High-efficiency acquisition method for low signal-to-noise ratio direct sequence spread spectrum channels

Abstract

This invention discloses a high-efficiency acquisition method for low signal-to-noise ratio (SNR) direct sequence spread spectrum channels, comprising: performing matched filtering and signal decimation on the analog-to-digital converted signal; selectively phase-shifting and vector accumulation on at least two sets of signal frame segments in the decimated signal, fusing them into an accumulated signal; performing frequency domain transformation and analysis on the accumulated signal to obtain spectral data; retrieving and identifying candidate peaks from the spectral data, and outputting the signal arrival position and frequency offset. Preferably, this invention applies a peak timing phase fingerprint verification to the candidate peaks, identifying the true signal peaks by establishing a phase evolution model and performing bidirectional consistency verification; and, during vector accumulation, employing a high-precision phase-shifting method based on a non-uniform lookup table and error diffusion mechanism to compensate for quantization errors. This invention effectively improves acquisition performance and reliability under low SNR conditions.

Description

Technical Field

[0001] This invention relates to a method for acquiring a direct sequence spread spectrum channel, and more particularly to a high-efficiency method for acquiring a low signal-to-noise ratio direct sequence spread spectrum channel. Background Technology

[0002] Direct sequence spread spectrum (DSSS) communication technology plays a crucial role in modern wireless communication systems due to its inherent advantages such as anti-interference, anti-fading, low probability of intercept, and support for code division multiple access. Especially in satellite communication, particularly for applications such as the Internet of Things (IoT) and low data rate transmission, DSSS communication allows the use of small-aperture, low-cost user terminal antennas while effectively reducing interference to nearby satellites, making it technically valuable. However, these applications often present challenges such as weak signals and harsh channel conditions. To establish a communication link quickly and reliably under extremely low signal-to-noise ratio (SNR) conditions, the receiver must prioritize accurate acquisition of the spread spectrum signal. Therefore, researching efficient acquisition methods in low SNR DSSS channels is not only a key technological bottleneck for improving DSSS system performance but also a necessary prerequisite for its widespread application in satellite IoT, emergency communications, and deep space exploration, possessing significant engineering application value.

[0003] Currently, various acquisition techniques have been developed for direct sequence spread spectrum signals. In hardware implementation, a fundamental and widely used method is the time-domain sliding correlator. This method performs point-by-point sliding and correlation operations between the received signal and a locally generated pseudo-random (PN) preamble sequence. When the correlation value exceeds a preset threshold, the signal is considered acquired. To improve computational efficiency, frequency-domain acquisition methods have been introduced. A typical implementation involves performing a Fast Fourier Transform (FFT) on the received signal, transforming the time-domain correlation operation into a frequency-domain multiplication operation, thus reducing computational complexity to some extent. When facing extremely low signal-to-noise ratio (SNR), existing technologies generally employ extending the preamble sequence length to accumulate more signal energy in exchange for improved acquisition performance. For example, Chinese Patent Publication No. CN109792771B provides a scheme that doubles the length of the preamble sequence, resulting in an approximately 3dB SNR gain. Furthermore, to address the frequency offset between the receiver and transmitter, some solutions employ a strategy based on multiple parallel processing channels. Each channel searches for a pre-defined frequency offset assumption, thus casting a wide net to cover potential frequency uncertainties. These methods constitute the mainstream technical framework in the current field of DSSS signal acquisition.

[0004] While existing technologies offer a range of signal acquisition solutions, their inherent limitations are becoming increasingly apparent as applications demand higher efficiency, greater robustness, and lower implementation complexity. For example, existing solutions still suffer from low acquisition performance at low signal-to-noise ratios and insufficient reliability in extremely harsh channel conditions. Summary of the Invention

[0005] This application provides a high-efficiency acquisition method for low signal-to-noise ratio direct sequence spread spectrum channels to solve the two core problems of low acquisition performance under low signal-to-noise ratio and insufficient reliability under extremely harsh channel conditions.

[0006] According to one aspect of this application, matched filtering and signal decimation are performed on the analog-to-digital converted signal to obtain the decimated signal; selective phase shifting and vector accumulation are performed on at least two sets of signal frame segments in the decimated signal to fuse them into an accumulated signal; frequency domain transformation and analysis are performed on the accumulated signal to obtain spectrum data; candidate peaks that meet a preset threshold are retrieved from the spectrum data and their candidate peak information is extracted; and the signal arrival position and frequency offset are calculated and output based on the candidate peak information.

[0007] In an exemplary embodiment, calculating and outputting the signal arrival position and frequency offset based on candidate peak information includes: applying a peak timing phase fingerprint verification to the candidate peak information in combination with the accumulated signal to obtain the real signal peak information; and calculating and outputting the signal arrival position and frequency offset based on the real signal peak information.

[0008] In an exemplary embodiment, combining the accumulated signal and applying a peak time-series phase fingerprint verification to obtain the true signal peak information includes: decomposing the accumulated signal into multiple sliding windows with time overlap, and performing frequency domain transformation on the data in each sliding window to construct a window spectrum sequence; in the window spectrum sequence, for each candidate peak in the candidate peak information, extracting its phase value at the corresponding frequency position in each window spectrum to form a phase evolution sequence characterizing the phase change of the candidate peak over time, and extracting the true signal peak information accordingly.

[0009] In an exemplary embodiment, extracting real signal peak information based on a phase evolution sequence includes: mathematically modeling the phase evolution sequence corresponding to each candidate peak, fitting a quadratic polynomial model using the least squares method to obtain a set of model parameters characterizing the dynamic characteristics of the sequence; quantifying the phase curvature or phase continuity score of the phase evolution sequence based on the model parameters, and determining whether the corresponding candidate peak is a real signal peak based on the quantization result, and extracting peak information.

[0010] In an exemplary embodiment, performing selective phase shifting and vector accumulation includes: based on a pre-stored target phase shift angle, searching or interpolating to calculate a complex phase shift coefficient with a predetermined precision in a pre-constructed non-uniform resolution lookup table with different angular resolutions in different phase intervals; performing complex multiplication on the signal frame segment using the complex phase shift coefficient to obtain a precisely phase-shifted signal frame segment, and then performing subsequent vector accumulation.

[0011] In one exemplary embodiment, a phase shift error diffusion mechanism is further included, which calculates a complex phase shift coefficient based on a target phase shift angle and performs calculations using the coefficient, including: reading the historical phase shift error stored in the previous phase shift operation and compensating the current target phase shift angle based on the historical phase shift error to obtain a compensated phase shift angle; determining a complex phase shift coefficient of predetermined precision in a non-uniform resolution lookup table based on the compensated phase shift angle; and after performing a phase shift operation on a signal frame segment using the complex phase shift coefficient, calculating the difference between the actual phase shift angle it represents and the target phase shift angle, generating and storing the current phase shift error as the historical phase shift error for the next phase shift operation.

[0012] In one exemplary embodiment, the non-uniform resolution lookup table is dynamically updated, and the update method includes: periodically statistically analyzing the frequency distribution of the target phase shift angle used in past phase shift operations to identify the high-frequency used angle range; adaptively increasing the angle resolution of the high-frequency used angle range and updating the non-uniform resolution lookup table to provide more suitable phase shift accuracy for commonly used angles in subsequent phase shift coefficient calculations.

[0013] In an exemplary embodiment, performing selective phase shifting and vector accumulation further includes a structured search strategy for the two sets of signal frame segments. Specifically, this involves: performing direct vector accumulation on the two sets of signal frame segments to obtain a first accumulated signal, and performing subsequent frequency domain transformation and analysis, as well as peak identification, on the first accumulated signal; if the first accumulated signal fails to identify a candidate peak that meets a preset threshold, applying a π-phase shift to one set of signal frame segments and performing vector accumulation with the other set of signal frame segments to obtain a second accumulated signal, and performing frequency domain transformation and analysis, as well as peak identification, on the second accumulated signal.

[0014] In an exemplary embodiment, the strategy further includes: if two attempts of the structured search strategy for the two sets of signal frame segments fail to identify a candidate peak that meets a preset threshold, the number of accumulated signal frame segments is increased to four; and for the four sets of signal frame segments, different preset phase combinations are used sequentially for phase shifting and vector accumulation, and the accumulated signal generated by each combination is used to perform frequency domain transformation and analysis and peak identification until a candidate peak is successfully identified.

[0015] In an exemplary embodiment, frequency domain transformation and analysis are performed on the accumulated signal to obtain spectral data, including: performing frequency domain transformation and analysis on the accumulated signal, specifically: performing point-by-point sliding window data acquisition on the accumulated signal and performing correlation operation with the local sequence to obtain a correlated signal; dividing the correlated signal into a predetermined number of groups according to the pre-stored chip rate representing the system symbol rate, and summing the data in each group to generate a grouped summed sequence; and performing a fast Fourier transform on the grouped summed sequence to obtain spectral data.

[0016] By adopting the above technical solution, the acquisition performance under low signal-to-noise ratio can be effectively improved, and the reliability under extremely harsh channel conditions can be enhanced. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0018] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0019] Figure 1 This is a flowchart of a method for high-efficiency acquisition of a low signal-to-noise ratio direct sequence spread spectrum channel according to an embodiment of this application.

[0020] Figure 2 This is a flowchart illustrating the calculation and output of signal arrival position and frequency offset based on candidate peak information according to an embodiment of this application.

[0021] Figure 3 This is a flowchart illustrating the process of obtaining real signal peak information by applying peak timing phase fingerprint verification according to an embodiment of this application.

[0022] Figure 4 This is a flowchart illustrating the extraction of real signal peak information based on a phase evolution sequence according to an embodiment of this application.

[0023] Figure 5 This is a flowchart illustrating the selective phase shifting and vector accumulation performed according to an embodiment of this application.

[0024] Figure 6 The spectrum is obtained by performing a PMF_FFT operation on the signal after matching filtering according to the embodiments of this application, without accumulation.

[0025] Figure 7 The spectrum is obtained by performing PMF_FFT operation after accumulating twice according to the embodiments of this application.

[0026] Figure 8 The spectrum is obtained by performing PMF_FFT operation after accumulating twice the amount of phase shift π according to the embodiment of this application.

[0027] Figure 9 The spectrum is obtained by performing PMF_FFT operation after accumulating the phase shifted π by 4 times according to the embodiment of this application. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. 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 should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in sequences other than those illustrated or described herein.

[0030] It should be noted that, for the purpose of clearly demonstrating the steps of this application, each step has been numbered in the specification. These numbers are for ease of explanation only and do not limit the execution order of the steps. In actual operation, depending on the technical requirements of the specific implementation scenario, the steps may be executed in a different order than that shown in the specification, and in some cases, parallel processing between steps can also be achieved.

[0031] like Figures 1 to 9 As shown in Example 1, a high-efficiency acquisition method for low signal-to-noise ratio direct sequence spread spectrum channels is proposed, including the following steps:

[0032] Step S101: Perform matched filtering on the analog-to-digital converted signal to generate a matched-filtered signal.

[0033] In the embodiments of this application, the signal after analog-to-digital conversion refers to the digital complex baseband signal obtained after the analog intermediate frequency signal from the antenna is down-converted, filtered, and sampled by the analog-to-digital converter (ADC).

[0034] The impulse response of matched filtering is the time inversion and conjugate of a known signal waveform. In this invention, it refers to the process of performing correlation operations between the received digital signal and a locally stored, known pseudo-random (PN) preamble sequence.

[0035] Specifically, the ADC samples at a frequency four times the chip rate. A matched filter module internally stores a preamble sequence identical to that of the transmitter (e.g., 3072 chips in length). The digital complex baseband signal stream output from the ADC is continuously fed into this matched filter for complex correlation operations, and the output signal is the matched-filtered signal Sm.

[0036] Specifically, the purpose of matched filtering is to maximize the signal-to-noise ratio (SNR) of a particular signal in a noisy environment. By correlating with a known preamble sequence, the signal energy scattered over a long symbol period can be concentrated into a single peak, making it easier to detect the presence of the signal in subsequent detection steps.

[0037] Step S102: Extract the signal from the matched-filtered signal to obtain the extracted signal. In the embodiments of this application, the matched-filtered signal is extracted by a factor of four to reduce the computational load of subsequent processing steps.

[0038] Decimation is a downsampling operation in digital signal processing (DSP) used to extract one point from a discrete signal sequence at regular intervals. Quadruple decimation means retaining one point from every four consecutive sample points.

[0039] Specifically, a signal decimation module processes the matched-filtered signal Sm output in step S101. Since the ADC uses a quadruple sampling rate, the sampling rate of Sm is also four times the chip rate. This module takes one point every four sampling points to generate the decimated signal Sd. For example, from Sm(0), Sm(1), Sm(2), Sm(3), Sm(4)..., the decimator may extract Sm(0), Sm(4), Sm(8)... to generate Sd(0), Sd(1), Sd(2)....

[0040] Specifically, this step reduces data processing volume and computational complexity. By quadrupling the sampling, the amount of data for all subsequent processing (such as accumulation, FFT, etc.) is reduced by 75%, which is crucial for implementing high-speed signal processing on resource-constrained hardware (such as FPGAs). It should be noted that this step introduces two controllable issues: a) Since the sampling position may be random, the optimal sampling point may be missed, resulting in a signal-to-noise ratio loss of approximately 1-2 dB; b) The detected signal position may have an error of approximately ±2 sampling points. These issues will be compensated for and corrected through subsequent accumulation gain and precise re-search steps.

[0041] As can be understood, the above means performing matched filtering and signal extraction on the signal after analog-to-digital conversion to obtain the extracted signal.

[0042] Step S103: Perform selective phase shifting and vector accumulation on at least two sets of signal frame segments in the extracted signal to fuse them into an accumulated signal.

[0043] A signal frame segment refers to a segment of signal length equal to the length L of a physical frame, extracted from the extracted signal Sd. For example, Sd(0) to Sd(L-1) is the first frame segment, and Sd(L) to Sd(2L-1) is the second frame segment. Vector accumulation refers to the complex summation of multiple complex signal frame segments at their corresponding positions.

[0044] Specifically, in this embodiment, the most basic two-frame direct accumulation is performed. An accumulator module stores the first signal frame segment Sd(0)...Sd(L-1) into local RAM. Further, when the second signal frame segment Sd(L)...Sd(2L-1) arrives, it is added point-by-point to the first frame segment stored in RAM to generate the accumulated signal Sa. That is: Sa(i)=Sd(i)+Sd(L+i), where i ranges from 0 to L-1.

[0045] Specifically, since the preamble sequence of each frame in a spread spectrum communication system is fixed and identical in continuous or quasi-continuous channel mode, by coherently accumulating corresponding parts of multiple frames of signals, the signal components will be superimposed in phase, while the noise components will be incoherently superimposed. This can effectively improve the signal-to-noise ratio (SNR). Theoretically, ideal accumulation of two frames can bring a 3dB SNR gain, which is sufficient to compensate for the loss caused by decimation in step S102 and obtain additional processing gain.

[0046] Step S104: Perform frequency domain transformation and analysis on the accumulated signal to obtain spectral data. This includes performing frequency domain transformation and analysis on the accumulated signal, specifically: performing point-by-point sliding window data acquisition on the accumulated signal and performing correlation operations with the local sequence to obtain the correlated signal; dividing the correlated signal into a predetermined number of groups according to the pre-stored chip rate representing the system symbol rate, and summing the data within each group to generate a grouped summed sequence; and performing a Fast Fourier Transform on the grouped summed sequence to obtain spectral data.

[0047] PMF-FFT (Partial Matched Filter-Fast Fourier Transform) is used for frequency domain acquisition, and this step is its specific implementation. Optionally, this step is completed by a PMF-FFT module, and the specific steps are as follows:

[0048] Step 1: Point-by-point sliding window and correlation. Perform point-by-point sliding window operation on the accumulated signal Sa(0)...Sa(L-1) generated in step S103. Each time, take out a window with a length equal to the length of the preceding sequence (e.g., 3072 points) and perform correlation operation with the locally stored preceding sequence to obtain the correlated signal Sc(0)...Sc(3071) with a length of 3072 points.

[0049] Step 2: Group Summation. Based on the system chip rate Rc, sum the Sc in groups.

[0050] Optionally, when the chip rate is 8 Mcps > Rc ≥ 4 Mcps, Sc is divided into 128 groups, with 24 points in each group. The sequence after grouping and summing is Sca(0) = Sc(0) + ... + Sc(23), Sca(1) = Sc(24) + ... + Sc(47), ..., until Sca(127).

[0051] When Rc≥8Mcps, the data is divided into 64 groups, with 48 points in each group.

[0052] When Rc < 4 Mcps, the data is divided into 256 groups, with 12 points in each group.

[0053] Step 3: Fast Fourier Transform. Perform a 128-point FFT on the grouped and summed sequence Sca (e.g., 128 points in length) to obtain the final spectral data F(0)...F(127).

[0054] Specifically, this PMF-FFT structure decomposes a large time-domain correlation operation into multiple smaller correlations (achieved through grouping and summing) and an FFT operation. The advantage is that the phase rotation caused by the carrier frequency offset manifests as a peak shift at the FFT frequency. By searching for the peak in the frequency domain, signal location detection and a coarse estimation of the frequency offset can be performed simultaneously, with a computational cost far less than compensating for and searching for each possible frequency offset in the time domain. Different grouping methods correspond to different frequency offset estimation ranges; for example, 128 groups can support a frequency offset range of no less than ±166.47 kHz (when Rc ≥ 1 Mcps).

[0055] Step S105: Search the spectrum data and identify candidate peaks that meet the preset threshold, and extract their candidate peak information; calculate and output the signal arrival position and frequency offset based on the candidate peak information.

[0056] Candidate peak information typically includes the peak amplitude, the FFT frequency index of the peak, and the starting position of the corresponding time window. Signal arrival position refers to the starting sampling point position of the preamble sequence in the received signal stream. Frequency offset refers to the difference between the carrier frequency of the received signal and the frequency of the receiver's local oscillator.

[0057] Specifically, the threshold detection module processes the spectrum data F output in step S104. Peak retrieval refers to searching for the point with the largest amplitude in the spectrum data F(0)...F(127). Threshold judgment refers to comparing the amplitude of the peak with a preset noise threshold. This threshold can be dynamically adjusted according to the channel environment or set to a fixed value based on experience. If the peak amplitude exceeds the threshold, it is determined to be a successful capture, and the peak becomes a candidate peak. Further, information calculation and output specifically include the signal arrival location and frequency offset.

[0058] The signal arrival location is the starting position of the time window corresponding to the PMF-FFT operation that generated the peak value, which is a rough estimate of the signal arrival location. Frequency offset calculation means that the FFT frequency index of the candidate peak value directly corresponds to the frequency offset magnitude. For example, if the peak value is at the k-th frequency point, then the frequency offset Δf = k * (sampling rate / number of FFT points). The system ultimately outputs these two pieces of information. If the peak value does not meet the threshold, the process will switch to other search strategies (as described in Example 4).

[0059] Regarding frequency offset calculation: The FFT frequency index of the candidate peak directly corresponds to the frequency offset magnitude. To obtain a more accurate frequency offset estimate, the Rife algorithm or other interpolation estimation algorithms can be used to calculate the frequency offset based on the detected spectral peak and its adjacent frequency points, thereby obtaining a frequency offset estimate with higher frequency resolution than that of the FFT.

[0060] By comparing the peak value with a threshold, it can be preliminarily determined whether the detected peak is a real signal or random noise. Once confirmed, the key synchronization parameters required for communication demodulation—time and frequency—can be extracted from the peak information.

[0061] Example 2 provides another implementation method: a real signal peak verification method based on peak time-series phase fingerprinting, which is a method to enhance the reliability of peak identification, especially suitable for extremely low signal-to-noise ratio scenarios. In step S105 of the previous example, this method is inserted before the final output after a candidate peak that meets a preset threshold is retrieved.

[0062] Step S201: For the candidate peak information, combine it with the accumulated signal and apply a peak time-series phase fingerprint verification to obtain the real signal peak information. Based on the real signal peak information, calculate and output the signal arrival position and frequency offset.

[0063] Specifically, at low signal-to-noise ratios, the amplitude of some noise spikes may occasionally exceed the threshold, leading to false alarms (misjudgments). The peaks of a real signal not only have high amplitudes, but their phase changes over time (caused by carrier frequency offset and timing drift) should also be continuous and smooth. In contrast, the phase of noise peaks is random and irregular in time. This step utilizes this fundamental difference as a more robust discrimination mechanism than a single amplitude threshold.

[0064] Step S202: Decompose the accumulated signal into multiple sliding windows with time overlap, and perform frequency domain transformation on the data in each sliding window to construct a window spectrum sequence.

[0065] The specific steps are as follows:

[0066] Step 1: Obtain the accumulated signal Sa from step S103.

[0067] Step 2: Set the sliding window parameters. For example, set the window length to 3072 points and the overlap rate to 50%.

[0068] Step 3: Generate the sliding window data sequence. If the total length of Sa is 6144, then 3 sliding windows can be generated: W1 is Sa (0-3071), W2 is Sa (1536-4607), and W3 is Sa (3072-6143).

[0069] Step 4: Perform FFT on each window of data W1, W2, W3 to obtain the window spectrum sequence Fw1, Fw2, Fw3.

[0070] Step S203: In the window spectrum sequence, for each candidate peak in the candidate peak information, extract the phase value of its corresponding frequency position in each window spectrum to form a phase evolution sequence that characterizes the phase change of the candidate peak over time, and extract the real signal peak information based on the phase evolution sequence.

[0071] Specifically, suppose that in the previous embodiment, n candidate peaks were detected, forming a set P = {p1, p2, ..., pn}. For any candidate peak pi (whose frequency position is fi):

[0072] Extract the complex value at frequency fi in Fw1 and calculate its phase to obtain φi(1).

[0073] Extract the complex value at frequency fi in Fw2 and calculate its phase to obtain φi(2).

[0074] Extract the complex value at frequency fi in Fw3 and calculate its phase to obtain φi(3).

[0075] Therefore, the phase evolution sequence of the candidate peak pi is obtained as {φi(1),φi(2),φi(3)}. The phase sequences of all peaks constitute a phase sequence matrix Φ.

[0076] Step S204: Extract the peak information of the real signal based on the phase evolution sequence.

[0077] Specifically, a mathematical model is performed on the phase evolution sequence corresponding to each candidate peak, and a set of model parameters characterizing the dynamic characteristics of the sequence is obtained by fitting a quadratic polynomial model using the least squares method.

[0078] Based on the model parameters, the phase curvature or phase continuity score of the phase evolution sequence is quantized, and the corresponding candidate peak is determined to be the real signal peak based on the quantization result, and the peak information is extracted.

[0079] The decision is based on an iterative robust estimation process and further includes bidirectional consistency verification. This is a multi-stage refinement process, with the following steps:

[0080] Step 1: Initial Modeling and Fitting: For the phase evolution sequence {φi(k)} of each peak, establish a time-phase model: φ(t) = a0 + a1*t + a2*t² / 2. Use the least squares method for fitting to solve for the model parameters [a0i, a1i, a2i] corresponding to each peak, forming the model parameter matrix A. Simultaneously calculate the phase curvature κi = |a2i|.

[0081] Step 2: Iterative robust estimation:

[0082] Loop 1: Using the current model parameters A, predict the phase of each window and calculate the prediction error e. Identify outlier phase points whose prediction errors exceed a preset statistical threshold (e.g., 2 standard deviations). Temporarily remove these outliers from the phase evolution sequence and refit the remaining phase points with a quadratic polynomial model to obtain an updated set of model parameters A.

[0083] Loop 2-3: Repeat the above process twice, for a total of three iterations. The final result is a more robust optimized model parameter matrix A_opt after outlier removal.

[0084] Bidirectional consistency check:

[0085] Forward prediction: Using A_opt, the phase φ_forward of the third window is predicted in a forward manner based on the phase value of the first window.

[0086] Backward prediction: Using A_opt, the phase φ_backward of the first window is derived from the actual phase value of the third window.

[0087] Calculate the consistency metric Ci = |φ_forward(i) - φ_backward(i)| / (2π). The smaller this value, the more the phase evolution trajectory conforms to the physical model, and the better the consistency.

[0088] Final decision: Set thresholds, such as consistency metric C < 0.1 and phase curvature κ < κ_threshold. Only candidate peaks that simultaneously meet both conditions are ultimately confirmed as the true signal peak p_true.

[0089] The following specific and reproducible example demonstrates the working process of this invention in a low signal-to-noise ratio environment.

[0090] Scene setting:

[0091] Communication system: Direct sequence spread spectrum system.

[0092] Signal parameters: chip rate Rc = 8 Mcps, physical frame preamble length is 3072 chips.

[0093] Channel conditions: The signal has a frequency offset of +61.23kHz and a signal-to-noise ratio (SNR) of -12.5dB, which is a typical low SNR harsh environment.

[0094] The capture process includes the following steps:

[0095] Step 1: Signal Input and Preliminary Processing (Example 1): The receiver performs matched filtering and decimation-by-four on the signal sampled by the ADC. Since SNR=-12.5dB, if PMF-FFT processing is performed directly on a single frame signal, the peak energy of the signal is about 100-120, which is submerged in noise, making it impossible to set an effective threshold for detection.

[0096] Step 2: Structured Search: The system initiates a structured search strategy.

[0097] First attempt (two sets of direct accumulation): The system performs direct vector accumulation Sa = Sd_1 + Sd_2 on two consecutive signal frame segments Sd_1 and Sd_2. PMF-FFT is then performed on Sa. The peak energy is increased to approximately 160, and the signal-to-noise ratio is somewhat improved, but still not significantly. It is assumed that this peak value does not reach the preset threshold.

[0098] The second attempt (two sets of phase-shifted accumulation): The system performs a π-phase shift on the second signal Sd_2, and then accumulates it using Sa = Sd_1 + phase_shift(Sd_2,π). The peak energy actually drops to between 100-120, resulting in even worse performance. This is because the +61.23kHz frequency offset between the two frames does not create an actual phase difference close to π.

[0099] Third attempt (4 groups of accumulation and phase shifting): Due to the failure of the first two attempts, the system increases the number of accumulation signal groups to 4. Different phase combinations are then used for phase shifting accumulation. When using the phase shifting combination [0, π / 2, π, 3π / 2], Sa = Sd_1 + shift(Sd_2, π / 2) + shift(Sd_3, π) + shift(Sd_4, 3π / 2).

[0100] Step 3: Peak identification and verification, specifically: from Figure 9 In the spectral data, a sharp peak with an energy value as high as 350 can be clearly seen, far exceeding other noise levels. This peak passed the initial amplitude threshold detection.

[0101] More preferably, the system initiates peak time-series phase fingerprint verification. A sliding window phase extraction is performed on the peak to obtain the phase evolution sequence. Since this is a real signal peak, its phase evolution sequence will exhibit good continuity and smoothness. After iterative robust estimation and bidirectional consistency verification, its consistency metric C will be much less than 0.1, and the phase curvature κ will also be at a very low level. Meanwhile, some other noise points with slightly higher amplitudes on the spectrogram will exhibit random phase sequences in this verification step, failing to pass model fitting and consistency verification, and thus will be eliminated.

[0102] The precise re-search and output process further includes the following steps: Through the aforementioned steps, the system confirms the existence of the true signal peak and calculates a frequency offset of approximately +61.23kHz from its frequency location, obtaining a coarse signal arrival location. The system then activates the precise re-search module. The estimated frequency offset is used to calibrate the original, undecimated matched-filtered signal Sm. Based on this, a fine correlation is performed within a small window near the coarse location (e.g., taking 3 samples forward and 3 samples backward), ultimately outputting a signal arrival location accurate to a single sample point.

[0103] This specific example demonstrates that the present invention, through the coordinated use of multiple methods such as multi-frame accumulation, structured phase-shift search, and advanced peak verification, can achieve good signal acquisition even under extremely harsh conditions with an SNR of -12.5dB.

[0104] Example 3 proposes a phase shifting method with predetermined accuracy based on adaptive lookup table and error diffusion. This example is a preferred implementation of the selective phase shifting and vector accumulation steps in Example 1, and provides a technical solution to solve the quantization error and accumulated energy leakage problems caused by traditional phase shifting operations (such as simple I / Q switching).

[0105] The method includes the following steps:

[0106] Step S301: Initialize the adaptive resolution lookup table.

[0107] A non-uniform resolution lookup table (LUT) is a data structure used to store the mapping relationship between phase angles and complex phase shift coefficients. Its characteristic is that it has different angular resolutions (step sizes) in different phase intervals.

[0108] Phase shift sensitivity refers to the magnitude of the quantization error caused by a specific phase angle when using simplified phase shift operations (such as I / Q switching). Typically, the sensitivity is highest around odd multiples of 45°.

[0109] This step is performed during system initialization and includes the following steps:

[0110] Step 1: Establish phase shift sensitivity spectrum:

[0111] The phase space from 0 to 2π is divided into 16 or more basic intervals. Through simulation or theoretical calculation, the errors arising from coarse phase shifting (such as an approximate π / 2 phase shift using I / Q swap) within each interval are analyzed. Regions with larger errors, such as 45°±5° and 135°±5°, are marked as high-sensitivity regions, while the rest are marked as low-sensitivity regions.

[0112] Step 2: Construct a Non-Uniform Lookup Table (LUT):

[0113] Based on the sensitivity map, a lookup table is generated at a non-uniform resolution. For highly sensitive regions, precise phase-shifting coefficient values ​​are generated at a higher resolution (e.g., a step size of 1°). For low-sensitive regions, phase-shifting coefficient values ​​are generated at a lower resolution (e.g., a step size of 5°). Each entry contains an angle θ and its corresponding precise complex phase-shifting coefficient c = [cos(θ), sin(θ)].

[0114] Step 3: Initialize error propagation parameters:

[0115] Set up a cumulative error vector E to store historical phase shift errors. Its length is equal to the maximum number of accumulations, and all its elements are initialized to zero. Also, set an error diffusion coefficient α, for example, α = 0.3.

[0116] This initialization step enables the highest possible phase shift accuracy with limited storage resources. By performing sensitivity analysis on the phase space and concentrating storage and computational resources in the most error-prone, highly sensitive regions, optimal resource allocation is achieved.

[0117] Step S302: Based on the pre-stored target phase shift angle, search or interpolate the pre-constructed non-uniform resolution lookup table with different angular resolutions in different phase intervals to calculate the complex phase shift coefficient with a predetermined accuracy; use the complex phase shift coefficient to perform complex multiplication on the signal frame segment to obtain the precisely phase-shifted signal frame segment, and then perform subsequent vector accumulation.

[0118] It further includes a phase shift error diffusion mechanism, which calculates a complex phase shift coefficient based on a target phase shift angle, and performs calculations using this coefficient.

[0119] This step is invoked when a phase-shifting operation needs to be performed on a segment of a signal frame.

[0120] Read the historical phase shift error stored from the previous phase shift operation, and compensate the current target phase shift angle based on this historical phase shift error to obtain the compensated phase shift angle:

[0121] Assuming the current iteration is the nth accumulation, and the required target phase shift angle is θ_target, the system reads the previous historical error e_prev = E(n-1) from the cumulative error vector E. Further, it calculates the compensated actual phase shift angle θ_comp = θ_target - α * e_prev.

[0122] Based on the compensated phase shift angle, the complex phase shift coefficients with a predetermined accuracy are determined in a non-uniform resolution lookup table. Specifically, in the non-uniform resolution lookup table, two adjacent entries on both sides of the target phase shift angle are located; a set of interpolation weights that satisfy the energy conservation constraint is calculated. This constraint ensures that the modulus of the final complex phase shift coefficient generated after weighting the complex phase shift coefficients in the two adjacent entries is 1; the complex phase shift coefficients of the two adjacent entries are weighted and summed using this set of interpolation weights to obtain the complex phase shift coefficients with a predetermined accuracy.

[0123] In the lookup table (LUT), the two nearest entries on both sides of θ_comp are found by binary search or other efficient search algorithms, denoted as [θ1,c1] and [θ2,c2].

[0124] Calculate a set of interpolation weights w1 and w2 that satisfy the energy conservation constraint (i.e., the modulus of the interpolated coefficients is 1). This can be achieved by solving the system of equations w1 + w2 = 1 and |w1*c1 + w2*c2|² = 1. To simplify the calculation, linear interpolation weights can also be used and subsequently normalized.

[0125] The interpolation weights are used to perform a weighted summation, resulting in the interpolated phase shift coefficient c_interp = w1*c1 + w2*c2. This coefficient c_interp is the final complex phase shift coefficient c_final with the predetermined precision.

[0126] Complex phase-shifting coefficients are used to perform complex multiplication on signal frame segments to obtain precisely phase-shifted signal frame segments, which are then followed by vector accumulation.

[0127] The input signal frame segment Sd is multiplied by the calculated c_final to obtain the phase-shifted signal Sd_shifted = Sd * c_final.

[0128] After performing phase shifting operations on a signal frame segment using this complex phase shift coefficient, the difference between the actual phase shift angle it represents and the target phase shift angle is calculated. This generates and stores the current phase shift error as the historical phase shift error for the next phase shift operation.

[0129] Calculate the actual phase shift angle θ_real = atan2(imag(c_final), real(c_final)). Based on this, calculate the phase shift error e_current = θ_real - θ_target for this operation. Further, store this error in the cumulative error vector E, i.e., E(n) = e_current, for use in the next phase shift.

[0130] Specifically, this step uses a closed-loop mechanism of prediction-compensation-correction to dynamically compensate for the minute errors introduced by table lookup and interpolation in each operation and transfer them to the next operation. This error diffusion mechanism can effectively prevent the one-way accumulation of errors, ensuring that the total phase deviation remains at an extremely low level after multiple accumulations, thereby guaranteeing the efficiency of vector accumulation and the final signal-to-noise ratio gain.

[0131] Step S303: The non-uniform resolution lookup table is dynamically updated, and its update method includes: periodically statistically analyzing the frequency distribution of the target phase shift angle used in past phase shift operations to identify the high-frequency used angle range;

[0132] Adaptively improve the angular resolution in the high-frequency operating angle range and update the non-uniform resolution lookup table to provide more suitable phase shift accuracy for commonly used angles in subsequent phase shift coefficient calculations.

[0133] Specifically, it includes the following steps:

[0134] Step 1: Count usage frequency: The system maintains an angle usage counter U. Each time a phase shift operation is performed, the counter for the corresponding angle interval is incremented by one.

[0135] Step 2: Dynamically adjust resolution: Every preset period (e.g., after 100 phase shifts), the system analyzes the counter U to identify the angle range that is used frequently (e.g., the range where the number of uses exceeds the average plus one standard deviation).

[0136] Step 3: Update the lookup table: For these frequently used intervals, if their current step size is large, the system will recalculate and update the lookup table entries for that interval at a higher resolution (e.g., from a 5° step size to a 1° step size). To prevent the lookup table from growing indefinitely, a total size limit can be set, and some entries for the least frequently used intervals can be removed when adding new entries.

[0137] Specifically, the dynamic update mechanism enables the lookup table to have adaptive learning capabilities. In actual operation, due to the existence of specific frequency offsets, the actual required phase shift angles may be concentrated in certain specific regions. This mechanism can dynamically reallocate limited storage resources to these hot spots, thereby further improving the phase shift accuracy of commonly used angles without increasing the overall cost.

[0138] Example 4 proposes a structured multi-stage phase-shifting accumulation search strategy. This example provides a systematic search strategy for selective phase-shifting and vector accumulation steps, aiming to address the phase difference problem with uncertain amplitude caused by unknown frequency offset.

[0139] Step S401: Perform direct vector accumulation on the two sets of signal frame segments to obtain the first accumulated signal, and then perform subsequent frequency domain transformation and analysis and peak identification on the first accumulated signal;

[0140] If the candidate peak that meets the preset threshold is not identified based on the first accumulated signal, a π-phase shift is applied to one set of signal frame segments, and vector accumulation is performed with another set of signal frame segments to obtain a second accumulated signal. Frequency domain transformation and analysis and peak identification are then performed on the second accumulated signal.

[0141] This strategy is the starting point of the capture process and includes the following steps:

[0142] Step 1: First Search (Direct Accumulation): Take the first set of signal frame segments Sd(0)...Sd(L-1) and the second set Sd(L)...Sd(2L-1) and perform direct vector accumulation to obtain the first accumulated signal Sa_1. Perform subsequent PMF-FFT and peak identification on Sa_1. If a peak that meets the threshold is successfully identified, the capture is successful and the process ends.

[0143] Step Two: Second Search (π-phase shift accumulation): If the first search fails, the phase shifter module is activated to apply a π-phase phase shift to the entire second group of signal frame segments Sd(L)...Sd(2L-1). Further, vector accumulation is performed with the first group of signals to obtain the second accumulated signal Sa_2. Then, PMF-FFT and peak identification are performed on Sa_2.

[0144] Specifically, the phase difference Δφ between the two frames is unknown. If Δφ is close to 0, direct accumulation yields the best results. If Δφ is close to π, direct accumulation will cause signal cancellation, while π-phase-shifted accumulation can achieve in-phase superposition. These two searches cover the two extreme cases of phase difference and can effectively handle most scenarios involving the accumulation of two frames.

[0145] Step S402: If the structured search strategy for the two sets of signal frame segments fails to identify candidate peaks that meet the preset threshold in both attempts, the number of accumulated signal frame segments will be increased to four sets.

[0146] The four signal frame segments are then sequentially phase-shifted and vector-accumulated using different preset phase combinations. The accumulated signal generated by each combination is used to perform frequency domain transformation and analysis as well as peak identification, until the candidate peak is successfully identified.

[0147] If both searches in step S401 fail, the system determines that a higher signal-to-noise ratio gain is needed and automatically increases the number of accumulated frames to four groups (Sd_1, Sd_2, Sd_3, Sd_4). Furthermore, the phase shifter will attempt various different phase combinations for phase-shifting accumulation in a preset order.

[0148] Optionally, the preset phase combination may include:

[0149] Combination a: Phase shift [π / 2, π, 3π / 2] for the second, third, and fourth groups of signals, respectively.

[0150] Combination b: Phase shift [π, 0, π] for the second, third, and fourth groups of signals respectively.

[0151] Combination c: Phase shift [3π / 2, π, π / 2] for the second, third, and fourth groups of signals, respectively.

[0152] For each combination attempted, an accumulated signal Sa is generated and a PMF-FFT and peak identification are performed. Once identification is successful, the search stops.

[0153] Specifically, the sum of the four signal sets can achieve a theoretical gain of 4.17dB to 6dB, which can cope with lower signal-to-noise ratios. By trying multiple preset phase combinations, it is equivalent to trying to align the signal vectors in a four-dimensional vector space using different rotation matrices. This further increases the probability of finding a solution that allows the four signal sets to be superimposed in phase, thus systematically solving the problem of unknown frequency offset.

[0154] Step S403: Phase shift the four sets of signal frame segments. When the phase shift angle in the preset phase combination is π / 2, π, or 3π / 2, the phase shift operation specifically involves swapping or inverting the in-phase and quadrature components of the signal frame segment to achieve the corresponding phase rotation without consuming multiplier computational resources. When the phase shift angle is π / 2, the swapping or inverting operation specifically involves inverting the quadrature component of the original signal frame segment and using it as the in-phase component of the phase-shifted signal frame segment; and directly using the in-phase component of the original signal frame segment as the quadrature component of the phase-shifted signal frame segment.

[0155] This is a preferred implementation for efficiently performing the specific angle phase shift in step S402. For a complex signal Sd = I + jQ:

[0156] Phase shift π / 2(j): Sd_shifted=(-Q)+j(I);

[0157] Phase shift π(-1): Sd_shifted=(-I)+j(-Q);

[0158] Phase shift 3π / 2(-j): Sd_shifted=(Q)+j(-I).

[0159] Specifically, the core advantage of this step lies in its lack of computational resource consumption. In hardware implementations such as FPGAs, complex multipliers are valuable logic resources. Phase rotations that are multiples of π / 2 can be achieved through simple interconnect swapping (I / Q swapping) and an inverter (NOT gate), avoiding the use of multipliers and further reducing hardware cost, power consumption, and processing latency.

[0160] Example 5 provides a method for searching the precise signal arrival location based on prior information.

[0161] Specifically, this embodiment describes a final refinement step performed after successfully capturing a signal and obtaining a rough signal arrival location and frequency offset using the method of any of the foregoing embodiments.

[0162] Step S501: The calculated signal arrival position is used as the prior signal arrival position (coarse signal arrival position), and the frequency offset is used to calibrate the matched-filtered signal. In the frequency offset-calibrated matched-filtered signal, a search window (small-range signal sequence) of predetermined width is defined with the prior signal arrival position as the center, and the signal within the search window is extracted as the signal sequence to be refined (small-range signal sequence). The signal sequence to be refined is correlated again, and the position of the maximum correlation peak is taken as the final output precise signal arrival position. The specific steps are as follows:

[0163] Step 1: Frequency Offset Calibration. Obtain the coarse frequency offset Δf and coarse position Pos_coarse from the acquisition stage. Take the un-decimated raw matched filter signal Sm. Perform frequency offset compensation on Sm to obtain the calibrated signal Sw(t) = Sm(t) * exp(-j * 2 * π * Δf * t).

[0164] Step 2: Define the search window and the extracted sequences. Define a search window of a predetermined width centered on Pos_coarse. Considering the maximum uncertainty brought about by four times decimation, the window width can be set to include three additional sampling points forward and three more backward. This extracts seven signal sequences from Sw, each with a length equal to the length of the preceding sequence and overlapping starting positions, to be refined for further searching.

[0165] Step 3: Re-correlation and precise localization. Perform a complete complex cross-correlation operation on each of the 7 sets of sequences with the local leading sequence. Compare the peak values ​​of the 7 correlation results; the starting position corresponding to the peak with the largest amplitude is determined as the final output signal arrival position, accurate to a single sampling point.

[0166] Specifically, this step eliminates the positional uncertainty introduced by step S102 (signal decimation) in Example 1. Frequency offset calibration is performed to maximize signal energy concentration during re-correlation. Furthermore, performing a small-range, predetermined-precision search around the coarse position on the undecimated raw signal can improve the signal position accuracy from ±2 sampling points to ±0 sampling points with minimal additional computation, providing a better initial value for subsequent demodulator locking.

[0167] In another preferred embodiment of the invention, the acquisition method can be designed and implemented based on the Xilinx xc7z100ffg900 chip as a hardware platform. This method is applicable to continuous or quasi-continuous channels.

[0168] A continuous channel refers to a channel mode in which signal frames are transmitted continuously without any data idle periods. The format of its transmitted frames is shown in Table 1, with the preamble sequence and frame content appearing alternately and continuously.

[0169] Table 1 Frames transmitted via continuous channel

[0170] Prelude 1 Frame content 1 Prelude 2 Frame content 2 Prelude 3 Frame content 3 Prelude 4 Frame content 4 …… …… …… …… Preceding N Frame content N

[0171] A quasi-continuous channel refers to a channel where the transmission of signal frames is irregular in time, with random-length idle periods between frames. The format of its transmitted frames is shown in Table 2.

[0172] Table 2 Quasi-continuous channel transmission frames

[0173] Prelude 1 Frame content 1 Prelude 2 Frame content 2 No data sent No data sent Prelude 3 Frame content 3 No data sent Prelude 4 Frame content 4 No data sent …… …… Preceding N Frame content N

[0174] This invention utilizes the characteristic that signal parameters (such as frequency offset, phase offset, and bit timing offset) in these two channels change continuously and slowly, and improves acquisition performance through multi-frame processing.

[0175] According to another aspect of this application, some steps in Embodiment 1 can also be (the same steps as in Embodiment 1, which will not be described in detail):

[0176] Step S103: Perform selective phase shifting and vector accumulation on at least two sets of signal frame segments in the extracted signal to fuse them into an accumulated signal.

[0177] Specifically, since the preamble sequence of each frame in a spread spectrum communication system is fixed and identical in continuous or quasi-continuous channel mode, by coherently accumulating corresponding parts of multiple frames of signals, the signal components will be superimposed in phase, while the noise components will be superimposed incoherently. This can effectively improve the signal-to-noise ratio of the signal.

[0178] Without considering the effects of bit timing offset, the ideal summation of the two sets of signals can achieve a signal-to-noise ratio gain of 3dB. Even considering the effects of frequency offset, the phase-shifting processing of this invention can still achieve a gain of 1.51dB to 3dB. This is sufficient to compensate for the loss caused by decimation in step S102 and to obtain additional processing gain.

[0179] Step S105: Search the spectrum data and identify candidate peaks that meet the preset threshold, and extract their candidate peak information; calculate and output the signal arrival position and frequency offset based on the candidate peak information.

[0180] According to one aspect of this application, the effect of multi-frame accumulation may also be affected by timing drift at both the transmitting and receiving ends. However, in the application scenarios of this invention, such as communication between a geostationary satellite and a ground station, the relative speed is low, and the resulting Doppler effect is not significant; at the same time, the crystal oscillator technology of modern terminal equipment is becoming increasingly mature. Based on this analysis, the normalized bit timing deviation caused by the Doppler effect and crystal oscillator technology in the system typically does not exceed 1×10⁻⁶. -7 Based on this deviation level, the bit timing deviation between the two signal vector groups does not exceed 1%, while the bit timing deviation between the four signal vector groups does not exceed 2.3%. Therefore, in such applications, the effect of the signal-to-noise ratio enhancement caused by the bit timing deviation is limited and almost negligible, which provides feasibility support for the multi-frame accumulation technique of this invention.

[0181] Example 6 provides a system environment and parameter configuration example to illustrate the specific parameters and environment of the system. In this example, the main technical specifications of the designed demodulator are set as follows:

[0182] Encoding and modulation method: 1 / 2 bit rate binary phase shift keying (1 / 2 BPSK).

[0183] Spreading factor: can be configured to 1, 2, 4, 8, or 16.

[0184] Chip rate range: Supports 1 Mcps to 16 Mcps.

[0185] Demodulation threshold: set to a bit energy to noise power spectral density ratio (Eb / n0) of not less than 2.5dB.

[0186] Specifically, these parameters collectively define a typical high spreading ratio and low signal-to-noise ratio (SNR) operating scenario. For example, when the spreading factor is selected as 16, the relationship between the ratio of symbol energy to noise power spectral density (Es / n0) and the demodulation threshold (Eb / n0) before despreading is Es / n0(dB) = Eb / n0(dB) - 10 * log10 (spreading factor). Substituting the values, we get Es / n0 = 2.5 - 10 * log10 (16) ≈ 2.5 - 12.04 = -9.54 dB. Considering the influence of the code rate, the SNR of the signal will be even lower when it enters the acquisition module. This indicates that the target operating environment of the acquisition device designed in this invention is precisely in such a harsh channel where the signal energy is far lower than the noise level. This invention combines a series of technical means, such as multi-frame accumulation, predetermined precision phase shifting, efficient frequency domain search, and high-reliability peak verification, as described in the foregoing embodiments, to reliably capture signals with Es / n0 as low as approximately -12.5dB under the system environment defined by the above indicators, thereby meeting the overall design requirements of the system.

[0187] Direct sequence spread spectrum technology in satellite communication has advantages such as strong resistance to fading, multipath interference, interception, and low transmit power density. It is suitable for small-aperture antennas, can reduce interference from neighboring satellites, and is mainly used for IoT or other low data rate transmission applications.

[0188] This application designs a demodulator, the main specifications of which are as follows:

[0189] Encoding and modulation: 1 / 2 BPSK;

[0190] Spreading factor: 1, 2, 4, 8, 16;

[0191] Chip rate: 1 Mcps-16 Mcps;

[0192] Demodulation threshold: Eb / n0 = 2.5 dB;

[0193] In direct sequence spread spectrum (DSS), 1 bit of information is spread into multiple bits using a spreading sequence. Furthermore, the symbol is modulated and transmitted without changing the system's transmit power, resulting in a lower frequency density of the spread signal. According to the demodulator specifications in this application, when the spreading factor is 16, the spread signal's Es / n0 is considered to be -12.5 dB. This invention provides a high-efficiency acquisition method for low signal-to-noise ratio (SNR) direct sequence spread spectrum channels, specifically:

[0194] Step S601: Start the matched filter module to perform matched filtering on the signal after analog-to-digital conversion (ADC). The signal after matched filtering is denoted as Sm.

[0195] Step S602: Start the signal extraction module, extract Sm by four times, and denote the extracted signal as Sd. Store the Sd signal of length L into the local RAM (where the channel physical frame length is denoteed as L, the ADC is 4 times sampling, so the number of sampling points per physical frame is 4L), and denote it as Sd(0), Sd(1)...Sd(L-1).

[0196] Step S603: Start the accumulator module, then take Sd(L)……Sd(2L-1) from the Sd sequence and add them to Sd(0), Sd(1)……Sd(L-1) in RAM to obtain Sa(0)……Sa(L-1);

[0197] Step S604: Start the PMF_FFT module, and perform FFT operation after matching the Sa(0)……Sa(L-1) parts;

[0198] Step S605: Start the threshold detection module to perform peak retrieval and threshold judgment on the spectrum of Sa(0)……Sa(L-1). If the threshold does not meet the condition, proceed to S606; otherwise, proceed to S610.

[0199] Step S606: If the current search is the first search with 2 groups of cumulative results, proceed to step S607; if the current search is the second search with 2 groups of cumulative results, proceed to step S608; if the current search is the first, second, and third search with 4 groups of cumulative results, proceed to step S609; otherwise, proceed to step S603 and reinitialize the parameters.

[0200] Step S607: Start the phase shifter module, perform π phase shift on the Sd(L)……Sd(2L-1) sequence, and add it with Sd(0), Sd(1)……Sd(L-1) to obtain Sa(0)……Sa(L-1), and then go to S604;

[0201] Step S608: Start the accumulator module, continuously take Sd(0), Sd(1)...Sd(L-1), Sd(L), Sd(L+1)...Sd(2L-1), Sd(2L), Sd(2L+1)...Sd(3L-1), Sd(3L), Sd(3L+1)...Sd(4L-1) from the Sd sequence, add them together to obtain Sa(0)...Sa(L-1), and then go to S604;

[0202] Step S609: Start the phase shifter module, perform phase shifting and accumulation on the Sd signal to obtain Sa(0)……Sa(L-1), and then proceed to step S604;

[0203] Step S610: Start the frequency offset detection module, calculate the frequency offset, and output the location and frequency offset of the signal arrival;

[0204] Step S611: Start the frequency offset calibration module to calibrate the signal frequency offset;

[0205] Step S612: Start the signal re-search module, use the prior information of the signal in step S610 to re-estimate the signal within a small range, and output the accurate signal arrival position.

[0206] In a further embodiment, the received signal chip rate is set to Rb = 8 Mcps, the frequency offset to +61.23 kHz, and the SNR to -12.5 dB.

[0207] The signal is decimated after matched filtering, without accumulation, and then subjected to a PMF_FFT operation. Its frequency domain characteristics are as follows: Figure 6 As shown.

[0208] The signal is accumulated by a factor of 2 according to the steps in the device, and then PMF_FFT operation is performed.

[0209] The signal is accumulated by a phase shift of π at twice the amount according to the steps in the device, and then PMF_FFT operation is performed.

[0210] According to one aspect of this application, the present invention successfully solves the technical problem of balancing computational complexity and search performance in traditional acquisition methods by performing a fourfold decimation of the signal after matched filtering and decomposing the subsequent correlation operation into a structure combining partially matched filtering and Fast Fourier Transform (PMF_FFT). Specifically, the fourfold decimation directly reduces the amount of data processed in subsequent operations by 75%, saving hardware buffering and power consumption. The PMF_FFT structure transforms a highly complex full-length time-domain correlation operation into a computationally fixed FFT operation, whose frequency point position directly corresponds to the frequency offset information of the signal. This time-frequency conversion approach allows the receiver to complete the search of a large frequency offset range in a single operation (e.g., covering a frequency offset range of no less than ±333.33 kHz at a specific code rate) without using multiple parallel hardware channels. Therefore, this method can achieve efficient and wide-range acquisition of high chip rate signals on resource-constrained hardware platforms (such as Xilinx FPGA chips), improving the practicality and economy of the system.

[0211] According to another aspect of this application, the invention effectively solves the core technical problem of uncertain gain in multi-frame signal accumulation under unknown frequency offset by designing a structured search strategy that ranges from two-group accumulation to four-group accumulation and from simple phase shifting to multi-phase combination. The essence of this strategy is an automated, hierarchical solution search process. Specifically, through two low-cost attempts—direct accumulation and π-phase shift accumulation—it quickly covers the two most extreme cases where the inter-frame phase difference caused by frequency offset is close to 0 or π. If this fails, it indicates extremely low signal-to-noise ratio and complex phase relationships. The system automatically upgrades to four-group accumulation to obtain a theoretical gain of up to 6dB, and systematically searches for the optimal solution that aligns the signal vectors in a four-dimensional vector space through attempts with multiple preset phase combinations (such as [π / 2, π, 3π / 2], etc.). This intelligent search mechanism eliminates reliance on prior knowledge of channel frequency offset during the acquisition process, improving the system's adaptability and robustness in real, variable satellite communication scenarios and increasing the success rate of link establishment under adverse channel conditions.

[0212] Specifically, this invention solves the signal energy leakage problem caused by inaccurate phase-shift quantization during multi-frame accumulation by introducing a predetermined-precision phase-shifting method based on a non-uniform lookup table and a phase-shift error diffusion mechanism. When accumulating signals for up to four or more frames, the tiny quantization errors introduced by each phase-shifting operation accumulate, potentially leading to a synthesized vector magnitude much smaller than the theoretical value, thus weakening the signal-to-noise ratio (SNR) gain from accumulation. The error diffusion mechanism in this invention reads the historical phase-shift error e_prev stored from the previous operation and pre-compensates the current target phase-shift angle θ_comp = θ_target - α * e_prev, which is equivalent to a dynamic correction process. This makes the long-term accumulated average phase error approach zero. This allows the signal vector to be aligned with high precision during multi-frame accumulation, thus approaching the theoretical SNR gain to a certain extent (e.g., a gain of nearly 6dB when 4 sets of accumulations are obtained). For systems that need to operate at extremely low signal-to-noise ratios (such as -12.5dB), this extra 1-2dB gain is crucial, directly determining the success or failure of acquisition, and can translate into an extension of communication distance or a reduction in the size requirements of the terminal antenna.

[0213] Building upon this foundation, this invention introduces a peak-time phase fingerprint verification technique, enhancing the reliability of acquisition decisions under extremely low signal-to-noise ratios and resolving the inherent high false alarm rate problem of traditional amplitude threshold-based decision methods. Instead of viewing a peak in the spectrum in isolation, it considers it as a snapshot of a signal's temporal evolution. By extracting the phase of this peak within multiple overlapping time windows, a phase evolution sequence Φ is formed and modeled using a quadratic polynomial: φ(t) = a0 + a1*t + a2*t² / 2. The phase evolution of a true signal should be smooth and deterministic (caused by a stable frequency offset), while the phase evolution of a noise spike is random and discontinuous. This invention further employs iterative robust estimation and bidirectional consistency verification to strengthen this criterion, ensuring that only peaks whose phase evolution trajectories fully conform to the physical model are confirmed as true signals. This method of dual verification of signals from both energy and behavior dimensions is equivalent to adding a pre-defined authenticity detector to the acquisition system. It can effectively filter out the interference of noise spurious peaks, ensuring that the subsequent demodulation loop is locked onto the real signal. This is of core value for ensuring the stable operation of unattended terminals or mission-critical communication links.

[0214] Furthermore, this invention addresses the position accuracy loss caused by initial signal decimation by adding a precise re-search step after coarse acquisition, achieving a better balance between acquisition efficiency and final accuracy. While the initial fourfold decimation improves search efficiency, it also introduces timing ambiguity of approximately ±2 sampling points. Directly passing this ambiguous position information to subsequent tracking loops would lead to longer convergence times and potentially lock failure. The re-search step in this invention uses the estimated frequency offset to calibrate the undecimated original matched-filtered signal Sm, eliminating phase rotation interference. Furthermore, it avoids a global search, performing full-resolution correlation calculations within a very small window (e.g., a range of 7 sampling points) near the coarse position. This coarse-to-fine strategy improves timing accuracy from the multi-sampling-point level to the single-sampling-point level with minimal additional computational cost. This provides a higher-quality initial anchor point for subsequent carrier tracking and symbol synchronization loops, shortens the lock time of the entire link, and improves the overall timeliness of the system from acquisition to data demodulation.

[0215] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A high-efficiency acquisition method for a low signal-to-noise ratio direct sequence spread spectrum channel, characterized in that, include: The analog-to-digital converted signal is subjected to matched filtering and signal extraction to obtain the extracted signal. Selective phase shifting and vector accumulation are performed on at least two sets of signal frame segments in the extracted signal, and they are fused into an accumulated signal. The accumulated signal is transformed and analyzed in the frequency domain to obtain the spectrum data; The frequency domain transformation and analysis includes: performing point-by-point sliding window data extraction on the accumulated signal and performing correlation operations with the local sequence to obtain the correlated signal; The spectrum data is retrieved and candidate peaks that meet the preset threshold are identified, and their candidate peak information is extracted. The signal arrival position and frequency offset are calculated and output based on candidate peak information; including: applying peak timing phase fingerprint verification to candidate peak information in combination with accumulated signal to obtain real signal peak information; and calculating and outputting signal arrival position and frequency offset based on real signal peak information.

2. The method according to claim 1, characterized in that, By combining the accumulated signal and applying peak time-series phase fingerprint verification, the true signal peak information is obtained, including: The accumulated signal is decomposed into multiple sliding windows with time overlap, and the data in each sliding window is transformed in the frequency domain to construct a window spectrum sequence. In the window spectrum sequence, for each candidate peak in the candidate peak information, the phase value of its corresponding frequency position in each window spectrum is extracted to form a phase evolution sequence that characterizes the phase change of the candidate peak over time, and the real signal peak information is extracted accordingly.

3. The method according to claim 2, characterized in that, Extracting peak information of the real signal based on the phase evolution sequence includes: Mathematical modeling is performed on the phase evolution sequence corresponding to each candidate peak. A quadratic polynomial model is fitted by the least squares method to obtain a set of model parameters characterizing the dynamic characteristics of the sequence. Based on the model parameters, the phase curvature or phase continuity score of the phase evolution sequence is quantized, and the corresponding candidate peak is determined to be the real signal peak based on the quantization result, and the peak information is extracted.

4. The method according to claim 1, characterized in that, Perform selective phase shifting and vector accumulation, including: Based on the pre-stored target phase shift angle, the complex phase shift coefficient with a predetermined accuracy is calculated by searching or interpolation in a pre-constructed non-uniform resolution lookup table with different angular resolutions in different phase intervals. Complex phase-shifting coefficients are used to perform complex multiplication on signal frame segments to obtain precisely phase-shifted signal frame segments, which are then subjected to subsequent vector accumulation.

5. The method according to claim 4, characterized in that, It further includes a phase shift error diffusion mechanism, which calculates a complex phase shift coefficient based on a target phase shift angle, and performs calculations using this coefficient, including: Read the historical phase shift error stored from the previous phase shift operation, and compensate the current target phase shift angle based on the historical phase shift error to obtain the compensated phase shift angle; Based on the compensated phase shift angle, the complex phase shift coefficient with a predetermined accuracy is determined in a non-uniform resolution lookup table; After performing phase shifting operations on signal frame segments using complex phase shift coefficients, the difference between the actual phase shift angle and the target phase shift angle is calculated, and the current phase shift error is generated and stored as the historical phase shift error for the next phase shifting operation.

6. The method according to claim 4, characterized in that, Non-uniform resolution lookup tables are dynamically updated, and their update methods include: Periodically analyze the frequency distribution of the target phase shift angles used in past phase shift operations to identify the high-frequency angle ranges. Improve the angular resolution in the high-frequency usage angle range and update the non-uniform resolution lookup table.

7. The method according to claim 1, characterized in that, Selective phase shifting and vector accumulation are performed, and a structured search strategy for the two sets of signal frame segments is also included, specifically: The two sets of signal frame segments are directly vector-accumulated to obtain the first accumulated signal, and then the first accumulated signal is subjected to subsequent frequency domain transformation and analysis as well as peak identification. If the candidate peak that meets the preset threshold is not identified based on the first accumulated signal, a π-phase shift is applied to one set of signal frame segments, and vector accumulation is performed with another set of signal frame segments to obtain a second accumulated signal. Frequency domain transformation and analysis and peak identification are then performed on the second accumulated signal.

8. The method according to claim 7, characterized in that, This structured search strategy further includes: If the structured search strategy fails to identify a candidate peak that meets the preset threshold in both attempts for the two sets of signal frame segments, the number of accumulated signal frame segments will be increased to four. The four signal frame segments are then sequentially phase-shifted and vector-accumulated using different preset phase combinations. The accumulated signal generated by each combination is used to perform frequency domain transformation and analysis as well as peak identification, until the candidate peak is successfully identified.

9. The method according to claim 1, characterized in that, The accumulated signal is transformed and analyzed in the frequency domain to obtain spectral data, which further includes: Based on the chip rate of the pre-stored characterization system symbol rate, the correlated signal is divided into a predetermined number of groups, and the data in each group is summed to generate a grouped summed sequence. The spectral data is obtained by performing a Fast Fourier Transform on the grouped and summed sequence.

Images