Anti-high dynamic Doppler low signal-to-noise ratio satellite signal blind detection method

Abstract

The invention discloses an anti-high dynamic Doppler low signal-to-noise ratio satellite signal blind detection method, which comprises the following steps of: performing frequency domain analysis on a delay correlation signal according to a signal symbol rate, and dividing a signal band and a noise band near the delay correlation signal to construct a detection quantity to perform constant false alarm rate (CFAR) detection; and finally, obtaining a final signal existence or disappearance result through time consistency judgment. According to the method disclosed by the invention, the analytic CFAR threshold and the state machine smoothing are combined, and the problem of detection failure in a low signal-to-noise ratio and high dynamic Doppler environment in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of satellite signal detection technology, specifically to a blind detection method for low signal-to-noise ratio satellite signals that is resistant to high dynamic Doppler. Background Technology

[0002] In recent years, with the booming development of LEO satellite internet and the Internet of Things industry, satellite communication systems have evolved into a key communication infrastructure for achieving seamless global coverage, including remote mountainous areas, oceans, and high altitudes. Against this backdrop, high-sensitivity detection of LEO satellite downlink signals constitutes a core technology for achieving integrated communication and navigation and enhanced positioning functions using LEO satellites.

[0003] LEO satellites are characterized by high speed and short overhead time, resulting in significant time-varying Doppler frequency shift and Doppler rate of change at the receiver. Furthermore, due to the long distance of the satellite-to-ground link, high path loss, and complex electromagnetic interference, the satellite signals received on the ground are often in a low signal-to-noise ratio (SNR) state, frequently accompanied by bursty or intermittent transmission characteristics. In practical signal detection tasks, receivers often face situations where the specific operating channel of the satellite is uncertain, requiring continuous monitoring of potential frequency points. How to quickly and accurately identify the truly existing weak satellite signals from numerous candidate channels, while avoiding false alarms and frequency estimation jumps caused by transient noise interference, is a key problem that current satellite signal detection and processing technologies urgently need to solve.

[0004] The existing technology is as follows: CN202010044194.0 discloses a method for extracting signal features using squaring operations. The specific steps are as follows: First, a detection decision threshold is set; squaring operations are performed on the collected discrete observation data; the alternating wave components generated by the squaring operation are filtered out by a low-pass filter; then, the processed data is subjected to FFT transformation; finally, the ratio of the maximum amplitude spectral line to the average power is calculated as a decision function and compared with the threshold to determine whether the signal exists. This scheme attempts to extract the energy or envelope features of the signal through nonlinear transformation (squaring) and combines it with the frequency domain peak ratio to combat noise.

[0005] CN202411995766.8 introduces an object detection algorithm from the field of computer vision. Its specific process is as follows: satellite signals are acquired and multiple signal waterfall plots (time-frequency maps) are generated; these waterfall plots are then stitched together at different scales, such as 2, 16, or 52 plots per frame, to form "multi-level" waterfall plots; these images are input into a pre-trained YOLO series object detection model; finally, the detection results are output through fusion strategies of the same level, different levels, and different times. This patent utilizes the powerful feature extraction capabilities of deep learning to identify the time-frequency texture features of signals at the image level, attempting to solve the detection problem in complex scenes.

[0006] The drawbacks of existing technologies are: 1. "Square loss" leads to missed detection of weak signals: in low signal-to-noise ratio (SNR) Under certain conditions, squaring introduces significant noise × noise terms, causing a sharp deterioration in the output signal-to-noise ratio, known as square loss. This makes the originally weak satellite signal submerged by the further amplified noise floor after squaring, making it difficult to extract effective spectral features through FFT, resulting in serious missed detections.

[0007] Sensitive to Doppler frequency shift, increasing processing difficulty: Simple squaring operations will multiply the carrier frequency and Doppler frequency shift of the signal, which will expand the influence range of frequency offset, increase the bandwidth requirements of subsequent processing and the difficulty of frequency search, and is not conducive to locking high dynamic signals.

[0008] Weak anti-interference capability: Squaring is essentially an energy or envelope extraction method, lacking the ability to identify the specific symbol rate of the signal's internal modulation structure, such as SDPSK; as long as there is strong narrowband interference or continuous wave CW interference in the frequency band, squaring it will also produce significant spectral peaks, which can easily trigger false alarms.

[0009] 2. Huge computational resource consumption, making real-time deployment difficult: This method requires generating high-resolution waterfall plots and running large deep learning models such as YOLO, which places extremely high demands on computing resources such as GPUs / NPUs and storage space. It is difficult to meet the engineering requirements of real-time performance and low power consumption in spaceborne processing, handheld terminals, or low-power monitoring devices.

[0010] There is a "signal-to-noise ratio wall": Visual detection based on waterfall plots still essentially relies on the "visibility" of the signal on the time-frequency plot. When the signal energy is below the noise floor, i.e., it is not visible on the waterfall plot, it cannot be detected and deep features cannot be extracted using long-term integral gain. Summary of the Invention

[0011] The purpose of this invention is to address the aforementioned problems by providing a blind detection method for low signal-to-noise ratio satellite signals resistant to high dynamic Doppler. This method involves performing frequency domain analysis on the delay-correlated signal based on the signal symbol rate. The surrounding area is divided into signal and noise bands to construct a detection quantity for constant false alarm rate (CFAR) detection. Finally, the presence or absence of the signal is determined by time consistency decision.

[0012] The technical solution adopted in this invention is as follows: A blind detection method for low signal-to-noise ratio satellite signals resistant to high dynamic Doppler, the method comprising: Signal acquisition and preprocessing: receiving the radio frequency signal from the satellite downlink and preprocessing it to obtain a complex baseband signal, the complex baseband signal including the target SDPSK modulated signal and background noise; Delay processing involves performing time-domain delay conjugate multiplication on the complex baseband signal, with the delay length set to the symbol period of the signal, to obtain the complex baseband signal after delay processing. Feature extraction involves performing a Fourier transform on the delayed complex baseband signal to obtain the delayed correlation spectrum, which includes the cyclic stationary features of the signal. Construct an energy ratio detection statistic. In the acquired delay correlation spectrum, determine the signal band according to the preset symbol rate, select the adjacent frequency band as the noise reference band, calculate the ratio of the signal band energy to the noise reference band energy, and construct the detection statistic. The adaptive CFAR threshold decision derives and sets an adaptive detection threshold independent of background noise power based on the preset false alarm probability and the bandwidth parameters of the noise reference band. The detection statistics are compared with the adaptive detection threshold to generate instantaneous detection results. The time-consistent state machine is smoothed by constructing a state machine model based on a sliding window. The instantaneous detection results of multiple consecutive time windows are accumulated in binary form and logically decided to output the final signal existence state and the corresponding Doppler frequency estimate.

[0013] Furthermore, in the signal acquisition and preprocessing, the acquired radio frequency signal undergoes down-conversion and analog-to-digital conversion to obtain a complex baseband signal, which is expressed as follows: (1) In the formula, It is an SDPSK modulated signal. It is zero-mean complex Gaussian white noise, and its power spectral density is ; The SDPSK modulated signal is as follows: (2) In the formula, For signal amplitude, For carrier frequency, For symbolic time, For impulse functions, For the first The modulation phase of the bit data; set up For baseband data information, for The mapped bipolar data information, As shown in the following formula: (3).

[0014] Furthermore, the feature extraction specifically includes: The complex baseband signal is subjected to symbol period delay correlation processing as follows: (4) Under ideal conditions with no noise influence, substituting the SDPSK signal model into equation (4), we have the signal assumption. As shown in the following formula: (5) No signal hypothesis As shown in the following formula: (6).

[0015] Furthermore, the feature extraction specifically includes: performing a Fourier transform on the signal after time-delay correlation processing, mapping the time-domain features to the frequency domain, and obtaining a time-delay correlation spectrum containing the cyclic stationary features of the signal, as shown in the following equation: (7).

[0016] Furthermore, the construction of the energy ratio detection statistic specifically includes: Let the frequency band near the symbol rate be the signal band. The frequency band adjacent to but not containing spectral peaks is the noise reference band. Define the average spectral energy within the signal band and noise band respectively. and for: (8) (9) Therefore, construct a statistic. As shown in the following formula: (10).

[0017] Furthermore, the adaptive CFAR threshold decision specifically includes: False alarm probability As shown in the following formula: (11) for The false alarm rate is calculated as follows: (Number of noise points) (12) Given a false alarm probability Detection threshold under certain conditions As shown in the following formula: (13) The calculated detection count is compared with the CFAR threshold, and the detection decision rule is as follows: (14).

[0018] Furthermore, the smoothing of the time-consistent state machine specifically includes: Input for current detection status: The input is the instantaneous detection result within the current time window. The system first reads the historical state of the signal that has been determined before the current moment, that is, the system currently believes that the signal exists. State branch judgment: Based on the current signal history state, the system enters different decision logic branches: Signal search / capture branch: When the current system state is unlocked or no signal, it is necessary to determine whether a signal has appeared; Signal tracking / maintenance branch: When the current system state is locked or has a signal, it is necessary to determine whether the signal has disappeared. Signal occurrence decision Logic: In the left branch, the system maintains a length of A sliding window with the following decision condition: count the number of hits contained within the window; if there are consecutive hits... Within a given time window, the cumulative number of hits must not be less than the set threshold. If the signal is valid, then it is considered to have occurred. If the condition is met, the system state flips, an output signal appears, and the channel is locked. If not: the decision condition is not met, the system maintains its current state, and no output is shown. Signal disappearance judgment Logic: In the right-hand branch, the system maintains a length of A sliding window; Decision condition: Count the number of misses contained in the window; if there are consecutive misses... Within a given time window, the cumulative number of misses is greater than or equal to a set threshold. If the signal has ended or been lost, then it is considered that the signal has ended or been lost. Perform the following actions: If: the judgment condition is met, the system state is reversed, the output signal disappears, and the lock is released; If not: If the judgment condition is not met, the system considers the signal to exist and the output remains present.

[0019] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. The method of this invention overcomes the squared loss effect and significantly improves the detection sensitivity under low signal-to-noise ratio; it employs symbol period delay conjugate correlation. The process utilizes conjugate operations to avoid the signal-to-noise ratio degradation caused by noise self-multiplication, and simultaneously leverages the cyclostationary characteristics of the SDPSK signal to coherently accumulate dispersed time-domain energy to specific discrete spectral lines in the transform domain. Experimental results show that this invention can effectively detect SNR as low as [insert value here]. The following weak signal solves the problem of high false negative rate in existing technologies under weak signal environments.

[0020] 2. The method of this invention is naturally immune to Doppler frequency shift, solving the frequency divergence problem of high dynamic signals; by utilizing the conjugate multiplication property, the time-varying Doppler phase rotation terms cancel each other out during the delay correlation process; the characteristic spectral peak positions in the transform domain of this invention depend only on a fixed symbol rate. It is independent of carrier frequency and Doppler shift; this allows the system to stably lock onto the signal without complex frequency offset compensation or broadband search when facing large dynamic Doppler slip of LEO satellites, greatly reducing computational complexity.

[0021] 3. The method of this invention has targeted signal recognition and anti-interference capabilities, effectively reducing the false alarm rate; it constructs a statistical measure of the energy ratio of the signal band and noise band based on a specific symbol rate position; it utilizes the unique cyclic frequency characteristics of SDPSK modulated signals as fingerprints; only signals with a specific symbol rate can form peaks in the detector, and other interference signals that do not have this rate characteristic (regardless of their energy) are regarded as background noise and automatically filtered out; this endows the system with natural anti-interference and signal recognition capabilities, significantly reducing the false alarm rate in complex electromagnetic environments.

[0022] 4. The method of this invention realizes constant false alarm rate (CFAR) detection with noiseless power estimation and strong environmental adaptability; it derives an analytical CFAR threshold formula based on ratio statistics; the detection threshold depends only on the preset false alarm probability and the number of reference units, and is completely decoupled from the actual received signal power and noise power; this means that no matter how the background noise fluctuates, the system can automatically maintain a constant false alarm probability, realizing robust detection that is configured once and adapts at all times.

[0023] 5. The method of this invention outputs a smooth and stable state, which better meets the needs of engineering tracking; it introduces state machine logic based on a dual-threshold sliding window; through binarization accumulation and smoothing of detection results from multiple consecutive frames, it effectively filters out instantaneous false alarms caused by sudden impulse noise and tolerates signal lock-in during short-term fading. The final output signal state stream has good temporal continuity and has extremely high engineering practical value. Attached Figure Description

[0024] Figure 1 This is a flowchart of a low signal-to-noise ratio satellite signal blind detection method resistant to high dynamic Doppler of the present invention; Figure 2 This is a flowchart illustrating the specific processing steps of the delay-correlation and CFAR detection part in the method of this invention; Figure 3 This is a flowchart illustrating the smooth decision-making process of the time-consistent state machine implemented in the method of the present invention. Detailed Implementation

[0025] The present invention will now be described in detail with reference to the accompanying drawings.

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0027] In detection scenarios involving low signal-to-noise ratio, high dynamic range, and low power consumption for modulated signals such as SDPSK, the following common technical bottlenecks generally exist: Performance degradation based on nonlinear transformations such as squaring and fourth-power methods: These techniques generally suffer from low signal-to-noise ratio (SNR) suppression effects. When the input SNR is low, nonlinear operations such as squaring cause noise terms to multiply themselves, resulting in an output SNR loss far greater than that of linear processing, leading to the complete suppression of weak signals. Furthermore, nonlinear transformations typically multiply the Doppler shift, further exacerbating the already severe frequency divergence problem, making them unsuitable for the high dynamic characteristics of LEO satellites. Simultaneously, these methods often lack selectivity for specific signal modulation characteristics such as symbol rate and are not immune to in-band interference.

[0028] Limitations of Deep Learning (Vision-based) Methods in Engineering Applications: While these technologies offer advantages in pattern recognition, their high computational and power consumption make real-time processing difficult on resource-constrained satellite terminals or embedded monitoring devices. More importantly, vision-based methods are limited by the physical resolution and visibility of time-frequency maps, preventing them from extracting statistical features of signals under negative signal-to-noise ratios through mathematical integration, unlike coherent accumulation or cyclostationary processing. This results in a ceiling on their sensitivity.

[0029] Lack of steady-state tracking mechanism for short burst signals: Existing technologies mostly focus on the existence judgment of single-frame signals, such as alarming when a threshold is crossed once, and lack the utilization of the signal's temporal continuity; for signals such as LEO satellites, which are often accompanied by fast fading, blockage, or burst transmission, traditional methods are prone to false alarms due to instantaneous noise fluctuations or missed detections due to short-term fading, resulting in frequent flickering of the detection status and failing to provide stable channel state information for subsequent demodulation or positioning.

[0030] To address the shortcomings of the existing technology, this invention provides a blind detection method for low signal-to-noise ratio satellite signals that is resistant to high dynamic Doppler. Overcoming squared loss and Doppler effects by using time delay correlation: By employing conjugate time delay correlation processing, not only is interference from the squared noise term avoided, but the carrier and most of the Doppler frequency shift can also be automatically canceled through conjugate multiplication (converting high-frequency bias into a fixed phase), which significantly improves the detection performance in low signal-to-noise ratio and high dynamic environments.

[0031] Improving detection sensitivity under low signal-to-noise ratio: By utilizing the cyclostationary characteristics of the signal (constructing a time-delayed correlation spectrum) instead of simple energy integration, and by extracting spectral features at a specific baud rate, satellite signals can be effectively highlighted in noisy environments, thus overcoming the limitations of the signal-to-noise ratio barrier.

[0032] Enhanced anti-interference and signal recognition capabilities: By extracting features for a specific symbol rate Rs, background noise and co-frequency interference that do not conform to the features can be automatically filtered out, significantly reducing the false alarm rate.

[0033] To achieve continuous and stable channel tracking: a sliding window state machine logic is introduced to smooth the instantaneous decision result by utilizing the historical detection state of the signal, effectively suppressing detection flicker caused by signal fading.

[0034] Example This embodiment provides a method for blind detection of low signal-to-noise ratio satellite signals that is resistant to high dynamic Doppler, such as... Figure 1 and Figure 2 As shown, please follow these steps: Step 1, Signal Acquisition and Preprocessing: Receive the satellite downlink RF signal, and obtain the complex baseband signal through down-conversion and analog-to-digital conversion. The complex baseband signal includes the target SDPSK modulated signal and background noise, as detailed below: The system first loads the raw baseband IQ signal data stream acquired by the receiver, assuming the received complex baseband signal... for (1) In the formula, It is an SDPSK modulated signal. It is zero-mean complex Gaussian white noise, and its power spectral density is ; The SDPSK signal can be represented as: (2) In the formula, For signal amplitude, For carrier frequency, For symbolic time, For impulse functions, For the first The modulation phase of the bit data.

[0035] set up For baseband data information, for The mapped bipolar data information, The following conditions must be met: (3) Initialization parameters: Set key system parameters, including sampling rate, symbol rate of ORBCOMM satellite signal, number of FFT points, false alarm probability, and sliding window size, etc. Dividing time windows: In order to adapt to the time-varying characteristics of satellite signals and perform continuous detection, the continuous signal stream is divided into several time windows of fixed length; Signal mixing to various channels: Since the receiving bandwidth contains multiple channels, the target frequency is shifted to zero intermediate frequency by digital down-conversion (DDC) to achieve parallel processing of multiple channels; Low-pass filtering & amplitude normalization: The mixed signal is low-pass filtered to remove out-of-band noise and amplitude normalized to eliminate the impact of excessive signal dynamic range on subsequent calculations.

[0036] Step 2, Symbol Period Delay Correlation Processing: The complex baseband signal undergoes time-domain delay conjugate multiplication, with the delay length set to the symbol period. This step eliminates the influence of unknown carrier phase and Doppler frequency offset in the signal, concentrating the signal energy to the characteristic frequency corresponding to the symbol rate. The specific details are as follows: Delay Correlation: To eliminate unknown carrier phase and highlight differential structure, the received signal is subjected to symbol period delay correlation processing, as shown in the following formula: (4) Under ideal conditions where noise is ignored, substituting the SDPSK signal model into the above equation, we have the signal assumption. As shown in the following formula: (5) That is, the delayed-correlated output exhibits a deterministic structure at the symbol rate; No signal hypothesis As shown in the following formula: (6) Its statistical characteristics are zero-mean complex Gaussian process.

[0037] Step 3, transform domain feature extraction: Perform a Fourier transform on the signal after delay correlation processing to map the time-domain features to the frequency domain, obtaining the delay correlation spectrum containing the cyclostationary features of the signal. The specific content is as follows: FFT spectral analysis: To further extract the statistical characteristics of the SDPSK signal, the delay-correlated signal is analyzed. Performing an FFT Fourier transform, its frequency domain As shown in the following formula: (7) Then in the frequency domain: exist Below, because the differential phase sequence of the SDPSK signal is based on the symbol period... As the time scale changes, its phase transition rate changes from the symbol rate. This is determined by the time domain; therefore, the time-delay correlated signal exhibits periodic characteristics of periodic correlation. This periodicity will manifest in the frequency domain as... The energy concentration in the vicinity. That is, the time-delay correlation spectrum when the SDPSK signal is present. A distinct spectral peak appears near the corresponding frequency; exist Down, It exhibits an approximately flat noise spectrum.

[0038] Therefore, detection can be achieved by comparing the energy near the symbol rate frequency point with the noise energy in the neighborhood.

[0039] Step 4: Construct the energy ratio detection statistic. In the time-delay correlation spectrum, determine the signal band based on the preset symbol rate, and select a neighboring frequency band as the noise reference band. Calculate the ratio of the signal band energy to the noise reference band energy to construct a dimensionless detection statistic. The specific content is as follows: The obtained time-delay correlation spectrum exist The peak characteristics at the peak value are analyzed, and a detection statistic based on the energy ratio is constructed; the frequency band near the symbol rate is defined as the signal band. The frequency band adjacent to but not containing spectral peaks is the noise reference band. .

[0040] Define the average spectral energy within the signal band and noise band respectively. and for: (8) (9) Therefore, construct the statistic for: (10) It effectively reflects the degree of enhancement of spectral energy near the symbol rate relative to background noise.

[0041] Step 5, Adaptive CFAR Threshold Decision: Based on the preset false alarm probability and the bandwidth parameters of the noise reference band, an adaptive detection threshold independent of the background noise power is derived and set. The detection statistics are compared with the adaptive detection threshold to generate an instantaneous detection result. The specific content is as follows: CFAR Detection: To achieve constant false alarm rate (CFAR) detection, the false alarm probability is... As shown in the following formula: (11) For a cell-averaged constant false alarm rate detector based on noise power averaging estimation, for The probability of the above-mentioned noise points can be analyzed as follows: (12) Therefore, given a false alarm probability... Detection threshold under certain conditions for: (13) Detection decision: The calculated detection quantity is compared with the CFAR threshold. The detection decision rule is as follows: (14) This threshold is independent of noise power, thus ensuring that the detector maintains a constant false alarm probability under varying noise conditions; Save detection information: Store the current decision result (hit / miss) and other information in the state cache or log. This information is usually passed as input to the subsequent "state machine logic" for multi-frame smoothing and continuity verification.

[0042] Step 6: Time-consistent state machine smoothing. A sliding window-based state machine model is constructed. Binary accumulation and logical decision-making are performed on the instantaneous detection results of multiple consecutive time windows. The final signal existence state and the corresponding Doppler frequency estimate are output. The specific details are as follows: A dual-state (locked / unlocked) sliding window logic with hysteresis characteristics is introduced to perform binarization filtering on the instantaneous detection results.

[0043] Logical Model and Details: Defined length as The sliding time window, currently the [number]th [time window]. The instantaneous frame detection result is ; Locking condition (signal occurrence): When the system is in "search state", if If the signal is detected, the system enters "tracking mode". Release condition (signal disappearance): When the system is in "tracking state", if If the signal is lost, the system returns to the "search state".

[0044] The specific steps are as follows: Input the current detection status: The input to the process is the instantaneous detection result within the current time window, such as hit / miss. The system first reads the historical signal status determined before the current moment, i.e., the system's current assessment of whether the signal exists.

[0045] State branch judgment: Based on the current signal history state, the system enters different decision logic branches: such as Figure 3The left-hand path, signal search / capture branch: When the system's current state is "unlocked" or "no signal," it is necessary to determine whether a signal has appeared; for example... Figure 3 The right-hand path, signal tracking / maintenance branch: When the current system state is "locked" or "signal present", it is necessary to determine whether the signal has disappeared; Signal occurrence decision Logic: In the left branch, the system maintains a length of Sliding window; Judgment criteria: Count the number of "hit" occurrences within the window. If there are consecutive hits... Within a given time window, the cumulative number of hits is greater than or equal to a set threshold. If the signal is valid, then it is considered to have occurred.

[0046] Perform the following actions: Yes: When the condition is met, the system state flips and outputs "Signal appears, channel locked". This step effectively filters out occasional noise spikes, ensuring that only a persistent signal can trigger locking.

[0047] No: The judgment condition is not met. The system maintains the current state and outputs "Remains not present".

[0048] Signal disappearance judgment Logic: In the right-hand branch, the system maintains a length of A sliding window.

[0049] Judgment criteria: Count the number of "misses" contained in this window. If there are consecutive Within a given time window, the cumulative number of misses is greater than or equal to a set threshold. If the signal is lost, it is considered to have ended or been lost.

[0050] Perform the following actions: Yes: When the judgment condition is met, the system state reverses and the output "Signal disappears, lock released" is displayed.

[0051] No: The judgment condition is not met. That is, although there may be a small number of missed detections, the loss threshold has not been reached. The system considers the signal to still exist (in short-term fading) and outputs "Keep Present". This step gives the system a certain "fault-tolerant memory" capability to prevent detection flickering caused by signal obstruction or fading. Update State History: Regardless of any branch, the system's state register and historical record list will be updated based on the judgment result of this round, such as locking or releasing, keeping present or keeping not present, as the prior state for the next frame.

[0052] In summary, existing satellite signal detection technologies, whether traditional methods based on nonlinear transformations or deep learning-based visual methods, generally face significant technical bottlenecks in low-Earth orbit satellite communication scenarios. Limited by the noise self-multiplication effect of nonlinear operations or the resolution limitations of time-frequency images, these methods struggle to effectively extract weak signal features at low signal-to-noise ratios and are sensitive to high-dynamic Doppler frequency shifts, easily leading to energy dissipation and an expanded frequency search range. Furthermore, due to a lack of identification of specific signal modulation structures and over-reliance on noise power estimation, their anti-interference capabilities and adaptability in complex electromagnetic environments are poor. In addition, the high computational cost and lack of time smoothing mechanisms make it difficult to achieve stable, real-time continuous tracking on resource-constrained terminals.

[0053] The present invention provides a signal detection architecture based on the ratio of conjugate delay correlation and spectral features, which combines analytical CFAR threshold and state machine smoothing, and solves the detection failure problem of existing technologies in low signal-to-noise ratio and high dynamic Doppler environments. The delay length of the signal is the symbol period. The conjugate multiplication operation, rather than a simple amplitude square or energy integral, is used in existing technologies. The operation will cause the carrier frequency and Doppler frequency offset to double. This increases the difficulty of frequency search; while the conjugate operation of this invention utilizes Its characteristics convert the time-varying carrier phase and most of the Doppler frequency shift into time-varying characteristics. Unrelated fixed phase terms. This allows the signal energy to be precisely focused in subsequent spectral analysis, without spectral dispersion due to Doppler shift.

[0054] The method for constructing the detection statistic proposed in this invention does not search for the maximum value across the entire frequency band, but instead locks onto a specific cyclic frequency corresponding to the symbol rate, calculating the ratio of the signal band energy to the noise reference band energy. Existing technologies typically cannot distinguish between strong narrowband interference and target signals. This invention utilizes the physical characteristics of SDPSK signals: only signals with a symbol rate of [missing information]... The signal, its delay correlation spectrum will only appear in [the specified timeframe]. A peak value is formed at this frequency. Any other modulation scheme or narrowband interference CW will appear as noise floor at this frequency. This statistic enables simultaneous detection and identification.

[0055] The detection threshold setting method proposed in this invention calculates the threshold value directly using an analytical formula. It is only related to the preset false alarm probability and the number of reference units, and is completely independent of the real-time power of the received signal and the background noise power. Traditional energy detection requires accurate estimation of the noise variance. Once the estimation deviation becomes extremely large, it is known as the "signal-to-noise ratio wall"; the ratio statistic of this invention eliminates this problem. Term, making the threshold This becomes a constant independent of the environment. This means that in non-stationary noise environments with drastic fluctuations in satellite signal strength or changes in antenna pointing, the system can maintain a constant false alarm rate without recalibration.

[0056] The proposed time consistency decision method for continuous signals introduces a dual-state (locked / unlocked) sliding window logic with hysteresis characteristics. It performs binarization filtering on the instantaneous detection results. ORBCOMM signals often experience missed detections due to channel fading or obstruction within a short period. Existing single-frame detection techniques are prone to "flickering" of the detection state. The state machine logic of this invention allows for missed detections within a short period without losing lock, while requiring multiple consecutive hits to trigger lock, greatly reducing the probability of false alarms caused by transient pulse interference.

[0057] This article uses specific embodiments to illustrate the principles and implementation methods of the present invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make several improvements and modifications to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A method for blind detection of low signal-to-noise ratio satellite signals resistant to high dynamic Doppler, characterized in that, The method includes: Signal acquisition and preprocessing: receiving the radio frequency signal from the satellite downlink and preprocessing it to obtain a complex baseband signal, the complex baseband signal including the target SDPSK modulated signal and background noise; Delay processing involves performing time-domain delay conjugate multiplication on the complex baseband signal, with the delay length set to the symbol period of the signal, to obtain the complex baseband signal after delay processing. Feature extraction involves performing a Fourier transform on the delayed complex baseband signal to obtain the delayed correlation spectrum, which includes the cyclic stationary features of the signal. Construct an energy ratio detection statistic. In the acquired delay correlation spectrum, determine the signal band according to the preset symbol rate, select the adjacent frequency band as the noise reference band, calculate the ratio of the signal band energy to the noise reference band energy, and construct the detection statistic. The adaptive CFAR threshold decision derives and sets an adaptive detection threshold independent of background noise power based on the preset false alarm probability and the bandwidth parameters of the noise reference band. The detection statistics are compared with the adaptive detection threshold to generate instantaneous detection results. The time-consistent state machine is smoothed by constructing a state machine model based on a sliding window. The instantaneous detection results of multiple consecutive time windows are accumulated in binary form and logically decided to output the final signal existence state and the corresponding Doppler frequency estimate.

2. The method for blind detection of low signal-to-noise ratio satellite signals resistant to high dynamic Doppler according to claim 1, characterized in that, In the signal acquisition and preprocessing, the acquired radio frequency signal is down-converted and analog-to-digital converted to obtain a complex baseband signal, which is expressed as follows: （1） In the formula, It is an SDPSK modulated signal. It is zero-mean complex Gaussian white noise, and its power spectral density is ; The SDPSK modulated signal is as follows: （2） In the formula, For signal amplitude, For carrier frequency, For symbolic time, For impulse functions, For the first The modulation phase of the bit data; set up For baseband data information, for The mapped bipolar data information, As shown in the following formula: （3）。 3. The method for blind detection of low signal-to-noise ratio satellite signals resistant to high dynamic Doppler according to claim 2, characterized in that, The feature extraction specifically includes: The complex baseband signal is subjected to symbol period delay correlation processing as follows: （4） Under ideal conditions with no noise influence, substituting the SDPSK signal model into equation (4), we have the signal assumption. As shown in the following formula: （5） No signal hypothesis As shown in the following formula: （6）。 4. The method for blind detection of low signal-to-noise ratio satellite signals resistant to high dynamic Doppler according to claim 3, characterized in that, The feature extraction specifically includes: performing a Fourier transform on the signal after time-delay correlation processing, mapping the time-domain features to the frequency domain, and obtaining a time-delay correlation spectrum containing the cyclic stationary features of the signal, as shown in the following equation: （7）。 5. The method for blind detection of low signal-to-noise ratio satellite signals resistant to high dynamic Doppler according to claim 4, characterized in that, The specific components of the constructed energy ratio detection statistics include: Let the frequency band near the symbol rate be the signal band. The frequency band adjacent to but not containing spectral peaks is the noise reference band. Define the average spectral energy within the signal band and noise band respectively. and for: （8） （9） Therefore, construct a statistic. As shown in the following formula: （10）。 6. The method for blind detection of low signal-to-noise ratio satellite signals resistant to high dynamic Doppler according to claim 5, characterized in that, The adaptive CFAR threshold decision specifically includes: False alarm probability As shown in the following formula: （11） for The false alarm rate is calculated as follows: (Number of noise points) （12） Given a false alarm probability Detection threshold under certain conditions As shown in the following formula: （13） The calculated detection count is compared with the CFAR threshold, and the detection decision rule is as follows: （14）。 7. The method for blind detection of low signal-to-noise ratio satellite signals resistant to high dynamic Doppler according to claim 6, characterized in that, The smoothing of the time-consistent state machine specifically includes: Input for current detection status: The input is the instantaneous detection result within the current time window. The system first reads the historical state of the signal that has been determined before the current moment, that is, the system currently believes that the signal exists. State branch judgment: Based on the current signal history state, the system enters different decision logic branches: Signal search / capture branch: When the current system state is unlocked or no signal, it is necessary to determine whether a signal has appeared; Signal tracking / maintenance branch: When the current system state is locked or has a signal, it is necessary to determine whether the signal has disappeared. Signal occurrence decision Logic: In the left branch, the system maintains a length of A sliding window with the following decision condition: count the number of hits contained within the window; if there are consecutive hits... Within a given time window, the cumulative number of hits must not be less than the set threshold. If the signal is valid, then it is considered to have occurred. If the condition is met, the system state flips, an output signal appears, and the channel is locked. If not: the decision condition is not met, the system maintains its current state, and no output is shown. Signal disappearance judgment Logic: In the right-hand branch, the system maintains a length of A sliding window; Decision condition: Count the number of misses contained in the window; if there are consecutive misses... Within a given time window, the cumulative number of misses is greater than or equal to a set threshold. If the signal has ended or been lost, then it is considered that the signal has ended or been lost. Perform the following actions: If: the judgment condition is met, the system state is reversed, the output signal disappears, and the lock is released; If not: If the judgment condition is not met, the system considers the signal to exist and the output remains present.

Images