediction results of the trained LSTM, a future predicted interference map is generated to obtain the interference intensity distribution of each frequency band within the prediction time.

5. The precise interference avoidance method in a radio system according to claim 1, characterized in that, The process of fusing the acquired instantaneous spectrum and the predicted interference map, and then using a pre-trained multi-level interference classification model to accurately classify and assess the threat level of the identified interference signals specifically includes: The instantaneous spectrogram obtained is fused with the predicted interference map output by the prediction mode by weighted averaging and splicing methods. The interference signal is accurately classified using a pre-trained multi-level interference classification model; The multi-level interference classification model includes: a first level to distinguish between intentional and unintentional interference; a second level to identify specific interference types based on a transfer learning framework; and a third level to coarsely locate the interference source by combining signal arrival strength and spatial information. Based on the classification results of the multi-level interference classification model, and combined with the intensity, type, duration of the interference and the positioning accuracy of the interference source, the threat level of the interference is comprehensively assessed.

6. The precise interference avoidance method in a radio system according to claim 1, characterized in that, The process of calculating the optimal beamforming weight vector based on the type and threat level of the interference signal, generating adaptive nulls in the direction of the interference source, and optimizing the beam gain in the direction of the target signal reception specifically includes: Based on the coarse localization data of the interference source, the direction of the interference source is determined. According to the angle between the interference source and the target signal receiving direction, the optimization focus of beamforming is determined. If the angle is greater than the threshold, the focus is on enhancing the gain of the target signal receiving direction. If the angle is less than the threshold, the focus is on suppressing the interference source. Based on the location of the interference source and the direction of the target signal, the preliminary beamforming weight vector is calculated by the array factor of the array antenna, and the beamforming weight vector is dynamically adjusted by the minimum mean square error method. An adaptive null is generated for the direction of the intentional interference source, and the depth and width of the null are dynamically adjusted based on the intensity of the interference signal. Based on interference suppression, beam gain is maximized by dynamically adjusting the gain weights of each element in the antenna array.

7. The precise interference avoidance method in a radio system according to claim 1, characterized in that, The power density optimization model, constructed based on the distribution of spectral holes and beamforming results of real-time pattern recognition using a dual-mode spectrum analysis model, dynamically allocates the transmit power spectral density of each sensing node and communication node, specifically including: A power density optimization model is constructed based on the distribution of spectral holes and beamforming results of real-time pattern recognition using a dual-mode spectrum analysis model. The goal is to maximize communication performance and minimize total transmit power. Available spectrum holes are matched with corresponding nodes for spectrum resources, and priorities are assigned based on the environment of each node, distance from the interference source, and beam gain factors. Establish a power spectral density allocation matrix, where each row represents a node, each column represents a frequency band, and the elements in the matrix represent the transmit power density of that node in that frequency band.

8. The precise interference avoidance method in a radio system according to claim 1, characterized in that, The prediction mode based on the dual-mode spectrum analysis model, and the dynamic generation of anti-interference frequency hopping sequences based on the chaotic mapping algorithm for dynamic spectrum camouflage and anti-tracking interference specifically include: Data on interference intensity, frequency distribution, and rate of change over future time periods are obtained through the prediction mode of a dual-mode spectrum analysis model. The initial value is obtained based on the current system state as the starting condition for the chaotic mapping, and a key sequence is generated through iterative calculation according to the chaotic mapping algorithm, which is used as the control signal for the subsequent frequency hopping sequence. The frequency bands of the hole region in the spectrum are used as the target frequency bands of the frequency hopping sequence. The generated chaotic key is combined with the spectrum hole information. Based on the value of the key sequence and the distribution of spectrum holes, the system performs frequency hopping between hole frequency bands. Each value of the key sequence corresponds to a new frequency or frequency band, forming a frequency hopping pattern. The frequency hopping rate is adaptively adjusted based on real-time interference conditions and the predicted rate of change, and the chaotic key and frequency hopping pattern are adjusted according to the real-time frequency hopping effect.

9. A precise interference avoidance method in a radio system according to claim 8, characterized in that, The chaotic mapping algorithm uses an improved Logistic-Tent composite chaotic system to generate a frequency hopping key sequence; the frequency hopping rate of the frequency hopping key sequence is adjusted from 10 hops per second to 10,000 hops per second.

10. A precision interference avoidance device in a radio system, used to implement the precision interference avoidance method in a radio system as described in any one of claims 1-9, characterized in that, include: Perception Matrix Module: This module collects and fuses the detection data from the sensing nodes to construct a perception matrix covering spatial, temporal, frequency, and signal modulation dimensions, providing data support for subsequent interference analysis. Transient interference signal acquisition module: The module uses asynchronous pulse acquisition technology to perform non-periodic scanning of the target frequency band, quickly acquire and obtain transient interference pulse signal data; Feature decoupling module: Based on the acquired signal data, the module uses filtering, Fourier transform and tensor decomposition algorithms to extract the time domain, frequency domain, spatial domain and modulation domain features of the interference signal; Dual-mode spectrum analysis model module: The module is used to construct a dual-mode spectrum analysis model of real-time mode and prediction mode, reconstruct the instantaneous spectrum map through compressed sensing algorithm, identify spectrum holes and interference regions, and predict interference modes and intensity distribution based on LSTM model; Interference source suppression module: The module calculates the optimal beamforming weight vector based on the type and threat level of the interference signal, suppresses the interference source through adaptive nulls, and optimizes the receiving direction gain of the target signal; Power density optimization and anti-interference module: The module combines the identification results of the dual-mode spectrum analysis model to dynamically allocate the transmit power spectral density of each sensing node and communication node. Based on the chaotic mapping algorithm, it dynamically generates an anti-interference frequency hopping sequence. It resists tracking interference through spectrum camouflage and frequency hopping, ensuring the anti-interference capability of the system. Processor: The processor is used to handle the calculation process of each formula and the construction calculation process of each model.

Images