Frequency conversion dynamic scanning method, device, equipment and medium

Abstract

The invention relates to the technical field of signal processing and frequency scanning control. The frequency conversion dynamic scanning method comprises the following steps: acquiring a radio frequency signal in a target frequency band, performing high-speed sampling on the radio frequency signal to obtain an analog sampling signal, performing analog-to-digital conversion on the analog sampling signal to obtain a digital signal, and performing spectral analysis on the digital signal to obtain a frequency conversion signal; and extracting frequency characteristic information of the digital signal, inputting the frequency characteristic information into the frequency regulation and control model to obtain a frequency control instruction, adjusting the output frequency of the local oscillator according to the frequency control instruction, and executing nonlinear hopping frequency scanning. The method has the effect of improving the frequency spectrum coverage efficiency.

Description

Technical Field

[0001] This invention relates to the technical field of signal processing and frequency scanning control, and in particular to a variable frequency dynamic scanning method, apparatus, device and medium. Background Technology

[0002] With the rapid development of technologies such as wireless communication, radar detection, and electromagnetic sensing, the management and utilization of spectrum resources are facing unprecedented challenges. Especially in complex electromagnetic environments, with a wide variety of signals, frequent interference, and dynamic changes in frequency band occupancy, traditional linear, fixed-step frequency scanning methods are unable to meet the real-time and accuracy requirements for rapid identification of target signals, spectrum coverage optimization, and interference avoidance.

[0003] In existing technologies, sequential scanning and equal-frequency-interval hopping are commonly used for spectrum detection and analysis. However, these methods have the following main problems: First, they lack the ability to respond to dynamic spectrum conditions and cannot dynamically adjust the scanning path according to the activity or changing trend of frequency band signals. Second, the control strategy is generally rule-based and cannot adapt to the nonlinear evolution of frequency hopping sequence in complex environments, resulting in uneven spectrum resource coverage or untimely interference identification. Third, traditional methods usually lack self-learning capabilities and lack the ability to utilize historical data feedback, making it impossible to achieve intelligent optimization of spectrum scanning and frequency control. Summary of the Invention

[0004] To improve spectrum coverage efficiency, this application provides a frequency conversion dynamic scanning method, apparatus, device, and medium.

[0005] The above-mentioned objective of this application is achieved through the following technical solution: A variable frequency dynamic scanning method, the variable frequency dynamic scanning method comprising: Acquire radio frequency signals within the target frequency band, and perform high-speed sampling on the radio frequency signals to obtain analog sampled signals; The analog sampled signal is converted from analog to digital to obtain a digital signal; Perform spectrum analysis on the digital signal to extract its frequency characteristic information; The frequency characteristic information is input into the frequency control model to obtain frequency control commands; The output frequency of the local oscillator is adjusted according to the frequency control command, and a nonlinear frequency sweep is performed.

[0006] By adopting the above technical solution, dynamic modeling and frequency control command generation based on frequency characteristics can be achieved after receiving the target frequency band radio frequency signal. This makes the frequency scanning process no longer dependent on fixed step size or linear sequence, but generates nonlinear jump paths through frequency regulation model inference, thereby dynamically adjusting the output frequency of the local oscillator. During the scanning process, idle frequency bands are effectively skipped and active spectral regions are prioritized. Combined with the regulation strategy of the model output, fast response and intelligent frequency hopping control are achieved, improving the adaptability, accuracy and coverage efficiency of spectrum scanning.

[0007] In a preferred embodiment, this application can be further configured such that: the high-speed sampling of the radio frequency signal to obtain an analog sampled signal includes: By performing spectrum transformation processing, the spectrum changes of the radio frequency signal are analyzed, the signal energy values ​​of each sub-band within the target frequency band are extracted, and the energy change trend of each sub-band is determined based on the changes of the energy values ​​within a continuous time window. Based on the energy change trend, the sampling frequency band and sampling rate are determined. Based on the sampling frequency band and the sampling rate, the sampling channel is controlled to perform high-speed sampling on the radio frequency signal to generate the analog sampling signal.

[0008] By adopting the above technical solution, the energy change trend within the target frequency band can be extracted based on the spectrum transformation of the radio frequency signal, and the sampling frequency band and sampling rate can be dynamically determined based on this trend. This enables the sampling control process to have the ability to perceive and respond to spectrum changes, no longer relying on static configuration or fixed frequency band acquisition strategies. It can adaptively adjust the allocation of sampling resources according to the spectrum activity level, realize more targeted and flexible analog sampling scheduling control, and thus improve the sampling efficiency and the effectiveness of information acquisition in complex spectrum environments.

[0009] In a preferred embodiment, this application can be further configured such that: inputting the frequency characteristic information into the frequency control model to obtain frequency control commands includes: The frequency feature information is converted into a structured feature vector; The structured feature vector is input into the frequency control model for inference processing. The frequency control model projects the structured feature vector to a high-dimensional decision space through nonlinear feature mapping, and classifies the structured feature vector based on the frequency control decision boundary formed by training to obtain the analysis results. Based on the analysis results, the frequency control command is confirmed.

[0010] By adopting the above technical solution, intelligent reasoning processing of frequency control models can be realized based on structured frequency feature information. The spectrum state is projected to a high-dimensional decision space using nonlinear feature mapping, and spectrum state classification and judgment are completed based on the decision boundary formed by training. Furthermore, the frequency control command is confirmed by combining the analysis results, so that the frequency jump strategy is transformed from static configuration to dynamic generation based on model judgment, thereby improving the adaptability and accuracy of control commands.

[0011] In a preferred embodiment, this application can be further configured as follows: the structured feature vector is input into the frequency control model for inference processing, the frequency control model projects the structured feature vector to a high-dimensional decision space through nonlinear feature mapping, and classifies the structured feature vector based on the frequency control decision boundary formed during training to obtain analysis results, including: The structured feature vector is transformed using a nonlinear mapping function to obtain a high-dimensional feature embedding representation corresponding to the structured feature vector; The high-dimensional feature embedding representation is projected onto the high-dimensional decision space to obtain the position coordinates of the high-dimensional feature embedding representation in the high-dimensional decision space; The position coordinates are compared with the frequency control decision boundary formed by training to obtain relative relationship information; Based on the relative relationship information, the spectral state category to which the high-dimensional feature embedding representation belongs is determined, and the spectral state category is output as the analysis result.

[0012] By adopting the above technical solution, frequency feature vectors can be converted into high-dimensional feature embedding representations through nonlinear mapping, and spatially located in the high-dimensional decision space. By comparing them with the frequency control decision boundary formed by training, relative relationship information can be obtained, thereby achieving accurate classification and judgment of the current spectrum state. This replaces the traditional low-dimensional feature direct mapping and fixed rule matching method in spectrum recognition with a model discrimination method based on high-dimensional feature relationships, thereby improving the accuracy and generalization ability of spectrum state recognition.

[0013] In a preferred embodiment, this application can be further configured such that: adjusting the output frequency of the local oscillator according to the frequency control command and performing a nonlinear frequency sweep includes: According to the frequency control command, the target frequency point and the corresponding transition priority information of the target frequency point are obtained; Based on the target frequency point and the transition priority information, a nonlinear transition path is constructed, and the output frequency of the local oscillator is controlled according to the nonlinear transition path; During the execution of the nonlinear transition path, frequency points identified as idle or low-risk among the target frequency points are skipped. At the target frequency points that are not skipped, the output frequency of the local oscillator is controlled, and the frequency scan of the nonlinear transition is performed.

[0014] By adopting the above technical solution, the target frequency point and its jump priority information can be obtained according to the frequency control command. On this basis, a nonlinear jump path can be constructed. Combined with the dynamic identification result of the spectrum status, idle frequency points or low-risk frequency points can be skipped, so as to realize precise control and flexible jump of the output frequency of the local oscillator. This makes the frequency scanning process have the characteristics of non-fixed jump order, selective frequency band coverage, and dynamic response mechanism, thereby improving the flexibility and adaptability of frequency jump control.

[0015] In a preferred embodiment, this application may be further configured such that: the step of adjusting the output frequency of the local oscillator according to the frequency control command and performing a nonlinear frequency sweep further includes: During the frequency scanning process, real-time spectrum response information is continuously acquired, and the real-time spectrum response information is compared with historical feature information to obtain the comparison result. Based on the comparison result, the frequency control model is adjusted to obtain an updated frequency control model. Based on the updated frequency modulation model, the digital signal is identified and processed, and interference suppression is performed.

[0016] By adopting the above technical solution, real-time spectrum response information can be continuously acquired during frequency scanning and compared with historical features, thereby dynamically capturing changes in spectrum status and adjusting the frequency control model accordingly. The incremental update mechanism of the model enhances its adaptability to the current spectrum environment. Furthermore, based on the updated model, digital signal recognition and interference suppression operations are performed, realizing a closed-loop process of spectrum perception, model optimization, and control execution, thereby enhancing the system's real-time identification capability of interference signals and the adaptability of spectrum control.

[0017] In a preferred embodiment, this application can be further configured as follows: adjusting the frequency modulation model based on the comparison results to obtain an updated frequency modulation model includes: The spectrum state identification deviation is determined based on the comparison results, and an online updated sample set is constructed based on the spectrum state identification deviation. The frequency regulation model is incrementally trained based on the online updated sample set, and the frequency regulation model is adjusted to obtain the updated frequency regulation model.

[0018] By adopting the above technical solution, model deviations can be dynamically identified based on the differences between the spectrum state identification results and historical features. The identification deviations can be used to construct online updated samples that reflect the current spectrum state, thereby enabling continuous incremental training and parameter optimization of the frequency regulation model. This allows the model to adapt to new spectrum change trends without relying on full retraining, improving the model's responsiveness to environmental evolution and the accuracy of classification.

[0019] The second objective of this invention is achieved through the following technical solution: A variable frequency dynamic scanning device, the variable frequency dynamic scanning device comprising: The radio frequency signal acquisition module is used to acquire radio frequency signals within the target frequency band, and to perform high-speed sampling of the radio frequency signals to obtain analog sampled signals; The analog-to-digital conversion and digital signal generation module is used to convert the analog sampled signal into a digital signal. The spectrum feature extraction module is used to perform spectrum analysis on the digital signal and extract the frequency feature information of the digital signal; The frequency control model processing module is used to input the frequency characteristic information into the frequency control model to obtain frequency control commands; The nonlinear jump control module is used to adjust the output frequency of the local oscillator according to the frequency control command and to perform a nonlinear jump frequency scan.

[0020] By adopting the above technical solution, dynamic modeling and frequency control command generation based on frequency characteristics can be achieved after receiving the target frequency band radio frequency signal. This makes the frequency scanning process no longer dependent on fixed step size or linear sequence, but generates nonlinear jump paths through frequency regulation model inference, thereby dynamically adjusting the output frequency of the local oscillator. During the scanning process, idle frequency bands are effectively skipped and active spectral regions are prioritized. Combined with the regulation strategy of the model output, fast response and intelligent frequency hopping control are achieved, improving the adaptability, accuracy and coverage efficiency of spectrum scanning.

[0021] The above-mentioned objective three of this application is achieved through the following technical solution: A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the aforementioned variable frequency dynamic scanning method.

[0022] The fourth objective of this application is achieved through the following technical solution: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the aforementioned frequency conversion dynamic scanning method.

[0023] In summary, this application includes at least one of the following beneficial technical effects: 1. After receiving the target frequency band RF signal, it can realize dynamic modeling based on frequency characteristics and generate frequency control commands, so that the frequency scanning process no longer depends on fixed step size or linear sequence. Instead, it generates nonlinear jump paths through frequency regulation model reasoning, thereby dynamically adjusting the output frequency of the local oscillator. During the scanning process, it can effectively skip idle frequency bands and give priority to covering active spectrum areas. Combined with the regulation strategy of model output, it can realize fast response and intelligent frequency hopping control, and improve the adaptability, accuracy and coverage efficiency of spectrum scanning. 2. It can convert frequency feature vectors into high-dimensional feature embedding representations through nonlinear mapping, and spatially locate them in the high-dimensional decision space. By comparing them with the frequency control decision boundary formed by training, it can obtain relative relationship information, thereby achieving accurate classification and judgment of the current spectrum state. It replaces the traditional low-dimensional feature direct mapping and fixed rule matching method in spectrum recognition with a model discrimination method based on high-dimensional feature relationships, thereby improving the accuracy and generalization ability of spectrum state recognition. 3. It can continuously acquire real-time spectrum response information during frequency scanning and compare it with historical features, thereby dynamically capturing changes in spectrum status and adjusting the frequency control model accordingly. Through the incremental update mechanism of the model, it can improve the adaptability to the current spectrum environment. Furthermore, it can perform digital signal recognition and interference suppression operations based on the updated model, realizing a closed-loop process of spectrum perception, model optimization, and control execution, thereby enhancing the system's real-time identification capability of interference signals and the adaptability of spectrum control. Attached Figure Description

[0024] Figure 1 This is a flowchart of a frequency conversion dynamic scanning method according to an embodiment of this application; Figure 2 This is a flowchart illustrating the implementation of step S10 in a frequency conversion dynamic scanning method according to an embodiment of this application. Figure 3 This is a flowchart illustrating the implementation of step S40 in a frequency conversion dynamic scanning method according to an embodiment of this application. Figure 4 This is another implementation flowchart of step S402 in a frequency conversion dynamic scanning method according to an embodiment of this application; Figure 5 This is a flowchart illustrating the implementation of step S50 in a frequency conversion dynamic scanning method according to an embodiment of this application. Figure 6 This is a flowchart illustrating the implementation of step S50 in a frequency conversion dynamic scanning method according to an embodiment of this application. Figure 7This is a flowchart illustrating the implementation of step S504 in a frequency conversion dynamic scanning method according to an embodiment of this application. Figure 8 This is a schematic block diagram of a frequency conversion dynamic scanning device according to an embodiment of this application; Figure 9 This is a schematic diagram of a device according to one embodiment of this application. Detailed Implementation

[0025] The present application will be further described in detail below with reference to the accompanying drawings.

[0026] In one embodiment, such as Figure 1 As shown, this application discloses a variable frequency dynamic scanning method, which specifically includes the following steps: S10: Acquire the radio frequency signal within the target frequency band, perform high-speed sampling on the radio frequency signal, and obtain the analog sampled signal.

[0027] In this embodiment, the target frequency band refers to a continuous or discontinuous frequency range selected according to task requirements in a specific application scenario. This frequency range includes frequency points where effective radio frequency signals may exist, and can be dynamically adjusted according to the external spectrum environment, adaptive strategies, or preset configurations.

[0028] Specifically, after setting the scan start frequency and end frequency, the local oscillator is driven by the frequency control signal to output the corresponding frequency signal, which is then mixed with the received RF signal to obtain the intermediate frequency signal. The intermediate frequency signal is amplified by an adjustable gain amplifier and the interference components of non-target frequency bands are suppressed by a bandpass filter. The output analog signal is introduced into the sampling channel. The sampling rate is set to more than twice the signal bandwidth by configuring the sampling clock. The sampling window is divided in a uniform time interval manner. Multiple analog voltage values ​​are continuously acquired as sampling points in each sampling window. During the sampling process, the signal active range is determined by the relative amplitude change between adjacent sampling points, and the subsequent sampling duration is adjusted based on the activity level. Throughout the process, the sampling clock is kept synchronized with the external frequency to ensure timing consistency, and finally, an analog sampling signal that meets the requirements of dynamic range and time resolution is obtained.

[0029] S20: Convert the analog sampled signal to a digital signal.

[0030] Specifically, the analog sampling signal is sequentially input into the analog-to-digital converter circuit. At each sampling moment, the sample-and-hold unit locks the instantaneous voltage value of the input signal. A reference voltage range is set and divided into multiple amplitude segments according to the selected quantization bit depth. A comparator is used to compare the current input voltage with each amplitude segment to determine its range and map the range to the corresponding digital encoded value. The above analog-to-digital conversion process is repeated throughout the entire sampling window so that each analog sampling point is quantized into a unique digital representation. After the conversion is completed, all digital encoded values ​​are output to the buffer interface in the sampling order to obtain a complete digital signal.

[0031] S30: Perform spectrum analysis on the digital signal to extract the frequency characteristic information of the digital signal.

[0032] Specifically, the digital signal is divided into several data segments according to the sampling window. Each data segment contains a fixed number of digital sampling points. In each data segment, the fast Fourier transform algorithm is applied to transform the time-domain sampling sequence, converting it into the corresponding frequency-domain complex spectrum value sequence. The amplitude spectrum of each frequency point is calculated by taking the modulus of the complex spectrum, and the amplitude spectrum is correlated with the frequency coordinates to obtain the complete spectrum distribution. Based on this, sub-frequency bands are divided by setting frequency intervals, and the average amplitude, maximum amplitude, rate of change, and other statistics in each sub-frequency band are calculated. The intensity distribution, energy concentration, and spectral shape change trend of the frequency domain signal in different frequency ranges are further analyzed to obtain frequency characteristic information for subsequent frequency control decisions.

[0033] S40: Input the frequency characteristic information into the frequency control model to obtain the frequency control command.

[0034] In this embodiment, the frequency control model is constructed using a support vector machine (SVM) model. The model maps the input structured feature vector to a high-dimensional feature space through a nonlinear kernel function, and determines the category of the input data based on the optimal classification hyperplane constructed from the training samples. The model supports an incremental training mechanism, which can adjust the local boundary according to the updated samples when the spectrum state changes, thereby realizing dynamic reasoning of the frequency control strategy.

[0035] Specifically, the frequency feature information extracted during spectrum analysis is structured and organized according to a preset dimension to construct a feature vector containing multiple frequency attributes. This feature vector includes numerical indicators such as signal amplitude, spectral profile, and energy distribution pattern at different frequency points. The organized feature vector is then fed into the frequency control model as input data. The frequency control model is pre-built through training and contains a nonlinear mapping function and decision rule structure. After the input feature vector is input, feature projection is first performed in the mapping space. Then, the input is inferred and judged based on the classification boundary or fitting function in the model. During the inference process, the category pattern or regression trend formed by historical training samples is combined to match the characteristics of the current input data. Finally, a frequency control command containing the jump frequency point and frequency modulation priority is generated to obtain the parameter output used to drive the frequency jump control behavior.

[0036] S50: Adjusts the output frequency of the local oscillator according to the frequency control command and performs a non-linear frequency sweep.

[0037] Specifically, the frequency point sequence and hopping sequence information contained in the frequency control command are analyzed. The specific frequency value of each hopping frequency point is extracted from the frequency point sequence, and a hopping path is established according to the specified hopping sequence. This path is used as the frequency control reference. The reference input of the local oscillator is dynamically adjusted by the control signal to change its output frequency. During the hopping execution, the frequency points are not traversed sequentially according to a linear law, but the oscillator frequency output is switched non-equidistantly according to the hopping sequence. After each hopping, the oscillator frequency is kept stable until the preset hold time ends, and then the next hopping frequency point is entered. If the hopping path contains a frequency point that is identified as having low interference intensity or being in an idle state, the frequency point is skipped and the hold phase is not entered. Finally, the non-continuous frequency scanning operation within the frequency range selected by the hopping path is completed to obtain the frequency hopping execution result covering the active region of the spectrum.

[0038] In one embodiment, such as Figure 2 As shown, in step S10, i.e., high-speed sampling of the radio frequency signal to obtain an analog sampling signal, the following steps are included: S101: Through spectrum transformation processing, analyze the spectrum changes of the radio frequency signal, extract the signal energy values ​​of each sub-band within the target frequency band, and determine the energy change trend of each sub-band based on the changes in energy values ​​within a continuous time window.

[0039] Specifically, the radio frequency signal is processed in frames within a specified time period. Each frame of signal is windowed and smoothed using a window function before being input to the frequency domain transformation algorithm module. The fast Fourier transform is used to convert the time domain signal of each frame into a frequency domain complex spectrum. The amplitude component in the transformed spectrum is obtained, and the amplitude is squared to calculate the power spectral density value. The entire spectrum is divided into multiple continuous sub-bands with a fixed bandwidth. The power values ​​of the frequency points in each sub-band are weighted and averaged to estimate the total energy value of the sub-band. The above steps are repeated in multiple consecutive time frames to form an energy value sequence. The energy change of each sub-band between adjacent time frames is calculated by difference, and the change is smoothed and filtered to suppress short-term fluctuations. Finally, the energy value evolution trend of each sub-band within the entire time window is fitted and estimated to obtain the energy change trend reflecting the spectrum activity state.

[0040] S102: Based on the energy change trend, determine the sampling frequency band and sampling rate, and based on the sampling frequency band and sampling rate, control the sampling channel to perform high-speed sampling of the radio frequency signal to generate an analog sampling signal.

[0041] Specifically, a sampling priority ranking table is constructed based on indicators such as the energy growth rate, fluctuation degree, and active duration of each sub-band reflected in the energy change trend. The sub-bands ranked higher are selected as sampling bands, and appropriate sampling rates are matched according to their energy change rates. The higher the energy change rate, the higher the allocated sampling rate. The activity level is mapped to the preset sampling rate level value through a lookup table. The sampling band and the corresponding sampling rate are combined to form a sampling scheduling instruction. The scheduling instruction is used to drive the sampling channel to complete frequency switching and rate setting. Then, the frequency selection circuit and sampling control unit are configured according to the scheduling instruction so that the radio frequency signal is continuously collected in the selected frequency band. During the sampling process, a synchronous clock signal is used to maintain a stable rate, and the input signal is intercepted at equal intervals. Finally, a continuous-time analog sampling signal containing the voltage amplitude of each sampling point is output to obtain an analog sampling signal that matches the dynamic characteristics of the spectrum.

[0042] In one embodiment, such as Figure 3 As shown, in step S40, the frequency characteristic information is input into the frequency control model to obtain frequency control commands, including: S401: Convert frequency feature information into structured feature vectors.

[0043] Specifically, effective feature elements for modeling are extracted from the frequency characteristic information obtained from spectrum analysis, including frequency domain amplitude distribution, energy differences between frequency points, spectral line density, dominant frequency position and its rate of change, etc. All feature elements are concatenated in a set order to form an original feature sequence. Each feature value in the original sequence is processed by normalization according to its physical quantity dimension to map its numerical range to a unified interval. The processed feature values ​​are arranged in sequence to form a feature vector. During the construction process, a fixed-dimensional template is used to maintain the consistency of the vector structure for different samples. Missing values ​​are filled with default placeholders. Finally, the frequency characteristic information is expressed as a fixed-length, structured feature vector that can be input into the frequency control model to obtain a standardized vector that can be used for model inference.

[0044] S402: The structured feature vectors are input into the frequency control model for inference processing. The frequency control model projects the structured feature vectors to a high-dimensional decision space through nonlinear feature mapping, and classifies the structured feature vectors based on the frequency control decision boundary formed during training to obtain the analysis results.

[0045] Specifically, structured feature vectors are loaded as input data into the inference interface of the frequency control model. The model has a pre-set nonlinear mapping function to map the input vector from the original feature space to a higher-dimensional feature space. During the mapping process, the activation function performs nonlinear transformation on the feature values ​​of each dimension to construct a high-dimensional representation consistent with the feature distribution of the training data. This high-dimensional representation is then sent to the classification and discrimination unit. The decision boundary model formed during the training phase is pre-loaded in the discrimination unit. This decision boundary is composed of multiple classification hyperplanes and is used to distinguish the different spectral state categories corresponding to the frequency control strategy. By calculating the positional relationship between the high-dimensional feature representations and each decision boundary, the category label is determined, and the category is output as the inference result to obtain the analysis results used for subsequent jump control decisions.

[0046] S403: Based on the analysis results, confirm the frequency control command.

[0047] Specifically, the category labels in the analysis results are indexed and matched with the preset frequency control strategy table. The frequency control strategy table configures a set of control parameters for each type of spectrum state, including a target frequency point list, hopping priority identifier, and hold duration parameter. By reading the strategy entry corresponding to the current category label, the parameters in the strategy are extracted to form a frequency control dataset. Each parameter in the dataset is filled into the control instruction format template to generate a complete instruction frame. The instruction frame sets the data encoding format and transmission order according to the control field definition. After the instruction is generated, it is encapsulated into a standard format by the encoder for use by the frequency hopping control loop to obtain a frequency control instruction that matches the current spectrum state.

[0048] In one embodiment, such as Figure 4 As shown, in step S402, the structured feature vectors are input into the frequency control model for inference processing. The frequency control model projects the structured feature vectors to a high-dimensional decision space through nonlinear feature mapping, and classifies the structured feature vectors based on the frequency control decision boundary formed during training, obtaining the analysis results, including: S4021: Transform the structured feature vector using a nonlinear mapping function to obtain the high-dimensional feature embedding representation corresponding to the structured feature vector.

[0049] Specifically, the input layer of the frequency modulation model receives a structured feature vector and inputs the values ​​of each dimension of the vector to the first transformation unit of the mapping network. In this unit, each dimension of the input is transformed by a preset nonlinear function. The nonlinear function can be implemented in the form of an activation function, such as using a hyperbolic tangent function to compress and map the input value or using a rectified linear function to truncate negative values. The result after processing by the first transformation unit is used as an intermediate expression and continues to be input to the subsequent multi-layer transformation structure. In each layer, the composite operation of linear weighting and nonlinear mapping is repeatedly performed, so that the original vector gradually expands its dimensions and captures the nonlinear relationship between potential features during the transformation process. Finally, a high-dimensional vector representation with fixed dimensions is formed in the output layer. This high-dimensional vector contains the representation of the original features after mapping in multi-dimensional space, so as to obtain the high-dimensional feature embedding representation corresponding to the structured feature vector.

[0050] S4022: Project the high-dimensional feature embedding representation onto the high-dimensional decision space to obtain the position coordinates of the high-dimensional feature embedding representation in the high-dimensional decision space.

[0051] Specifically, the high-dimensional feature embedding representation generated by the nonlinear mapping function is input into the projection calculation unit inside the frequency control model. During the projection calculation, each dimension of the embedding feature is normalized according to the feature space mapping rules constructed during the model training phase. Then, the normalized high-dimensional vector is multiplied by the feature space basis vectors corresponding to the training data. The projection value of the vector in each dimension basis vector direction is calculated by linear combination, thereby obtaining the geometric position of the embedding representation relative to each classification boundary in the high-dimensional decision space. Finally, a vector result with fixed coordinate values ​​in the multi-dimensional space is formed. This vector contains the spatial position index of the high-dimensional feature embedding representation in each feature dimension, so as to obtain the position coordinates of the high-dimensional feature embedding representation in the high-dimensional decision space.

[0052] S4023: Compare the position coordinates with the frequency control decision boundary formed during training to obtain relative relationship information.

[0053] Specifically, the position coordinates in the high-dimensional decision space are input into the boundary comparison unit of the frequency control model. Spatial relationship judgments are made against multiple frequency control decision boundaries generated during the training phase. Each decision boundary is expressed in the form of a hyperplane, corresponding to a specific category classification region. By substituting the position coordinates into the discriminant function of each hyperplane, the distance from the coordinate point to each hyperplane and its sign are calculated. The sign of the distance determines which side of the boundary the point is located on. The discrimination results of all boundaries are combined into a relation vector. Each item in the relation vector corresponds to the relative positional relationship of a boundary. The absolute value of the distance is used to represent the degree of proximity, while the sign information is retained to represent the belonging trend, so as to obtain the relative relationship information reflecting the classification relationship between the position coordinates and each frequency control decision boundary.

[0054] S4024: Based on the relative relationship information, determine the spectral state category to which the high-dimensional feature embedding representation belongs, and output the spectral state category as the analysis result.

[0055] Specifically, the boundary discrimination results in the relative relationship information are input into the classification decision unit of the frequency control model. By searching the preset category mapping table, the relative position combination pattern is matched with the corresponding spectrum state category. If multiple categories overlap or conflict, the category with the smallest absolute value of the boundary distance is selected as the main judgment result. At the same time, for inputs with relatively ambiguous classification boundaries, a weighted voting method can be used for category fusion judgment. During the fusion process, the weights are assigned according to the reciprocal of the boundary distances, and the candidate categories are weighted and sorted. The category with the highest score is selected as the final judgment output. The category number or label is encoded as the identification and attribution of the high-dimensional feature embedding representation of the current input, so as to obtain the spectrum state analysis results that can be used to drive the frequency control strategy.

[0056] In one embodiment, such as Figure 5 As shown, in step S50, that is, adjusting the output frequency of the local oscillator according to the frequency control command and performing a nonlinear frequency sweep, the following steps are included: S501: Based on the frequency control command, obtain the target frequency point and the corresponding transition priority information of the target frequency point.

[0057] Specifically, the control fields contained in the frequency control command are parsed to identify the data segment containing frequency transition parameters. The numerical codes of multiple frequency points are read item by item in the data segment, and the transition priority number corresponding to each frequency point is extracted. The extracted frequency values ​​are converted into the actual operating frequency according to the unit conversion. The corresponding priority number reflects the importance of its transition order according to the preset rules. If the frequency control command uses a compressed index method to encode the pairing relationship between frequency points and priorities, the index is restored to the specific parameters through the decoding table. Finally, a set of frequency point sequences sorted by transition priority is formed, with each frequency point corresponding to its priority one-to-one, so as to obtain the target frequency point and the transition priority information corresponding to each target frequency point.

[0058] S502: Based on the target frequency point and transition priority information, a nonlinear transition path is constructed, and the output frequency of the local oscillator is controlled according to the nonlinear transition path.

[0059] Specifically, the target frequency points extracted in sequence are paired with their corresponding switching priorities. During the processing, the frequency switching priority values ​​are sorted in ascending order or according to rules to generate a frequency switching sequence table. After sorting, a random disturbance factor or nonlinear adjustment function is introduced to intervene and rearrange some switching sequences, so that the switching path does not have a linear pattern as a whole. Then, the frequency values ​​of each frequency point in the path are transmitted to the local oscillator control interface one by one as control inputs. After each switching command is issued, the oscillator output frequency is adjusted by frequency control voltage or phase-locked loop configuration parameters, so that the oscillator switches sequentially between multiple target frequencies. Throughout the switching process, the frequency switching interval and hold time are kept synchronized to obtain a local oscillator output frequency sequence that changes in a nonlinear order.

[0060] S503: During the execution of the nonlinear jump path, skip the target frequency points that are identified as idle or low-risk. At the target frequency points that are not skipped, control the output frequency of the local oscillator and perform a nonlinear jump frequency scan.

[0061] Specifically, during the sequential execution of the nonlinear hopping path, the spectral status index of the current target frequency point is determined by real-time spectrum monitoring results. Idle frequency points are marked as not participating in the scan, and low-risk frequency points are determined whether to skip based on priority setting conditions. Skipped frequency points do not generate frequency setting commands or enter the hopping maintenance phase. At target frequency points that are not skipped, the local oscillator is controlled to set its output frequency according to the hopping path sequence, and a certain time window is maintained during the set frequency maintenance phase. The scanning acquisition process at the corresponding frequency is then initiated. The scanning process includes signal reception, analog sampling, and frequency domain processing preparation operations. This control hopping and scanning process is repeated until the hopping path ends, so as to obtain the dynamically filtered frequency hopping sequence and the frequency scanning process completed at the corresponding frequency point.

[0062] In one embodiment, such as Figure 6 As shown, after step S50, that is, adjusting the output frequency of the local oscillator according to the frequency control command and performing a nonlinear frequency sweep, the process also includes: S504: During the frequency scanning process, real-time spectrum response information is continuously acquired, and the real-time spectrum response information is compared with historical feature information to obtain the comparison results. Based on the comparison results, the frequency control model is adjusted to obtain the updated frequency control model.

[0063] Specifically, during the frequency scanning process, response information such as the spectral amplitude distribution, frequency component intensity, and spectral structure at the current frequency point is extracted in real time as a current spectrum snapshot. This snapshot is cached within a time window, and feature quantities are extracted and compared item by item with feature quantities at the same frequency point in the historical record. During the comparison, indicators such as frequency peak shift, energy change amplitude, and spectral similarity are calculated as comparison parameters. The comparison parameters are organized into comparison feature vectors and input to the model update judgment unit. Based on preset conditions, it is determined whether to trigger the model adjustment operation. If the update conditions are met, a new sample is constructed using the current snapshot information. The original frequency control model is locally incrementally updated by expanding the sample or fine-tuning the parameters, so that the model adapts to the current spectrum change trend on the basis of the original structure, so as to obtain a frequency control model whose spectrum sensing capability is dynamically updated with the environment.

[0064] S505: Based on the updated frequency control model, it identifies and processes digital signals and performs interference suppression operations.

[0065] Specifically, the digital signal acquired by the current frequency scan is input into the updated frequency control model in the order of sampling time. Feature extraction and mapping operations are performed within the model to convert the spectral features in the digital signal into a structured representation. Based on the latest model boundary, the spectral state category to which it belongs is determined. At the same time, for frequency bands identified as interference signals, a matching suppression strategy is selected according to the interference level or interference type parameters output by the model. Suppression strategies include band-stop filtering, spectral reduction, amplitude suppression, etc. The corresponding filter parameters or control thresholds in the processing flow are dynamically adjusted according to the strategy requirements to adjust the amplitude or remove the frequency domain of the signal components in the target frequency band, retain the effective part of the target signal, and complete the digital processing correction of the frequency band with interference.

[0066] In one embodiment, such as Figure 7 As shown, in step S504, based on the comparison results, the frequency control model is adjusted to obtain an updated frequency control model, including: S5041: Determine the spectrum state identification deviation based on the comparison results, and construct an online updated sample set based on the spectrum state identification deviation.

[0067] Specifically, quantitative indicators such as the dominant frequency offset, amplitude difference, and spectral inconsistency degree of the current spectral response are extracted from the comparison results. These indicators are compared item by item with the benchmark features corresponding to the model prediction results under historical spectral conditions. The spectral state identification deviation value is obtained through residual calculation. When the identification deviation value exceeds the set threshold, the corresponding frequency point is determined as a sample with inconsistent state identification. Feature extraction processing is performed on the spectral response of the frequency point in the current scanning period, and the corresponding actual observation category or spectral label is labeled. The extracted feature data and label data are combined into input-output pairs and cached in the training sample area according to the receiving time order. The upper limit of the number of samples and the update window period are set as needed to obtain an online updated sample set that can be used for incremental model learning.

[0068] S5042: Incremental training of the frequency regulation model is performed based on the online updated sample set to adjust the frequency regulation model and obtain the updated frequency regulation model.

[0069] Specifically, the input feature vectors and corresponding spectral state labels contained in the online updated sample set are input to the training interface of the frequency regulation model. During training, the original parameter structure of the model remains unchanged. By freezing some of the weights of the stable layer, gradient updates are performed only on the parameters of the output layer or key mapping layer. Mini-batch sample grouping is used to control the amount of data in each training round. The error value between the current predicted output and the true label is calculated through the loss function. Then, the updatable parameters are adjusted based on the error value through backpropagation. During the training iteration, an early stopping mechanism is enabled to monitor the changing trend of the verification error and avoid the model from overfitting to new samples. A historical parameter moving average strategy is introduced to balance the degree of fusion between the old and new parameters. Finally, the local weights within the model structure are adjusted to obtain an updated frequency regulation model with the ability to adapt to the current spectrum.

[0070] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0071] In one embodiment, a variable frequency dynamic scanning device is provided, which corresponds one-to-one with the variable frequency dynamic scanning method described in the above embodiments. For example... Figure 8 As shown, this frequency conversion dynamic scanning device includes an RF signal acquisition module, an analog-to-digital conversion and digital signal generation module, a spectrum feature extraction module, a frequency control model processing module, and a nonlinear jump control module. Detailed descriptions of each functional module are as follows: The radio frequency signal acquisition module is used to acquire radio frequency signals within the target frequency band, perform high-speed sampling of the radio frequency signals, and obtain analog sampled signals. The analog-to-digital conversion and digital signal generation module is used to convert analog sampled signals into digital signals. The spectrum feature extraction module is used to perform spectrum analysis on digital signals and extract frequency feature information of digital signals; The frequency control model processing module is used to input frequency characteristic information into the frequency control model to obtain frequency control commands; The nonlinear jump control module is used to adjust the output frequency of the local oscillator according to the frequency control command and to perform nonlinear jump frequency scanning.

[0072] Optionally, the radio frequency signal acquisition module includes: The spectrum trend analysis submodule is used to analyze the spectrum changes of radio frequency signals through spectrum transformation processing, extract the signal energy values ​​of each sub-band within the target frequency band, and determine the energy change trend of each sub-band based on the changes of energy values ​​within a continuous time window. The adaptive sampling control submodule is used to determine the sampling frequency band and sampling rate according to the energy change trend. Based on the sampling frequency band and sampling rate, it controls the sampling channel to perform high-speed sampling of the radio frequency signal and generate an analog sampling signal.

[0073] Optionally, the frequency modulation model processing module includes: The feature vector construction submodule is used to convert frequency feature information into structured feature vectors; The frequency decision reasoning submodule is used to input structured feature vectors into the frequency control model for reasoning processing. The frequency control model projects the structured feature vectors to a high-dimensional decision space through nonlinear feature mapping, and classifies the structured feature vectors based on the frequency control decision boundary formed during training to obtain the analysis results. The frequency command generation submodule is used to confirm frequency control commands based on the analysis results.

[0074] Optional, the frequency decision reasoning submodule includes: The nonlinear feature mapping unit is used to transform the structured feature vector using a nonlinear mapping function to obtain the high-dimensional feature embedding representation corresponding to the structured feature vector; The feature space projection unit is used to project the high-dimensional feature embedding representation onto the high-dimensional decision space to obtain the position coordinates of the high-dimensional feature embedding representation in the high-dimensional decision space. The boundary comparison and discrimination unit is used to compare the position coordinates with the frequency control decision boundary formed during training to obtain relative relationship information; The spectrum state classification unit is used to determine the spectrum state category to which the high-dimensional feature embedding representation belongs based on relative relationship information, and outputs the spectrum state category as the analysis result.

[0075] Optional, the nonlinear jump control module includes: The frequency hopping parameter parsing submodule is used to obtain the target frequency point and the corresponding hopping priority information based on the frequency control command. The nonlinear jump path generation submodule is used to construct a nonlinear jump path based on the target frequency point and jump priority information, and control the output frequency of the local oscillator according to the nonlinear jump path; The dynamic jump execution submodule is used to skip frequency points identified as idle or low-risk in the target frequency points during the execution of the nonlinear jump path. At the target frequency points that are not skipped, the output frequency of the local oscillator is controlled and the frequency scan of the nonlinear jump is performed.

[0076] Optionally, the nonlinear jump control module may also include... The spectrum feedback comparison and model update submodule is used to continuously acquire real-time spectrum response information during frequency scanning, compare the real-time spectrum response information with historical feature information to obtain comparison results, and adjust the frequency control model based on the comparison results to obtain the updated frequency control model. The model-driven identification and interference suppression submodule is used to identify and process digital signals based on the updated frequency modulation model and perform interference suppression operations.

[0077] Optionally, the spectrum feedback alignment and model update submodule includes: The identification deviation analysis and sample construction unit is used to determine the spectrum state identification deviation based on the comparison results, and to construct an online updated sample set based on the spectrum state identification deviation; The incremental training and parameter update unit is used to incrementally train the frequency regulation model based on the online updated sample set, adjust the frequency regulation model, and obtain the updated frequency regulation model.

[0078] For specific limitations regarding a variable frequency dynamic scanning device, please refer to the limitations of a variable frequency dynamic scanning method described above, which will not be repeated here. Each module in the aforementioned variable frequency dynamic scanning device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0079] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a variable frequency dynamic scanning method.

[0080] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: The radio frequency signal within the target frequency band is acquired, and the radio frequency signal is sampled at high speed to obtain an analog sampled signal; The analog sampled signal is converted from analog to digital to obtain a digital signal; Perform spectrum analysis on digital signals to extract frequency characteristic information; Frequency characteristic information is input into the frequency control model to obtain frequency control commands; The output frequency of the local oscillator is adjusted according to the frequency control command, and a non-linear frequency sweep is performed.

[0081] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: The radio frequency signal within the target frequency band is acquired, and the radio frequency signal is sampled at high speed to obtain an analog sampled signal; The analog sampled signal is converted from analog to digital to obtain a digital signal; Perform spectrum analysis on digital signals to extract frequency characteristic information; Frequency characteristic information is input into the frequency control model to obtain frequency control commands; The output frequency of the local oscillator is adjusted according to the frequency control command, and a non-linear frequency sweep is performed.

[0082] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0083] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0084] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A variable frequency dynamic scanning method, characterized in that, The variable frequency dynamic scanning method includes: Acquire radio frequency signals within the target frequency band, and perform high-speed sampling on the radio frequency signals to obtain analog sampled signals; The analog sampled signal is converted from analog to digital to obtain a digital signal; Perform spectrum analysis on the digital signal to extract its frequency characteristic information; The frequency characteristic information is input into the frequency control model to obtain frequency control commands; The output frequency of the local oscillator is adjusted according to the frequency control command, and a nonlinear frequency sweep is performed.

2. The variable frequency dynamic scanning method according to claim 1, characterized in that, The step of high-speed sampling of the radio frequency signal to obtain an analog sampled signal includes: By performing spectrum transformation processing, the spectrum changes of the radio frequency signal are analyzed, the signal energy values ​​of each sub-band within the target frequency band are extracted, and the energy change trend of each sub-band is determined based on the changes of the energy values ​​within a continuous time window. Based on the energy change trend, the sampling frequency band and sampling rate are determined. Based on the sampling frequency band and the sampling rate, the sampling channel is controlled to perform high-speed sampling on the radio frequency signal to generate the analog sampling signal.

3. The variable frequency dynamic scanning method according to claim 1, characterized in that, The step of inputting the frequency characteristic information into the frequency control model to obtain frequency control commands includes: The frequency feature information is converted into a structured feature vector; The structured feature vector is input into the frequency control model for inference processing. The frequency control model projects the structured feature vector to a high-dimensional decision space through nonlinear feature mapping, and classifies the structured feature vector based on the frequency control decision boundary formed by training to obtain the analysis results. Based on the analysis results, the frequency control command is confirmed.

4. The variable frequency dynamic scanning method according to claim 3, characterized in that, The structured feature vector is input into the frequency control model for inference processing. The frequency control model projects the structured feature vector to a high-dimensional decision space through nonlinear feature mapping, and classifies the structured feature vector based on the frequency control decision boundary formed during training, obtaining analysis results, including: The structured feature vector is transformed using a nonlinear mapping function to obtain a high-dimensional feature embedding representation corresponding to the structured feature vector; The high-dimensional feature embedding representation is projected onto the high-dimensional decision space to obtain the position coordinates of the high-dimensional feature embedding representation in the high-dimensional decision space; The position coordinates are compared with the frequency control decision boundary formed by training to obtain relative relationship information; Based on the relative relationship information, the spectral state category to which the high-dimensional feature embedding representation belongs is determined, and the spectral state category is output as the analysis result.

5. The variable frequency dynamic scanning method according to claim 1, characterized in that, The step of adjusting the output frequency of the local oscillator according to the frequency control command and performing a nonlinear frequency sweep includes: According to the frequency control command, the target frequency point and the corresponding transition priority information of the target frequency point are obtained; Based on the target frequency point and the transition priority information, a nonlinear transition path is constructed, and the output frequency of the local oscillator is controlled according to the nonlinear transition path; During the execution of the nonlinear transition path, frequency points identified as idle or low-risk among the target frequency points are skipped. At the target frequency points that are not skipped, the output frequency of the local oscillator is controlled, and the frequency scan of the nonlinear transition is performed.

6. The variable frequency dynamic scanning method according to claim 1, characterized in that, The step of adjusting the output frequency of the local oscillator according to the frequency control command and performing a nonlinear frequency sweep further includes: During the frequency scanning process, real-time spectrum response information is continuously acquired, and the real-time spectrum response information is compared with historical feature information to obtain the comparison result. Based on the comparison result, the frequency control model is adjusted to obtain an updated frequency control model. Based on the updated frequency modulation model, the digital signal is identified and processed, and interference suppression is performed.

7. The variable frequency dynamic scanning method according to claim 6, characterized in that, The step of adjusting the frequency control model based on the comparison results to obtain an updated frequency control model includes: The spectrum state identification deviation is determined based on the comparison results, and an online updated sample set is constructed based on the spectrum state identification deviation. The frequency regulation model is incrementally trained based on the online updated sample set, and the frequency regulation model is adjusted to obtain the updated frequency regulation model.

8. A variable frequency dynamic scanning device, characterized in that, The variable frequency dynamic scanning device includes: The radio frequency signal acquisition module is used to acquire radio frequency signals within the target frequency band, and to perform high-speed sampling of the radio frequency signals to obtain analog sampled signals; The analog-to-digital conversion and digital signal generation module is used to convert the analog sampled signal into a digital signal. The spectrum feature extraction module is used to perform spectrum analysis on the digital signal and extract the frequency feature information of the digital signal; The frequency control model processing module is used to input the frequency characteristic information into the frequency control model to obtain frequency control commands; The nonlinear jump control module is used to adjust the output frequency of the local oscillator according to the frequency control command and to perform a nonlinear jump frequency scan.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the frequency conversion dynamic scanning method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the frequency conversion dynamic scanning method as described in any one of claims 1 to 7.

Images