Sensing radio front-end radio frequency signal detection system

Abstract

The invention relates to the technical field of wireless communication, and discloses a sensing radio front-end radio frequency signal detection system, which comprises a broadband radio frequency front-end module used for receiving radio frequency signals in an environment and supporting multi-band signal capture and anti-interference preprocessing; the analog-to-digital conversion module is used for converting the radio frequency signal into a digital signal by adopting an ADC (Analog to Digital Converter) with 14-bit resolution and 2GSPS sampling rate; the signal processing module is connected to the output end of the analog-to-digital conversion module and is used for performing spectrum analysis, feature extraction and signal detection on the digital signal and outputting a detection result; and a synchronous control bus. According to the method, energy detection, cyclostationary feature analysis and a deep learning classification algorithm are fused, a time-frequency domain joint analysis framework is constructed, and the time domain transient feature extraction capability of the 1D-CNN and the frequency domain long-range dependence modeling advantage of Transform are combined, so that high-precision recognition of complex modulation signals is realized, and the detection sensitivity and the anti-interference capability are improved.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, and in particular to a cognitive radio front-end radio frequency signal detection system. Background Art

[0002] With the rapid development of wireless communication technology, efficient spectrum utilization and reliable signal detection in complex electromagnetic environments have become core challenges. In modern communication scenarios, signal types are diverse, dynamic, and hidden, and face challenges such as high-density spectrum occupancy, strong interference noise, and low signal-to-noise ratio signal capture. Existing systems face bottlenecks in dynamic noise suppression, multi-mode signal classification, and cross-module timing coordination, making it difficult to meet the complex requirements of real-time spectrum sensing, accurate signal analysis, and adaptive interference suppression. Traditional RF signal detection systems often use fixed-architecture analog mixers and discrete filter chains for signal processing.

[0003] However, in current technologies, fixed-architecture analog mixers and discrete filter chains for signal processing are unable to cope with dynamic noise interference and multi-mode signal aliasing in complex electromagnetic environments. Summary of the Invention

[0004] In order to overcome the above shortcomings, the present invention provides a cognitive radio front-end RF signal detection system, which aims to improve the dynamic noise interference and multi-mode signal aliasing problems that are difficult to deal with in complex electromagnetic environments.

[0005] In a first aspect, the present invention provides the following technical solution: a cognitive radio front-end radio frequency signal detection system, comprising:

[0006] Wideband RF front-end module, used to receive RF signals in the environment, supporting multi-band signal capture and anti-interference preprocessing;

[0007] The analog-to-digital conversion module uses an ADC with 14-bit resolution and 2GSPS sampling rate to convert the RF signal into a digital signal;

[0008] a signal processing module, connected to the output end of the analog-to-digital conversion module, for performing spectrum analysis, feature extraction and signal detection on the digital signal and outputting the detection result;

[0009] Synchronous control bus, based on the time-sensitive network protocol to coordinate the operation timing of each module, and synchronize the real-time RF front-end signal reception, analog-to-digital conversion and signal processing;

[0010] The wideband RF front-end module, analog-to-digital conversion module and signal processing module are connected in series in sequence and closed-loop controlled via a synchronous control bus.

[0011] Preferably, the broadband radio frequency front-end module includes:

[0012] The photon-assisted down-conversion unit down-converts the RF signal to an intermediate frequency through a lithium niobate optical modulator, which is used for mixing-free processing of high-frequency signals.

[0013] Metamaterial reconfigurable antenna, based on a graphene patch array, dynamically adjusts the radiation pattern with a gain of ≥12dBi for directional reception of target signals;

[0014] GaN low-noise amplifier, used to amplify received RF signals and reduce signal link noise;

[0015] The anti-aliasing filter group is used to filter out out-of-band interference signals and retain valid signals within the target frequency band.

[0016] Preferably, the analog-to-digital conversion module includes:

[0017] The clock source unit generates a synchronous clock signal based on a phase-locked loop to control the ADC sampling timing;

[0018] The quantizer unit converts the analog signal into a digital signal through the Delta-Sigma architecture;

[0019] Digital interface unit, supporting JESD204B protocol, used to transmit digital signals to the signal processing module;

[0020] The oversampling mode unit achieves 8 times oversampling through interpolation filters to improve signal quantization accuracy.

[0021] Preferably, the signal processing module includes:

[0022] The digital down-conversion unit converts the intermediate frequency signal into a baseband I / Q signal through the orthogonal local oscillator signal generated by the orthogonal mixer and the digitally controlled oscillator. The signal is then decimated and down-sampled through a multi-stage cascaded CIC filter and FIR compensation filter to output a low-rate baseband signal.

[0023] The fast Fourier transform unit performs a 4096-point FFT operation on the baseband signal, suppresses spectrum leakage through the Blackman-Harris window function, and generates a spectrum density map. The FFT calculation formula is:

[0024] Where N = 4096 is the number of FFT points, x(n) is the input baseband signal time domain sequence, and w(n) is the Blackman-Harris window function;

[0025] A detection algorithm unit performs energy detection, cyclostationary feature analysis, and deep learning-based signal classification on the spectrum density map. The following algorithms are performed on the spectrum density map:

[0026] Energy detection, the detection threshold is set by the dynamic threshold formula: Among them, T is the detection threshold, α is the false alarm rate factor, is the estimated value of the sliding window noise power;

[0027] Cyclostationary characteristic analysis: Where α is the cycle frequency, τ is the time delay, and K is the number of signal samples;

[0028] Deep learning classification: implemented through a hybrid network, including:

[0029] 1D-CNN convolution formula: Extract time domain transient features, where w k is the convolution kernel weight, b is the bias term, x t+k is the t+kth sampling value of the time domain baseband signal.

[0030] Preferably, the hybrid network also includes a Transformer self-attention mechanism to extract long-range dependency features in the frequency domain, concatenate the time domain feature vector and the frequency domain feature vector in the channel dimension, and output the probability distribution of the signal type through a softmax function.

[0031] Preferably, the synchronous control bus includes:

[0032] The clock source unit generates a global clock reference signal based on the GPS disciplined crystal oscillator and synchronizes the local clocks of each module through a phase-locked loop;

[0033] A delay compensation mechanism inserts timestamps between the RF front-end, analog-to-digital conversion module, and signal processing module, dynamically adjusting transmission delays in conjunction with data buffer queues to compensate for link jitter and temperature drift.

[0034] Priority scheduling strategy assigns transmission priority based on data type, with detection results and control instructions given the highest priority and raw signal data given the next highest priority, using traffic shaping rules defined by the IEEE 802.1Qbv standard.

[0035] Error recovery mechanism: When clock lock loss or data packet loss is detected, hardware-level redundant clock source switching and software-level state synchronization protocol are initiated.

[0036] Preferably, the system further includes an environmental noise modeling module, including:

[0037] The data preprocessing unit obtains the real-time spectrum data and baseband signal from the signal processing module, performs segmented windowing processing on the time domain signal, and generates an input matrix;

[0038] Independent component analysis unit, which performs blind source separation on the input signal matrix, extracts background noise components, and constructs a noise power spectrum density feature library;

[0039] The principal component analysis unit performs covariance matrix decomposition on the frequency domain spectrum data, extracts the principal components by eigenvalue sorting, and separates the interference signal components;

[0040] The target signal extraction unit determines the remaining unclassified components as target signals.

[0041] In a second aspect, the present invention provides the following technical solution: a method for detecting radio frequency signals at a cognitive radio front end, the method comprising:

[0042] S1, RF signal reception and preprocessing, directionally receiving RF signals, down-converting to intermediate frequency and filtering out interference;

[0043] S2, signal digitization, converting the intermediate frequency signal into a digital signal and synchronizing the timing;

[0044] S3, signal processing and detection, generating baseband signals and spectrograms, and identifying signal types through energy detection, cyclostationary analysis, and deep learning classification;

[0045] S4, noise modeling and calibration, separation of noise and interference signals, and dynamic adjustment of detection thresholds;

[0046] S5, synchronization control, global clock synchronization, switching to redundant clock source in case of error.

[0047] In a third aspect, the invention provides the following technical solution: a computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor; the processor implements the above-mentioned method for detecting RF signals of a cognitive radio front end when executing the computer program.

[0048] In a fourth aspect, the present invention provides the following technical solution: a readable storage medium having a computer program stored thereon, wherein the computer program implements the above-mentioned cognitive radio front-end RF signal detection method when executed by a processor.

[0049] The present invention has the following beneficial effects:

[0050] 1. In the present invention, by integrating energy detection, cyclostationary feature analysis and deep learning classification algorithms, a time-frequency domain joint analysis framework is constructed. By combining the time-domain transient feature extraction capability of 1D-CNN with the frequency-domain long-range dependency modeling advantage of Transformer, high-precision recognition of complex modulated signals is achieved, significantly improving the detection sensitivity and anti-interference capability in dynamic electromagnetic environments.

[0051] 2. In the present invention, the operation timing of each module is accurately scheduled through the time-sensitive network protocol. Based on the global clock synchronization and dynamic delay compensation mechanism, the timing drift and data conflict in multi-module collaboration are eliminated, ensuring strict time alignment of the entire link of the RF front-end, analog-to-digital conversion and signal processing, and ensuring the stability and real-time performance of the system under complex working conditions.

[0052] 3. In the present invention, by acquiring the real-time spectrum and baseband signal, the time domain signal is segmented and windowed, and divided into signal frames of fixed duration to generate an input matrix, thereby suppressing spectrum leakage and boundary effects, providing a high-fidelity data basis for subsequent noise separation, and ensuring the stability of noise feature extraction by eliminating non-stationary noise interference, thereby enhancing the reliability of modeling in dynamic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is an architecture diagram of a cognitive radio front-end RF signal detection system proposed by the present invention;

[0054] Figure 2 This is a flow chart of a method for detecting radio frequency signals at a cognitive radio front end proposed by the present invention. DETAILED DESCRIPTION

[0055] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0056] Example 1

[0057] Reference Figure 1 In a first embodiment of the present invention, the present invention provides a cognitive radio front-end radio frequency signal detection system, comprising:

[0058] Wideband RF front-end module, used to receive RF signals in the environment, supporting multi-band signal capture and anti-interference preprocessing;

[0059] The analog-to-digital conversion module uses an ADC with 14-bit resolution and 2GSPS sampling rate to convert the RF signal into a digital signal;

[0060] The signal processing module is connected to the output end of the analog-to-digital conversion module and is used to perform spectrum analysis, feature extraction and signal detection on the digital signal and output the detection results;

[0061] Synchronous control bus, based on the time-sensitive network protocol to coordinate the operation timing of each module, and synchronize the real-time RF front-end signal reception, analog-to-digital conversion and signal processing;

[0062] The wideband RF front-end module, analog-to-digital conversion module and signal processing module are connected in series in sequence and closed-loop controlled through a synchronous control bus.

[0063] Specifically, the wideband RF front-end module covers a wide spectrum range (e.g., 0.1-40 GHz) through multi-band signal capture capabilities, performs anti-interference pre-processing for complex electromagnetic environments, and outputs intermediate frequency signals with a high signal-to-noise ratio, providing a pure signal substrate for back-end analog-to-digital conversion and signal analysis, ensuring the sensitivity and reliability of RF signal detection in complex scenarios.

[0064] The analog-to-digital conversion module, serving as the digital hub of the signal chain, converts the analog intermediate frequency (IF) signal output by the RF front-end into a digital signal with high fidelity. Its 14-bit resolution ensures signal dynamic range (≥84dB) and weak signal quantization accuracy (LSB ≤61μV). Its 2GSPS sampling rate ensures alias-free capture of wideband signals (instantaneous bandwidth ≥1GHz). It also supports the JESD204B high-speed interface for low-latency data transmission, providing a high signal-to-noise ratio (SNR) and low-distortion digital signal substrate for subsequent signal processing modules. This effectively addresses the weak signal missed detection issues caused by spectral aliasing and lack of high resolution associated with traditional low-sampling-rate ADCs, enhancing the system's ability to analyze complex modulated signals and low SNR scenarios.

[0065] The signal processing module performs multi-dimensional information analysis on the digital signal after analog-to-digital conversion. Through spectrum analysis, feature extraction (extraction of cyclostationarity, modulation parameters, etc.) and detection, it can accurately identify and classify target signals in complex electromagnetic environments, thereby converting the original digital signal into actionable detection results (such as signal type, spectrum occupancy, and interference source location). This significantly improves the system's detection probability in low signal-to-noise ratio scenarios, the classification accuracy of complex modulated signals, and the anti-interference ability in dynamic environments, providing highly reliable and low-latency decision support for applications such as spectrum monitoring and communication perception.

[0066] The synchronous control bus uses the Time Sensitive Network (TSN) protocol to precisely schedule the operation timing of each module at the nanosecond level, ensuring strict time alignment of RF front-end signal reception, analog-to-digital conversion sampling, and signal processing and analysis. Based on global clock synchronization (such as GPS disciplined crystal oscillators) and a dynamic delay compensation mechanism, it eliminates timing drift and data conflicts caused by parallel operation of multiple modules, achieving a deterministic response with end-to-end processing latency ≤ 2ms. Priority scheduling and error recovery strategies ensure the transmission reliability of critical data, achieving high real-time performance, low jitter (clock deviation ≤ 100ps), and anti-interference capabilities for full-link signal processing even in complex electromagnetic environments.

[0067] The wideband RF front-end module, analog-to-digital conversion module and signal processing module are connected in series to form a unidirectional signal processing link. At the same time, through the closed-loop feedback control of the synchronous control bus, dynamic coordination of signal acquisition, conversion and analysis is achieved. The series connection ensures the efficient transmission of the signal flow and avoids data conflicts in the parallel architecture. The closed-loop control of the synchronous bus monitors the status of each module in real time, dynamically adjusts the anti-interference parameters, sampling rate and detection threshold, and solves the signal distortion or false detection problems caused by environmental disturbances in traditional open-loop systems.

[0068] The wideband RF front-end module includes:

[0069] The photon-assisted down-conversion unit down-converts the RF signal to an intermediate frequency through a lithium niobate optical modulator, which is used for mixing-free processing of high-frequency signals.

[0070] Metamaterial reconfigurable antenna, based on a graphene patch array, dynamically adjusts the radiation pattern with a gain of ≥12dBi for directional reception of target signals;

[0071] GaN low-noise amplifier, used to amplify received RF signals and reduce signal link noise;

[0072] The anti-aliasing filter group is used to filter out out-of-band interference signals and retain valid signals within the target frequency band.

[0073] Specifically, the photon-assisted down-conversion unit directly converts high-frequency RF signals (such as millimeter wave 28-40GHz) into intermediate frequency signals (such as 2-6GHz). It uses the interaction between the optical carrier and the RF signal to complete the frequency shift of the high-frequency signal in the optical domain through the electro-optical modulation of the Mach-Zehnder modulator, effectively avoiding the problems of harmonic interference and intermodulation distortion introduced by nonlinear devices (such as diodes) in traditional mixers, supporting a spurious-free dynamic range (SFDR) ≥ 75dB, and breaking through the frequency bottleneck of traditional circuits in the millimeter wave band, achieving ultra-wideband coverage, and providing a low-noise, high-linear intermediate frequency signal substrate for back-end analog-to-digital conversion and signal analysis;

[0074] The metamaterial reconfigurable antenna uses the graphene patch array's dynamic dielectric constant control capability to achieve real-time reconstruction of the radiation pattern, enhancing target signal reception and suppressing multipath interference in complex electromagnetic environments. By applying a bias voltage to change the graphene surface impedance, the antenna can dynamically adjust the resonant frequency and radiation characteristics within the 0.1-40GHz frequency band, achieving a gain of ≥12dBi and a sidelobe suppression ratio of ≤-15dB, significantly improving weak signal reception sensitivity.

[0075] The GaN low-noise amplifier performs low-noise amplification of weak RF signals received by the antenna over a wide bandwidth (0.1-40 GHz). It utilizes a distributed amplification architecture and adaptive bias technology to maintain high linearity while amplifying the signal, avoiding intermodulation distortion caused by strong interference signals. Combined with an anti-aliasing filter bank, it can meet the amplification requirements of signals with an instantaneous bandwidth ≥ 2 GHz, boosting the microvolt-level signal received by the antenna to the effective quantization range of the analog-to-digital conversion module, ensuring the dynamic range and accuracy of subsequent digital signal processing, and significantly enhancing the system's ability to handle weak signals and complex modulated signals.

[0076] The anti-aliasing filter group filters out out-of-band interference signals through LC or SAW filter cascades while retaining valid signals within the target frequency band, providing high-purity, low-distortion target frequency band signals for back-end analog-to-digital conversion and signal processing, significantly improving the system's dynamic range and weak signal detection capabilities in dense interference environments.

[0077] The analog-to-digital conversion module includes:

[0078] The clock source unit generates a synchronous clock signal based on a phase-locked loop to control the ADC sampling timing;

[0079] The quantizer unit converts the analog signal into a digital signal through the Delta-Sigma architecture;

[0080] Digital interface unit, supporting JESD204B protocol, used to transmit digital signals to the signal processing module;

[0081] The oversampling mode unit achieves 8 times oversampling through interpolation filters to improve signal quantization accuracy.

[0082] Specifically, the clock source unit synchronizes an external reference clock (such as the global clock provided by the time-sensitive network bus) through a phase-locked loop (PLL) to generate a low-jitter, high-stability sampling clock signal (frequency error ≤ 1ppm), ensuring precise timing control of the ADC at a 2GSPS sampling rate.

[0083] The quantizer unit converts broadband analog signals into digital signals and suppresses the effects of clock jitter and power supply interference on quantization accuracy through noise shaping, providing a high-fidelity, low-noise digital signal substrate for subsequent signal processing.

[0084] The digital interface unit uses the JESD204B protocol to achieve high-speed, low-latency, multi-channel synchronous digital signal transmission. It replaces the traditional parallel interface with a serial link to achieve strict timing matching with the signal processing module, avoiding data loss or misalignment. Combined with adaptive equalization and clock correction functions, it maintains signal integrity in complex electromagnetic environments.

[0085] The oversampling mode unit interpolates the digital signal after analog-to-digital conversion, extending the original sampling rate (such as 2GSPS) to 16GSPS. Digital filtering eliminates the image frequency components introduced by interpolation, and combines noise shaping technology to push the quantization noise energy to the high-frequency band. Digital extraction filtering is then used to retain the target frequency band signal, ultimately increasing the effective resolution from 14 bits to 16 bits, extending the dynamic range to ≥96dB, and reducing the in-band noise floor.

[0086] The signal processing module includes:

[0087] The digital down-conversion unit converts the intermediate frequency signal into a baseband I / Q signal through the orthogonal local oscillator signal generated by the orthogonal mixer and the digitally controlled oscillator. The signal is then decimated and down-sampled through a multi-stage cascaded CIC filter and FIR compensation filter to output a low-rate baseband signal.

[0088] The fast Fourier transform unit performs a 4096-point FFT operation on the baseband signal, suppresses spectrum leakage through the Blackman-Harris window function, and generates a spectrum density map. The FFT calculation formula is:

[0089] Where N = 4096 is the number of FFT points, x(n) is the input baseband signal time domain sequence, and w(n) is the Blackman-Harris window function;

[0090] The detection algorithm unit performs energy detection, cyclostationary feature analysis, and deep learning-based signal classification on the spectrum density map. It also performs the following algorithms on the spectrum density map:

[0091] Energy detection, the detection threshold is set by the dynamic threshold formula: Among them, T is the detection threshold, α is the false alarm rate factor, is the estimated value of the sliding window noise power;

[0092] Cyclostationary characteristic analysis: Where α is the cycle frequency, τ is the time delay, and K is the number of signal samples;

[0093] Deep learning classification: implemented through a hybrid network, including:

[0094] 1D-CNN convolution formula: Extract time domain transient features, where w k is the convolution kernel weight, b is the bias term, x t+k is the t+kth sampling value of the time domain baseband signal.

[0095] Specifically, the digital down-conversion unit converts the intermediate frequency signal into a baseband I / Q signal through an orthogonal local oscillator signal generated by an orthogonal mixer and a digitally controlled oscillator. The signal is then decimated and downsampled by a multi-stage cascaded CIC filter and FIR compensation filter. The CIC filter achieves efficient integer multiple downsampling through a cascaded structure of integration and comb filtering, while the FIR compensation filter compensates for the passband attenuation defect of the CIC filter by optimizing the passband ripple and stopband attenuation.

[0096] The Fast Fourier Transform (FFT) unit performs a 4096-point FFT operation on the baseband signal, suppressing spectrum leakage through a Blackman-Harris window function. The mainlobe width and sidelobe attenuation characteristics of the window function are utilized to significantly reduce spectrum leakage, generating a high-resolution spectrum density map. This allows precise positioning of signal frequency domain characteristics in complex electromagnetic environments, supporting dynamic signal detection and spectrum occupancy analysis.

[0097] The detection algorithm unit performs energy detection, cyclostationary feature analysis, and deep learning classification on the spectrum density map to achieve multi-dimensional signal analysis. The energy detection dynamically sets the detection threshold using a dynamic threshold formula (a sliding window noise power estimate), and extracts the signal modulation periodicity characteristics by combining the cyclostationary feature analysis formula. A hybrid network is constructed using the 1D-CNN convolution formula and the Transformer self-attention mechanism to achieve time-frequency domain joint feature modeling. The dynamic threshold calibration mechanism is combined with multi-algorithm collaborative decision-making to effectively balance detection sensitivity and false alarm suppression. At the same time, the classification accuracy and real-time performance are optimized based on the hybrid network architecture.

[0098] The hybrid network also includes the Transformer self-attention mechanism, which extracts long-range dependency features in the frequency domain, concatenates the time domain feature vector and the frequency domain feature vector in the channel dimension, and outputs the probability distribution of the signal type through the softmax function.

[0099] Specifically, the hybrid network extracts long-range dependency features in the frequency domain through the Transformer self-attention mechanism, uses the self-attention weight to dynamically focus on key frequency bands in the spectral density map (such as the main frequency of the signal and harmonic components), combines position encoding to retain the temporal correlation of the frequency domain sequence, and concatenates the time domain transient features and frequency domain features extracted by 1D-CNN into a joint feature vector in the channel dimension. After inputting into the fully connected layer, the probability distribution of the signal type is generated through the softmax function, realizing the deep fusion and complementarity of time-frequency domain features, and significantly improving the classification accuracy of complex modulated signals under low signal-to-noise ratio.

[0100] The synchronous control bus includes:

[0101] The clock source unit generates a global clock reference signal based on the GPS disciplined crystal oscillator and synchronizes the local clocks of each module through a phase-locked loop;

[0102] A delay compensation mechanism inserts timestamps between the RF front-end, analog-to-digital conversion module, and signal processing module, dynamically adjusting transmission delays in conjunction with data buffer queues to compensate for link jitter and temperature drift.

[0103] Priority scheduling strategy assigns transmission priority based on data type, with detection results and control instructions given the highest priority and raw signal data given the next highest priority, using traffic shaping rules defined by the IEEE 802.1Qbv standard.

[0104] Error recovery mechanism: When clock lock loss or data packet loss is detected, hardware-level redundant clock source switching and software-level state synchronization protocol are initiated.

[0105] Specifically, the clock source unit of the synchronous control bus generates a global clock reference signal based on the GPS disciplined crystal oscillator, and dynamically adjusts the local clock phase error of each module through a phase-locked loop to ensure the timing consistency of the full-link signal processing;

[0106] The delay compensation mechanism embeds precise timestamps in the data streams of the RF front-end, analog-to-digital conversion module, and signal processing module. This mechanism, combined with an adjustable-depth data buffer, dynamically compensates for random delays in the transmission link. Linear interpolation is used to predict delay trends and adjust the buffer depth in real time, ensuring accurate cross-module data alignment and avoiding signal aliasing or analysis errors caused by link asynchrony.

[0107] The priority scheduling strategy dynamically allocates transmission bandwidth and processing resources based on data type. It uses the IEEE802.1Qbv standard's Time-Aware Shaper (TAS) to set exclusive transmission time windows for high-priority data such as test results and control instructions, forcibly isolating low-priority traffic such as raw signal data, and achieving deterministic, low-latency transmission of critical instructions.

[0108] The error recovery mechanism works in conjunction with a hardware-level redundant clock source (such as an oven-controlled crystal oscillator + atomic clock backup) and a software-level state synchronization protocol. When clock lock loss or data packet loss is detected, the seamless switching logic is triggered and the context state of the signal processing module is quickly restored based on an incremental synchronization algorithm. This ensures continuous operation of the system under extreme interference or hardware failure, minimizing the risk of service interruption.

[0109] The system also includes an environmental noise modeling module, including:

[0110] The data preprocessing unit obtains the real-time spectrum data and baseband signal from the signal processing module, performs segmented windowing processing on the time domain signal, and generates an input matrix;

[0111] Independent component analysis unit, which performs blind source separation on the input signal matrix, extracts background noise components, and constructs a noise power spectrum density feature library;

[0112] The principal component analysis unit performs covariance matrix decomposition on the frequency domain spectrum data, extracts the principal components by eigenvalue sorting, and separates the interference signal components;

[0113] The target signal extraction unit determines the remaining unclassified components as target signals.

[0114] Specifically, the data preprocessing unit obtains real-time spectrum data and baseband signals from the signal processing module, performs segmented windowing processing (such as Hanning window) on the time domain signal, divides it into signal frames of fixed length, and generates an input matrix. By suppressing spectrum leakage and boundary effects, it provides a high-fidelity, low-distortion data foundation for subsequent noise separation, while reducing the interference of non-stationary noise on feature extraction, ensuring the stability and reliability of noise modeling.

[0115] The Independent Component Analysis (ICA) unit performs blind source separation on the input signal matrix, utilizing the FastICA algorithm to maximize the signal's non-Gaussianity, isolate statistically independent background noise components from the mixed signal, and construct a dynamically updated noise power spectral density feature library. This real-time noise feature matching improves the system's adaptability to complex electromagnetic environments and significantly reduces the negative impact of noise floor fluctuations on signal detection sensitivity.

[0116] The principal component analysis unit performs covariance matrix decomposition on the frequency domain spectrum data, extracts the first K principal components by eigenvalue sorting, separates high-power narrowband interference signals (such as radar pulses and adjacent frequency communication leakage), suppresses the contamination of the target frequency band by the interference signal through the frequency domain energy focusing feature, retains the low-power broadband target signal component, and provides high signal-to-noise ratio input data for subsequent signal detection and classification;

[0117] The target signal extraction unit determines the remaining components of the mixed signal that are not classified as noise and interference as the target signal, and conducts joint verification based on the time domain transient characteristics (such as modulation transients) and the frequency domain cyclostationary characteristics (such as symbol rate periodicity) to eliminate false alarms caused by residual interference and improve the probability of target signal detection. At the same time, it outputs the time and frequency domain parameters of the target signal (center frequency, bandwidth, modulation type), providing a decision-making basis for dynamic spectrum access and interference avoidance.

[0118] Example 2:

[0119] Reference Figure 2 In a second embodiment of the present invention, the present invention provides a method for detecting radio frequency signals at a cognitive radio front end, the method comprising:

[0120] S1, RF signal reception and preprocessing, directionally receiving RF signals, down-converting to intermediate frequency and filtering out interference;

[0121] S2, signal digitization, converting the intermediate frequency signal into a digital signal and synchronizing the timing;

[0122] S3, signal processing and detection, generating baseband signals and spectrograms, and identifying signal types through energy detection, cyclostationary analysis, and deep learning classification;

[0123] S4, noise modeling and calibration, separation of noise and interference signals, and dynamic adjustment of detection thresholds;

[0124] S5, synchronization control, global clock synchronization, switching to redundant clock source in case of error.

[0125] Specifically, S1 uses a metamaterial reconfigurable antenna to directionally receive RF signals in the target frequency band. It then uses photon-assisted down-conversion technology to convert high-frequency signals into intermediate frequencies without mixing. It then uses an anti-aliasing filter bank to filter out out-of-band interference and harmonic components, ensuring the purity and sensitivity of signal reception and providing high-integrity intermediate frequency signal input for subsequent processing.

[0126] S2 uses a high-resolution analog-to-digital converter to quantize the intermediate frequency signal, triggers multi-channel synchronous sampling through a synchronous control bus, and transmits digital signals in combination with a high-speed interface protocol to ensure sampling timing consistency and avoid spectrum aliasing and phase distortion problems.

[0127] The S3 generates baseband I / Q signals through digital down-conversion, performs high-resolution spectrum analysis to generate spectral density maps, and integrates energy detection, cyclostationary analysis, and deep learning classification algorithms to achieve accurate recognition of complex modulated signals, improving detection probability and false alarm mitigation capabilities in dynamic environments.

[0128] S4 uses blind source separation and frequency domain decomposition technology to separate noise and interference from mixed signals, build a dynamic noise feature library, and adjust the detection threshold in real time through event triggering and periodic calibration mechanisms to suppress the impact of environmental noise fluctuations on system performance;

[0129] S5 tames the local clock phase based on the global clock reference signal, ensures the continuous operation of the system in the event of clock anomalies through redundant clock source switching and state synchronization protocol, and optimizes end-to-end processing latency and reliability.

[0130] Example 3

[0131] The third embodiment of the present invention is based on the same inventive concept. The present invention proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of a cognitive radio front-end RF signal detection method of the above embodiment.

[0132] Example 4

[0133] The fourth embodiment of the present invention is based on the same inventive concept. The present invention proposes a computer terminal including: a processor and a memory; the processor and the memory communicate with each other; the memory is used to store instructions; the processor is used to execute the instructions in the memory to perform a cognitive radio front-end RF signal detection method of the above embodiment.

[0134] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0135] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A cognitive radio front-end radio frequency signal detection system, characterized in that: include: Wideband RF front-end module, used to receive RF signals in the environment, supporting multi-band signal capture and anti-interference preprocessing; The analog-to-digital conversion module uses an ADC with 14-bit resolution and 2GSPS sampling rate to convert the RF signal into a digital signal; a signal processing module, connected to the output end of the analog-to-digital conversion module, for performing spectrum analysis, feature extraction and signal detection on the digital signal and outputting the detection result; Synchronous control bus, based on the time-sensitive network protocol to coordinate the operation timing of each module, and synchronize the real-time RF front-end signal reception, analog-to-digital conversion and signal processing; The wideband RF front-end module, analog-to-digital conversion module and signal processing module are connected in series in sequence and closed-loop controlled via a synchronous control bus.

2. The cognitive radio front-end radio frequency signal detection system according to claim 1, characterized in that: The broadband radio frequency front-end module includes: The photon-assisted down-conversion unit down-converts the RF signal to an intermediate frequency through a lithium niobate optical modulator, which is used for mixing-free processing of high-frequency signals. Metamaterial reconfigurable antenna, based on a graphene patch array, dynamically adjusts the radiation pattern with a gain of ≥12dBi for directional reception of target signals; GaN low-noise amplifier, used to amplify received RF signals and reduce signal link noise; The anti-aliasing filter group is used to filter out out-of-band interference signals and retain valid signals within the target frequency band.

3. The cognitive radio front-end radio frequency signal detection system according to claim 1, characterized in that: The analog-to-digital conversion module includes: The clock source unit generates a synchronous clock signal based on a phase-locked loop to control the ADC sampling timing; The quantizer unit converts the analog signal into a digital signal through the Delta-Sigma architecture; Digital interface unit, supporting JESD204B protocol, used to transmit digital signals to the signal processing module; The oversampling mode unit achieves 8 times oversampling through interpolation filters to improve signal quantization accuracy.

4. The cognitive radio front-end radio frequency signal detection system according to claim 1, characterized in that: The signal processing module includes: The digital down-conversion unit converts the intermediate frequency signal into a baseband I / Q signal through the orthogonal local oscillator signal generated by the orthogonal mixer and the digitally controlled oscillator. The signal is then decimated and down-sampled through a multi-stage cascaded CIC filter and FIR compensation filter to output a low-rate baseband signal. The fast Fourier transform unit performs a 4096-point FFT operation on the baseband signal, suppresses spectrum leakage through the Blackman-Harris window function, and generates a spectrum density map. The FFT calculation formula is: Where N = 4096 is the number of FFT points, x(n) is the input baseband signal time domain sequence, and w(n) is the Blackman-Harris window function; A detection algorithm unit performs energy detection, cyclostationary feature analysis, and deep learning-based signal classification on the spectrum density map. The following algorithms are performed on the spectrum density map: Energy detection, the detection threshold is set by the dynamic threshold formula: Among them, T is the detection threshold, α is the false alarm rate factor, is the estimated value of the sliding window noise power; Cyclostationary characteristic analysis: Where α is the cycle frequency, τ is the time delay, and K is the number of signal samples; Deep learning classification: implemented through a hybrid network, including: 1D-CNN convolution formula: Extract time domain transient features, where w k is the convolution kernel weight, b is the bias term, x t+k is the t+kth sampling value of the time domain baseband signal.

5. The cognitive radio front-end radio frequency signal detection system according to claim 4, characterized in that: The hybrid network also includes a Transformer self-attention mechanism, which extracts long-range dependency features in the frequency domain, concatenates the time domain feature vector and the frequency domain feature vector in the channel dimension, and outputs the probability distribution of the signal type through the softmax function.

6. The cognitive radio front-end radio frequency signal detection system according to claim 1, characterized in that: The synchronous control bus comprises: The clock source unit generates a global clock reference signal based on the GPS disciplined crystal oscillator and synchronizes the local clocks of each module through a phase-locked loop; A delay compensation mechanism inserts timestamps between the RF front-end, analog-to-digital conversion module, and signal processing module, dynamically adjusting transmission delays in conjunction with data buffer queues to compensate for link jitter and temperature drift. Priority scheduling strategy assigns transmission priority based on data type, with detection results and control instructions given the highest priority and raw signal data given the next highest priority, using traffic shaping rules defined by the IEEE 802.1Qbv standard. Error recovery mechanism: When clock lock loss or data packet loss is detected, hardware-level redundant clock source switching and software-level state synchronization protocol are initiated.

7. The cognitive radio front-end radio frequency signal detection system according to claim 1, characterized in that: The system also includes an environmental noise modeling module, including: The data preprocessing unit obtains the real-time spectrum data and baseband signal from the signal processing module, performs segmented windowing processing on the time domain signal, and generates an input matrix; Independent component analysis unit, which performs blind source separation on the input signal matrix, extracts background noise components, and constructs a noise power spectrum density feature library; The principal component analysis unit performs covariance matrix decomposition on the frequency domain spectrum data, extracts the principal components by eigenvalue sorting, and separates the interference signal components; The target signal extraction unit determines the remaining unclassified components as target signals.

8. A method for detecting radio frequency signals in a cognitive radio front end, characterized in that: A cognitive radio front-end radio frequency signal detection system according to any one of claims 1 to 7, the method comprising: S1, RF signal reception and preprocessing, directionally receiving RF signals, down-converting to intermediate frequency and filtering out interference; S2, signal digitization, converting the intermediate frequency signal into a digital signal and synchronizing the timing; S3, signal processing and detection, generating baseband signals and spectrograms, and identifying signal types through energy detection, cyclostationary analysis, and deep learning classification; S4, noise modeling and calibration, separation of noise and interference signals, and dynamic adjustment of detection thresholds; S5, synchronization control, global clock synchronization, switching to redundant clock source in case of error.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for detecting radio frequency signals of a cognitive radio front-end is implemented as claimed in claim 8.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by the processor, the method for detecting radio frequency signals of a cognitive radio front end according to claim 8 is implemented.

Images