A double sliding window energy detection method based on intelligent adaptive cepstrum filtering

Abstract

This invention belongs to the field of wireless spectrum signal detection technology, specifically relating to a dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering, including: S1, signal time-frequency transformation and cepstral decomposition: the received RF baseband signal is framed and preprocessed by windowing, and the noisy power spectrum is obtained by fast Fourier transform. The power spectrum is then subjected to logarithmic compression transform and inverse fast Fourier transform in sequence. This invention abandons the fixed filter parameter mode and adaptively solves the optimal cutoff cepstral frequency in real time based on the global energy distribution and attenuation gradient characteristics of each frame's cepstral spectrum. Combined with a cosine-gradient soft threshold filtering method, it can accurately remove high cepstral random noise and eliminate spectral glitches, while preserving the detailed features of the low cepstral spectrum envelope to the greatest extent. It effectively adapts to complex time-varying noise environments, takes into account the spectral integrity of broadband continuous signals and narrowband weak signals, and greatly improves the adaptability and stability of signal detection in complex electromagnetic environments.

Description

Technical Field

[0001] This invention belongs to the field of wireless spectrum signal detection technology, specifically relating to a dual sliding window energy detection method based on intelligent adaptive cepstral filtering. Background Technology

[0002] In wireless communication and electromagnetic sensing scenarios such as cognitive radio, spectrum monitoring, weak signal reconnaissance, and electromagnetic environment detection, energy detection has become the most widely used spectrum carrier detection technology due to its advantages of not requiring prior signal information, low computational complexity, and ease of hardware implementation.

[0003] Traditional cepstral filtering techniques all employ manually preset fixed cutoff frequencies for noise reduction, and these parameters cannot be adaptively adjusted according to real-time noise conditions. In real-world complex time-varying noise environments, noise intensity and distribution fluctuate dynamically with the electromagnetic scene. Fixed filtering parameters have inherent drawbacks: when the ambient noise intensity is high, a fixed low cutoff frequency cannot completely filter out high-frequency noise spikes, leaving numerous random interference peaks in the spectrum, which can easily trigger false alarms; when the ambient noise intensity is low and there are narrowband weak signals, a fixed high cutoff frequency will excessively smooth spectral details, obliterating the spectral envelope characteristics of narrowband signals and transient weak signals, resulting in missed signal detection. This type of fixed-parameter filtering mode cannot adapt to dynamic noise scenarios and cannot simultaneously achieve noise suppression effectiveness and signal spectral integrity, leading to a significant decrease in the stability and adaptability of signal detection in complex electromagnetic environments. Summary of the Invention

[0004] The purpose of this invention is to provide a dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering, which aims to solve the problem that traditional cepstral filtering techniques all use manually preset fixed cutoff frequencies to complete noise reduction, and the parameters cannot be adaptively adjusted according to real-time noise conditions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering, comprising the following steps:

[0006] S1. Signal Time-Frequency Transformation and Cepstral Decomposition: The received RF baseband signal is preprocessed by framing and windowing. The noisy power spectrum is obtained by fast Fourier transform. The power spectrum is then subjected to logarithmic compression transform and inverse fast Fourier transform in sequence to obtain the original cepstral sequence. The characteristics of the cepstral domain are used to separate the low cepstral envelope component carrying the real signal and the high cepstral noise spike component carrying random interference.

[0007] S2. Intelligent adaptive cepstral low-pass filtering for noise reduction: Statistically analyze the global energy distribution and energy attenuation gradient characteristics of the original cepstral sequence frame by frame, and adaptively solve the optimal cutoff frequency corresponding to each frame signal based on the cumulative energy ratio of the cepstral dimension, thus abandoning the fixed cutoff frequency filtering mode.

[0008] A dynamic soft-threshold low-pass filter is constructed based on the optimal cutoff frequency. The low-frequency cepstral envelope component is fully preserved, while the high-frequency noise spike component is gradually attenuated and suppressed. The filtered clean cepstral sequence is restored by fast Fourier transform to obtain a smooth spectral envelope curve without noise interference.

[0009] S3, Dual sliding window hierarchical collaborative energy detection: Configure large and small sliding windows with fixed sizes and differentiating sizes to traverse the smooth spectral envelope curve synchronously with the same length;

[0010] By integrating the overall energy of the frequency band through a large sliding window, a coarse screening of broadband continuous carriers is completed. By using a small sliding window, the inflection point of spectral energy change is precisely captured, and the local location of weak narrowband signals is locked.

[0011] By integrating the average energy threshold, local energy gradient, and window energy difference multidimensional features of the dual-window system, a set of candidate frequency bands suspected of having effective carriers is obtained.

[0012] S4. Accurate removal of false alarms during post-demodulation: For all candidate frequency bands, the multi-standard general demodulation algorithm is called to perform carrier synchronization, symbol synchronization and baseband decision processing in sequence, and output the corresponding demodulation confidence index.

[0013] Candidate frequency bands with demodulation confidence levels below a preset effective threshold are identified as false carriers caused by noise interference and are removed. Frequency bands with valid demodulation features and normal symbol output are retained as the final effective carrier detection results, thus completing the signal energy detection.

[0014] As a preferred method for energy detection based on intelligent adaptive cepstral filtering in this invention, the specific process of adaptively solving for the optimal cutoff frequency in S2 is as follows:

[0015] The energy values ​​of individual frequency points in the original cepstrum sequence are extracted one by one, and the total energy of the original cepstrum sequence is calculated. The energy of individual frequency points is accumulated segment by segment along the ascending direction of the cepstrum, and the proportion of accumulated energy to total energy and the energy attenuation slope of adjacent cepstrum intervals are calculated in real time. A fixed energy proportion threshold is preset, and the cepstrum point that first meets the threshold of accumulated energy proportion is determined as the optimal cutoff frequency of the current frame signal. Each frame signal is independently iterated and updated to achieve adaptive filtering adaptation in scenarios where the noise intensity varies over time.

[0016] As a dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering according to the present invention, preferably, the filtering rule of the dynamic soft-threshold low-pass filter in S2 is as follows:

[0017] Using the optimal cutoff frequency obtained in real time as the boundary, the low frequency components within the boundary are preserved with an all-pass strategy and no amplitude attenuation; the high frequency components outside the boundary are suppressed with a cosine gradient attenuation function to avoid spectral edge distortion caused by hard threshold filtering, and to preserve the subtle spectral characteristics of the narrowband signal to the greatest extent while completely filtering out random noise spikes.

[0018] As a dual sliding window energy detection method based on intelligent adaptive cepstral filtering according to the present invention, preferably, in S3 the size ratio of the large sliding window to the small sliding window is a fixed adaptation ratio, and the size of the large window is 8 to 16 times the size of the small window.

[0019] The large sliding window is used to smooth random noise in the frequency band and statistically analyze the base energy of the broadband spectrum, adapting to broadband continuous carrier detection; the small sliding window is used to amplify the abrupt change characteristics of local spectrum energy, adapting to the inflection point identification of weak narrowband signals and transient signals. The dual windows slide and sample synchronously to ensure the timing alignment of the spectrum position.

[0020] As a dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering according to the present invention, preferably, the multi-dimensional judgment rule for screening candidate frequency bands in S3 is as follows:

[0021] A broadband noise reference threshold and a narrowband energy mutation gradient threshold are pre-configured. When the average energy of the large sliding window is greater than the broadband noise reference threshold at the same frequency point, and the local energy gradient corresponding to the small sliding window is greater than the narrowband energy mutation gradient threshold, and the energy difference between the two windows is within the preset effective range, the frequency band where the frequency point is located is determined to be a candidate carrier frequency band; otherwise, it is determined to be a pure noise frequency band and is directly discarded.

[0022] As a dual sliding window energy detection method based on intelligent adaptive cepstral filtering according to the present invention, preferably, the multi-standard universal demodulation algorithm in S4 is compatible with mainstream digital modulation standards such as ASK, FSK, PSK, and QAM.

[0023] By matching the time-domain and frequency-domain characteristics of different modulation schemes, blind synchronization and blind demodulation are completed, and the symbol accuracy, constellation diagram convergence, phase deviation and multi-dimensional demodulation confidence index are output to comprehensively determine whether there is a valid carrier signal in the candidate frequency band.

[0024] As a dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering according to the present invention, preferably, when performing cepstral decomposition in S1, a complex cepstral transformation method is adopted, retaining only the amplitude dimension cepstral component for subsequent filtering and detection processing, and discarding the phase dimension cepstral component to avoid detection interference caused by phase random perturbation.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] 1. This invention abandons the fixed filter parameter mode and adaptively solves the optimal cutoff frequency in real time based on the energy distribution and attenuation gradient characteristics of the global cepstrum of each frame. Combined with the cosine gradient soft threshold filtering method, it can accurately remove high cepstrum random noise and eliminate spectral glitches, while preserving the detailed features of the low cepstrum spectral envelope to the greatest extent. It can effectively adapt to complex time-varying noise environments, take into account the spectral integrity of broadband continuous signals and narrowband weak signals, and greatly improve the adaptability and stability of signal detection in complex electromagnetic environments.

[0027] 2. This invention employs a synchronous collaborative traversal mechanism with dual sliding windows of varying sizes. The large window is responsible for integrating and statistically analyzing the global energy of the frequency band and smoothing random noise, efficiently completing the coarse screening of broadband carriers. The small window accurately captures the inflection point of spectral energy mutations and amplifies local weak features, precisely locating the position of narrowband weak signals. Through joint judgment of multi-dimensional energy features of the dual windows, it achieves full coverage detection of broadband, narrowband, and transient weak signals, effectively compensating for the shortcomings of single-window detection technology in scene adaptation and significantly improving the detection accuracy and scene versatility of multi-type coexisting signals.

[0028] 3. This invention constructs a two-layer false alarm suppression architecture: "front-end cepstral filtering for coarse screening + back-end multi-standard trial demodulation for fine verification." The front end uses precise cepstral domain denoising to reduce the generation of false noise peaks. The back end relies on universal blind demodulation algorithms for multiple standards such as ASK, FSK, PSK, and QAM. Through multi-dimensional indicators such as demodulation confidence, symbol accuracy, and constellation diagram convergence, it distinguishes between real effective carriers and random noise interference from the fundamental level of signal modulation, completely eliminating false carriers that meet energy standards but lack effective modulation characteristics. Compared to traditional single-energy determination methods, this invention completes detection and verification from both physical characteristics and signal system dimensions, significantly reducing the false alarm rate in complex interference scenarios and significantly improving the accuracy and reliability of spectrum carrier detection in low signal-to-noise ratio environments. Attached Figure Description

[0029] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0030] Figure 1 This is a flowchart illustrating an embodiment of the present application. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] Please see Figure 1 The present invention provides the following technical solution: a dual sliding window energy detection method based on intelligent adaptive cepstral filtering, comprising the following steps:

[0033] S1. Signal Time-Frequency Transformation and Cepstral Decomposition: The received RF baseband signal is preprocessed by framing and windowing. The noisy power spectrum is obtained by fast Fourier transform. The power spectrum is then subjected to logarithmic compression transform and inverse fast Fourier transform in sequence to obtain the original cepstral sequence. The characteristics of the cepstral domain are used to separate the low cepstral envelope component carrying the real signal and the high cepstral noise spike component carrying random interference.

[0034] Let the received discrete baseband signal be... After frame segmentation and windowing, a single frame signal is obtained. Performing an FFT on it yields the complex spectrum:

[0035] ;

[0036] The power spectrum is taken and logarithmically compressed. The compressed spectrum is:

[0037] ;

[0038] in The constant is a very small constant to avoid the logarithm becoming meaningless. Performing an IFFT on the logarithmic spectrum yields the original complex cepstrum:

[0039] ;

[0040] In the formula The frequency inversion point number corresponds to the cepstral dimension; only the amplitude cepstral component is retained. By discarding phase components, random phase jitter interference is avoided. The low frequency reciprocal range corresponds to the overall spectrum envelope, while the high frequency reciprocal range corresponds to high-frequency glitches and noise disturbances in the spectrum, thus achieving initial decoupling between the signal and noise dimensions.

[0041] S2. Intelligent adaptive cepstral low-pass filtering for noise reduction: Statistically analyze the global energy distribution and energy attenuation gradient characteristics of the original cepstral sequence frame by frame, and adaptively solve the optimal cutoff frequency corresponding to each frame signal based on the cumulative energy ratio of the cepstral dimension, thus abandoning the fixed cutoff frequency filtering mode.

[0042] A dynamic soft-threshold low-pass filter is constructed based on the optimal cutoff frequency. The low-frequency cepstral envelope component is fully preserved, while the high-frequency noise spike component is gradually attenuated and suppressed. The filtered clean cepstral sequence is restored by fast Fourier transform to obtain a smooth spectral envelope curve without noise interference.

[0043] S3, Dual sliding window hierarchical collaborative energy detection: Configure large and small sliding windows with fixed sizes and differentiating sizes to traverse the smooth spectral envelope curve synchronously with the same length;

[0044] By integrating the overall energy of the frequency band through a large sliding window, a coarse screening of broadband continuous carriers is completed. By using a small sliding window, the inflection point of spectral energy change is precisely captured, and the local location of weak narrowband signals is locked.

[0045] By integrating the average energy threshold, local energy gradient, and window energy difference multidimensional features of the dual-window system, a set of candidate frequency bands suspected of having effective carriers is obtained.

[0046] S4. Accurate removal of false alarms during post-demodulation: For all candidate frequency bands, the multi-standard general demodulation algorithm is called to perform carrier synchronization, symbol synchronization and baseband decision processing in sequence, and output the corresponding demodulation confidence index.

[0047] Candidate frequency bands with demodulation confidence levels below a preset effective threshold are identified as false carriers caused by noise interference and are removed. Frequency bands with valid demodulation features and normal symbol output are retained as the final effective carrier detection results, thus completing the signal energy detection.

[0048] The preferred method is as follows: The specific process of adaptively solving for the optimal cutoff reciprocal frequency in S2 is as follows:

[0049] The energy values ​​of individual frequency points in the original cepstrum sequence are extracted one by one, and the total energy of the original cepstrum sequence is calculated. The energy of individual frequency points is accumulated segment by segment along the ascending direction of the cepstrum, and the proportion of accumulated energy to total energy and the energy attenuation slope of adjacent cepstrum intervals are calculated in real time. A fixed energy proportion threshold is preset, and the cepstrum point that first meets the threshold of accumulated energy proportion is determined as the optimal cutoff frequency of the current frame signal. Each frame signal is independently iterated and updated to achieve adaptive filtering adaptation in scenarios where the noise intensity varies over time.

[0050] In practical applications, the single-frequency reversal point energy is defined as follows: Total energy of a single frame cepstral spectrum:

[0051] ;

[0052] Inverted ascending sequence accumulated energy: Cumulative energy percentage:

[0053] ;

[0054] Preset energy percentage threshold , representing the dominant energy proportion of the effective signal envelope, is solved to satisfy The smallest reciprocal number This is the optimal adaptive cutoff inverse frequency. Simultaneously, the energy decay gradient is calculated:

[0055] ;

[0056] By constraining the noise mutation range through energy gradient, we can ensure... It always falls at the natural boundary between the signal envelope and noise glitches. This mechanism completely abandons the traditional fixed cutoff frequency parameter and updates it independently for each frame. When the noise is strong, the cutoff frequency is automatically reduced to completely filter out glitches; when the noise is weak, the cutoff frequency is automatically increased to retain the details of weak signals, thus solving the over-smoothing and under-smoothing defects caused by fixed parameters.

[0057] Preferably, the filtering rule of the dynamic soft-threshold low-pass filter in S2 is:

[0058] Using the optimal cutoff frequency obtained in real time as the boundary, the low frequency components within the boundary are preserved with an all-pass strategy and no amplitude attenuation; the high frequency components outside the boundary are suppressed with a cosine gradient attenuation function to avoid spectral edge distortion caused by hard threshold filtering, and to preserve the subtle spectral characteristics of the narrowband signal to the greatest extent while completely filtering out random noise spikes.

[0059] In practical applications, a cosine-gradient soft threshold filter function is constructed. :

[0060] ;

[0061] The pure cepstrum after filtering is The smoothed and denoised spectral envelope is then obtained by FFT reconstruction:

[0062] ;

[0063] This soft threshold gradual attenuation method has no spectral truncation abrupt changes and no edge distortion. While completely removing high-frequency inverted random noise and eliminating spectral glitches, it retains the details of the low-frequency inverted spectral envelope to the greatest extent. It is suitable for wide and narrow band mixed signal scenarios and greatly improves the detection stability in complex time-varying noise environments.

[0064] Preferably, in S3, the size ratio between the large sliding window and the small sliding window is a fixed adaptation ratio, with the size of the large window being 8 to 16 times the size of the small window;

[0065] The large sliding window is used to smooth random noise in the frequency band and statistically analyze the base energy of the broadband spectrum, adapting to broadband continuous carrier detection; the small sliding window is used to amplify the abrupt change characteristics of local spectrum energy, adapting to the inflection point identification of weak narrowband signals and transient signals. The dual windows slide and sample synchronously to ensure the timing alignment of the spectrum position.

[0066] In practical use, let the length of the small window be... Large window length The scaling factor Dual windows with fixed step size Synchronous sliding for smoothing the spectrum Calculate the window integral energy separately:

[0067] Local energy in a small window: ;

[0068] Global energy of large window: ;

[0069] The large window suppresses random noise fluctuations through long integral averaging, accurately characterizes the broadband spectrum base energy, and achieves broadband carrier coarse screening; the small window focuses on local frequency bands, amplifies the energy change inflection point of weak signals, accurately captures narrowband and transient signals, and solves the technical shortcomings of traditional single windows that cannot simultaneously detect both wideband and narrowband signals.

[0070] Preferred: The multi-dimensional judgment rule for selecting candidate frequency bands in S3 is as follows:

[0071] A broadband noise reference threshold and a narrowband energy mutation gradient threshold are pre-configured. When the average energy of the large sliding window is greater than the broadband noise reference threshold at the same frequency point, and the local energy gradient corresponding to the small sliding window is greater than the narrowband energy mutation gradient threshold, and the energy difference between the two windows is within the preset effective range, the frequency band where the frequency point is located is determined to be a candidate carrier frequency band; otherwise, it is determined to be a pure noise frequency band and is directly discarded.

[0072] In practical application, the average energy of the large window is defined. Small window energy gradient:

[0073] ;

[0074] Set broadband noise threshold Narrowband gradient threshold Effective range of energy difference between dual windows A frequency band is considered a candidate frequency band if it meets all of the following conditions:

[0075] ;

[0076] By using a three-dimensional joint criterion of global energy base, local abrupt gradient, and window energy difference, the system distinguishes between the stationary characteristics of broadband carriers and the abrupt characteristics of narrowband signals, achieving full coverage detection of broadband, narrowband, and transient weak signals, and significantly improving the detection accuracy and scenario versatility of multiple coexisting signals.

[0077] Preferably, the multi-standard universal demodulation algorithm described in S4 is compatible with mainstream digital modulation standards such as ASK, FSK, PSK, and QAM.

[0078] By matching the time-domain and frequency-domain characteristics of different modulation schemes, blind synchronization and blind demodulation are completed, and the symbol accuracy, constellation diagram convergence, phase deviation and multi-dimensional demodulation confidence index are output to comprehensively determine whether there is a valid carrier signal in the candidate frequency band.

[0079] In practical application, the candidate frequency band signal is bandpass filtered and then blindly demodulated to construct a comprehensive demodulation confidence scoring model:

[0080] ;

[0081] in Score the code element accuracy. Score the convergence of the constellation diagram. The phase deviation score is given. These are normalized weighting coefficients. Pre-set reliability thresholds. ,when If it is determined to be a real carrier, it is determined to be a false alarm caused by noise spikes and is rejected.

[0082] This step constructs a two-layer false alarm suppression architecture of "front-end cepstral noise reduction coarse screening + back-end demodulation feature fine judgment". It breaks away from the limitations of traditional methods that rely solely on energy threshold judgment, and identifies effective signals and random interference from the essential dimension of signal modulation, thereby significantly reducing the false alarm rate in low signal-to-noise ratio scenarios.

[0083] When performing cepstral decomposition in S1, a complex cepstral transformation method is adopted, retaining only the amplitude dimension cepstral component for subsequent filtering and detection processing, and discarding the phase dimension cepstral component to avoid detection interference caused by random phase perturbations.

[0084] In practical applications, radio frequency channel noise and transmission delay can cause random jitter in the cepstral phase component, resulting in no effective signal characteristics. Therefore, only the following parameters are extracted: Amplitude cepstrum is used in energy calculation, filtering and detection, completely eliminating false triggering caused by random phase perturbations and further improving the robustness of the algorithm.

[0085] Working principle: The invention forms a closed-loop detection principle of "cepstral domain adaptive noise reduction - dual-window hierarchical energy coarse detection - demodulation feature fine false alarm elimination", which specifically solves the three core defects of traditional spectrum energy detection. First, logarithmic-IFFT cepstral transform is used to decouple the spectral envelope from noise spikes, unlike traditional frequency domain noise reduction. It utilizes the physical characteristics of high and low cepstral components in the cepstral domain to achieve natural separation of signal and interference. Second, the optimal cutoff cepstral frequency is adaptively solved based on the cepstral energy distribution and attenuation gradient of each frame. Combined with cosine soft threshold gradient filtering, it intelligently adapts to time-varying noise scenarios, completely solving the over-smoothing and under-smoothing problems of fixed parameter filtering, and fully preserving the spectral details of wide and narrow band signals. Third, a synchronous traversal of 8 to 16 times the size of the differential dual sliding windows is adopted. Through a hierarchical mechanism of large window stabilizing the substrate and small window capturing abrupt changes, it takes into account the detection requirements of wideband continuous carrier and narrowband weak transient signals, making up for the scene adaptation shortcomings of single window detection. Finally, the single energy threshold decision logic is abandoned, and a multi-standard universal blind demodulation confidence comprehensive evaluation mechanism is introduced to eliminate noise false peaks from the essential level of signal modulation characteristics, and construct a two-layer false alarm suppression system. This invention, through the innovative coupling of multiple technologies, significantly improves the adaptability, accuracy, and reliability of signal detection in scenarios involving complex electromagnetic interference, low signal-to-noise ratio, and coexistence of wide and narrow band signals.

[0086] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering, characterized in that, Includes the following steps: S1. Signal Time-Frequency Transformation and Cepstral Decomposition: The received RF baseband signal is preprocessed by framing and windowing. The noisy power spectrum is obtained by fast Fourier transform. The power spectrum is then subjected to logarithmic compression transform and inverse fast Fourier transform in sequence to obtain the original cepstral sequence. The characteristics of the cepstral domain are used to separate the low cepstral envelope component carrying the real signal and the high cepstral noise spike component carrying random interference. S2. Intelligent adaptive cepstral low-pass filtering for noise reduction: Statistically analyze the global energy distribution and energy attenuation gradient characteristics of the original cepstral sequence frame by frame, and adaptively solve the optimal cutoff frequency corresponding to each frame signal based on the cumulative energy ratio of the cepstral dimension, thus abandoning the fixed cutoff frequency filtering mode. A dynamic soft-threshold low-pass filter is constructed based on the optimal cutoff frequency. The low-frequency cepstral envelope component is fully preserved, while the high-frequency noise spike component is gradually attenuated and suppressed. The filtered clean cepstral sequence is restored by fast Fourier transform to obtain a smooth spectral envelope curve without noise interference. S3, Dual sliding window hierarchical collaborative energy detection: Configure large and small sliding windows with fixed sizes and differentiating sizes to traverse the smooth spectral envelope curve synchronously with the same length; By integrating the overall energy of the frequency band through a large sliding window, a coarse screening of broadband continuous carriers is completed. By using a small sliding window, the inflection point of spectral energy change is precisely captured, and the local location of weak narrowband signals is locked. By integrating the average energy threshold, local energy gradient, and window energy difference of the dual windows, a set of candidate frequency bands suspected of having effective carriers is obtained. S4. Accurate removal of false alarms during post-demodulation: For all candidate frequency bands, the multi-standard general demodulation algorithm is called to perform carrier synchronization, symbol synchronization and baseband decision processing in sequence, and output the corresponding demodulation confidence index. Candidate frequency bands with demodulation confidence levels below a preset effective threshold are identified as false carriers caused by noise interference and are removed. Frequency bands with valid demodulation features and normal symbol output are retained as the final effective carrier detection results, thus completing the signal energy detection.

2. The dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering according to claim 1, characterized in that: The specific process of adaptively solving for the optimal cutoff reciprocal frequency in S2 is as follows: Extract the energy values ​​of individual frequency points of the original cepstral sequence one by one, and calculate the total energy of the original cepstral sequence; accumulate the energy of individual frequency points segment by segment along the cepstral ascending direction, and calculate in real time the proportion of accumulated energy to the total energy and the energy attenuation slope of adjacent cepstral intervals; A fixed energy percentage threshold is preset, and the inverted frequency point that first meets the cumulative energy percentage threshold is determined as the optimal cutoff frequency of the current frame signal. Each frame signal is independently iterated and updated to achieve adaptive filtering adaptation in scenarios where noise intensity varies over time.

3. The dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering according to claim 1, characterized in that: The filtering rule of the dynamic soft-threshold low-pass filter described in S2 is as follows: Using the optimal cutoff frequency obtained in real time as the boundary, the low frequency components within the boundary are preserved with an all-pass strategy and no amplitude attenuation; the high frequency components outside the boundary are suppressed with a cosine gradient attenuation function to avoid spectral edge distortion caused by hard threshold filtering, and to preserve the subtle spectral characteristics of the narrowband signal to the greatest extent while completely filtering out random noise spikes.

4. The dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering according to claim 1, characterized in that: In S3, the size ratio between the large sliding window and the small sliding window is a fixed adaptation ratio, with the size of the large window being 8 to 16 times that of the small window. The large sliding window is used to smooth random noise in the frequency band and statistically analyze the base energy of the broadband spectrum, adapting to broadband continuous carrier detection; the small sliding window is used to amplify the abrupt change characteristics of local spectrum energy, adapting to the inflection point identification of weak narrowband signals and transient signals. The dual windows slide and sample synchronously to ensure the timing alignment of the spectrum position.

5. The dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering according to claim 1, characterized in that: The multi-dimensional judgment rules for selecting candidate frequency bands in S3 are as follows: A broadband noise reference threshold and a narrowband energy mutation gradient threshold are pre-configured. When the average energy of the large sliding window is greater than the broadband noise reference threshold at the same frequency point, and the local energy gradient corresponding to the small sliding window is greater than the narrowband energy mutation gradient threshold, and the energy difference between the two windows is within the preset effective range, the frequency band where the frequency point is located is determined to be a candidate carrier frequency band; otherwise, it is determined to be a pure noise frequency band and is directly discarded.

6. The dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering according to claim 1, characterized in that: The multi-standard universal demodulation algorithm described in S4 is compatible with mainstream digital modulation standards such as ASK, FSK, PSK, and QAM. By matching the time-domain and frequency-domain characteristics of different modulation schemes, blind synchronization and blind demodulation are completed, and the symbol accuracy, constellation diagram convergence, phase deviation and multi-dimensional demodulation confidence index are output to comprehensively determine whether there is a valid carrier signal in the candidate frequency band.

7. The dual-sliding-window energy detection method based on intelligent adaptive cepstral filtering according to claim 1, characterized in that: When performing cepstral decomposition in S1, a complex cepstral transformation method is adopted, retaining only the amplitude dimension cepstral component for subsequent filtering and detection processing, and discarding the phase dimension cepstral component to avoid detection interference caused by random phase perturbations.

Images