Low earth orbit opportunity signal doppler estimation method for multi-beam dynamic switching

Abstract

The application discloses a kind of low orbit opportunity signal Doppler estimation methods for multi-beam dynamic switching, belong to satellite positioning and navigation technical field.The method is first to Iridium time frame signal according to beam polling period is blocked and band-pass filter preprocessing, then utilize segmented FFT to carry out Doppler frequency rough estimation, and the starting position of pilot signal is accurately positioned in combination with the prior independent word information generated locally.Next, the frequency offset factor and signal-to-noise ratio features such as two-dimensional parameters are extracted for the pilot signal accurately stripped, and the pre-trained KNN model is used for dynamic decision, and accordingly Rife or MLE algorithm is used for Doppler frequency fine estimation.The application solves the problem of insufficient accuracy of traditional method in the complex scene of large signal-to-noise ratio fluctuation and large frequency change through adaptive algorithm matching mechanism, significantly improves the robustness of Doppler estimation, and is suitable for non-cooperative signal processing of low-orbit satellite navigation system.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of satellite positioning and navigation, and particularly relates to a low-orbit opportunity signal Doppler estimation method for multi-beam dynamic switching. BACKGROUND

[0002] The global navigation satellite system (GNSS) is the most widely used navigation and positioning technology at present, but its signal power is low, it is easy to be disturbed and blocked, and there is a problem of service performance decline or even failure in complex environments. Low-orbit satellites (such as Iridium systems) become important backup or enhancement means in GNSS denial environments due to their high landing power (about 30 dB stronger than GNSS), wide coverage, and fast geometric configuration changes. However, the low-orbit satellite signal has a significant Doppler frequency shift (up to ±40 kHz), and the signal-to-noise ratio fluctuates dramatically due to beam dynamic switching, which brings challenges to signal processing.

[0003] In the prior art, the Doppler estimation methods for low-orbit non-cooperative signals include phase-time algorithm, QSA-IDE algorithm, and implicit pilot extraction algorithm. These algorithms have the following disadvantages: first, they rely on prior information, but low-orbit opportunity signals often lack stable pilots or frame structures; second, the algorithm complexity is high, and it is difficult to achieve efficient processing in millisecond-level short signal frames; third, the adaptability to signal-to-noise ratio fluctuations and large dynamic Doppler is poor, resulting in unstable estimation accuracy. These shortcomings limit the practical application of low-orbit opportunity signals in navigation and positioning.

[0004] Therefore, there is an urgent need for a robust estimation method that can adapt to large dynamic Doppler and signal-to-noise ratio fluctuations of low-orbit signals, to achieve high-precision and high-efficiency Doppler frequency estimation under the condition of lacking prior information, and to improve the usability of low-orbit opportunity signals in complex environments. SUMMARY

[0005] To solve the above technical problems, the application provides a low-orbit opportunity signal Doppler estimation method for multi-beam dynamic switching. Based on the Iridium time frame pilot signal, the method performs Doppler frequency coarse estimation through global segmented FFT, then synchronizes the frame header using the prior information of the Iridium frame, and identifies and extracts the frequency offset factor and signal-to-noise ratio. The characteristic parameters are input into the pre-trained KNN (K Near Neighbor) two-dimensional characteristic parameter space (frequency offset factor, instantaneous signal-to-noise ratio) dynamic decision model to establish a "parameter space-refinement algorithm" matching mapping mechanism, realize the dynamic coupling of algorithm architecture and signal characteristics, and can adaptively select algorithms according to complex scene characteristics to optimize the Doppler estimation accuracy and robustness.

[0006] To achieve the above object, the technical scheme adopted by the present application is as follows:

[0007] In a first aspect, the present application provides a low-orbit opportunity signal Doppler estimation method for multi-beam dynamic switching, comprising:

[0008] Step 1, block the received low-orbit opportunity signal according to a preset period, and preprocess each block signal to obtain a target time frame signal, wherein the target time frame signal at least includes the pilot and unique word of the low-orbit opportunity signal; wherein the preset period matches the beam polling period of the low-orbit opportunity signal;

[0009] Step 2, segment the target time frame signal, and perform fast Fourier transform (FFT) on each segment signal of the target time frame signal to obtain the frequency domain peak value of each segment signal, determine the segment signal whose frequency domain peak value meets the preset peak value condition as a target segment signal, and determine the coarse estimation value of the Doppler frequency of the pilot based on the frequency domain peak value in the target segment signal; wherein the preset peak value condition includes that the frequency domain peak value of the segment signal is greater than a preset peak value threshold, and after sorting the frequency domain peak values of all segment signals from large to small, the frequency domain peak value of the segment signal is located in the first N positions, and N is a positive integer;

[0010] Step 3, generate a local independent word signal using prior information, determine the peak value of the cross-correlation function between the local independent word signal and a general pilot rough position signal as the pilot starting position, and the general pilot rough position signal is a signal whose distance from the pilot rough position is less than a preset distance threshold; wherein the prior information at least includes prior pilot information and prior unique word information;

[0011] Step 4, determine the pilot signal based on the pilot starting position, perform FFT processing on the pilot signal again to obtain the pilot signal spectrum, and determine the frequency offset factor estimation value and the instantaneous signal-to-noise ratio estimation value based on the pilot signal spectrum; wherein the number of points of the FFT processing performed again is the same as the time domain sampling point number of the pilot signal;

[0012] Step 5, input the frequency offset factor estimation value and the instantaneous signal-to-noise ratio estimation value into a pre-trained KNN decision model to obtain a first algorithm weight and a second algorithm weight; wherein the first algorithm is a frequency estimation Rife algorithm based on the phase difference of discrete time signals, and the second algorithm is a maximum likelihood estimation (MLE) algorithm;

[0013] Step 6, determine the Doppler frequency compensation value of the pilot using the target algorithm, determine the accurate Doppler frequency value of the pilot based on the coarse Doppler frequency estimation value of the pilot and the Doppler frequency compensation value of the pilot; wherein the target algorithm is the first algorithm when the weight of the first algorithm is greater than the weight of the second algorithm; the target algorithm is the second algorithm when the weight of the second algorithm is greater than the weight of the first algorithm.

[0014] The beneficial effects of the present application are:

[0015] High precision and strong robustness: the present application extracts signal characteristic parameters (frequency offset factor and signal-to-noise ratio) and dynamically selects the optimal estimation algorithm (Rife or MLE) using a pre-trained KNN model to realize algorithm advantage complementation. Experiments show that the root mean square error is significantly reduced and the estimation accuracy is improved by more than 30% under the conditions of signal-to-noise ratio fluctuation (-15dB to 5dB) and large dynamic Doppler (±40kHz).

[0016] Adaptive scene matching capability: the present application can automatically select the optimal processing path according to the real-time characteristics of the signal by constructing a "feature parameter-algorithm" mapping mechanism. For example, the Rife algorithm with less calculation is preferred at high signal-to-noise ratio, and the MLE algorithm with stronger noise resistance is switched to at low signal-to-noise ratio, so as to balance accuracy and efficiency.

[0017] Low complexity and high real-time performance: compared with algorithms such as phase-time method that need full-band search, the present application greatly reduces the calculation amount through step-by-step processing of coarse estimation and fine estimation. In addition, Doppler estimation can be completed only by using pilot signals (without demodulating data frames), avoiding complex demodulation processes such as implicit pilot method, and the operation efficiency is improved by more than 50%, which is suitable for scenes with high real-time requirements.

[0018] Wide applicability: the present application does not depend on prior information of specific satellite systems and can be directly applied to Iridium and other low-orbit opportunity signals, and is compatible with non-cooperative signal processing of other similar systems, providing reliable technical support for navigation and positioning in GNSS denial environment. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 is a flowchart of a low-orbit opportunity signal Doppler estimation method for multi-beam dynamic switching provided by the present application.

[0020] Figure 2 is a time frame structure diagram of Iridium signal.

[0021] Figure 3 is a flowchart of a method for determining the coarse Doppler frequency estimation value of the pilot provided by the present application.

[0022] Figure 4is a flowchart of a method for determining a pilot starting position provided by an embodiment of the present application.

[0023] Figure 5 is a schematic diagram of a pilot signal determined by taking the iridium star as an example.

[0024] Figure 6 is a flowchart of a method for training a KNN decision model provided by an embodiment of the present application.

[0025] Figure 7 is a schematic diagram of a label point set in a pre-trained KNN decision model provided by an embodiment of the present application.

[0026] Figure 8 is a flowchart of a method for determining a first algorithm weight and a second algorithm weight using a pre-trained KNN decision model provided by an embodiment of the present application.

[0027] Figure 9 is a flowchart of another low-orbit opportunity signal Doppler estimation method for multi-beam dynamic switching provided by the present application. DETAILED DESCRIPTION

[0028] The present application will be further described below in conjunction with the drawings and embodiments.

[0029] Figure 1 is a flowchart of a low-orbit opportunity signal Doppler estimation method for multi-beam dynamic switching provided by the present application. As shown in Figure 1 , the method comprises the following steps:

[0030] In step 1, the received low-orbit opportunity signal is divided into blocks according to a preset period, and each block signal is preprocessed to obtain a target time frame signal.

[0031] Among the target time frame signal, at least a pilot and a unique word of the low-orbit opportunity signal are included, and the preset period matches a beam polling period of the low-orbit opportunity signal.

[0032] In step 2, the target time frame signal is processed in segments, and each segmented signal of the target time frame signal is subjected to fast Fourier transform (FFT) to obtain a frequency domain peak value of each segmented signal, a segmented signal whose frequency domain peak value meets a preset peak value condition is determined as a target segmented signal, and a Doppler frequency coarse estimation value of the pilot is determined based on the frequency domain peak value in the target segmented signal.

[0033] The preset peak value condition includes that the frequency domain peak value of the current segmented signal is greater than a preset peak value threshold, and after the frequency domain peak values of all segmented signals are sorted from large to small, the frequency domain peak value of the current segmented signal is located in the first N positions, and N is a positive integer.

[0034] In step 3, the local independent word signal is generated using the prior information, and a peak value of a cross-correlation function of the local independent word signal and the general pilot rough position signal is determined as a pilot starting position.

[0035] The general pilot rough position signal is a signal with a distance from the pilot rough position less than a preset distance threshold; and the prior information at least includes prior pilot information and prior unique word information.

[0036] In step 4, the pilot signal is determined based on the pilot starting position, and the pilot signal is subjected to FFT processing again to obtain a pilot signal spectrum, and a frequency offset factor estimation value and an instantaneous signal-to-noise ratio estimation value are determined based on the pilot signal spectrum.

[0037] The number of points of the FFT processing performed again is the same as the number of time domain sampling points of the pilot signal.

[0038] In step 5, the frequency offset factor estimation value and the instantaneous signal-to-noise ratio estimation value are input into a pre-trained KNN decision model to obtain a first algorithm weight and a second algorithm weight.

[0039] The first algorithm is a frequency estimation Rife algorithm based on a phase difference of a discrete time signal, and the second algorithm is a maximum likelihood estimation (MLE) algorithm.

[0040] In step 6, a Doppler frequency compensation value of the pilot is determined using a target algorithm, and a Doppler frequency accurate value of the pilot is determined based on a Doppler frequency coarse estimation value of the pilot and the Doppler frequency compensation value of the pilot.

[0041] When the first algorithm weight is greater than the second algorithm weight, the target algorithm is the first algorithm; and when the second algorithm weight is greater than the first algorithm weight, the target algorithm is the second algorithm.

[0042] In some embodiments of the present application, the method can be executed by a server or a terminal device with certain processing capability. In some embodiments, the method can be used for accurate Doppler spectrum estimation of a low-orbit satellite opportunity signal, and the estimation result is used to assist positioning and navigation.

[0043] In some embodiments of the present application, a low-orbit opportunity signal can be received first, and the received low-orbit opportunity signal is blocked according to a preset period. The low-orbit opportunity signal can be a low-orbit satellite opportunity signal. In an example, the low-orbit satellite opportunity signal can be an Iridium signal.

[0044] The preset period matches with the beam polling period of the low-orbit opportunity signal, so that the pilot section is included in each segmented signal. Then, the pilot signal in each segmented signal can be obtained to perform Doppler frequency estimation by using the pilot signal as an unmodulated continuous wave signal and without prior information.

[0045] Taking the Iridium signal as the low-orbit opportunity signal, since the Iridium system polls 48 beams with a period of 4.32 seconds (s) when broadcasting the signal, the obtained Iridium downlink signal can be segmented with a period of 4.32 s, so that the Iridium time frame signal is ensured to exist in each signal block.

[0046] In some embodiments of the present application, each segmented signal after segmentation can be preprocessed to obtain a target time frame signal, and the target time frame signal at least includes the pilot and unique word of the low-orbit opportunity signal.

[0047] Still taking the Iridium signal as the low-orbit opportunity signal, Figure 2 is a schematic diagram of the time frame structure of the Iridium signal. As Figure 2 shown, the Time division multiple access (TDMA) frame length of the Iridium signal is 90 milliseconds (ms), in which the simplex channel allocation time slot is 20.32 ms, the duplex channel occupies 4 uplink channels of 8.28 ms and 4 downlink channel time slots of 8.28 ms, the frame header allocates a protection interval of 1 ms, a protection interval of 1.24 ms is allocated between the simplex channel and the duplex uplink channel, a protection interval of 0.22 ms is allocated between each time slot of the duplex uplink channel, a protection interval of 0.1 ms is allocated between each time slot of the duplex downlink channel, and a protection interval of 0.24 ms is allocated between the duplex uplink channel and the duplex downlink channel.

[0048] The simplex channel can include a pilot, a unique word and data, in which the pilot occupies a time slot of 2.56 ms, the unique word includes 12 characters, each character occupies a time slot of 0.04 ms, the unique word occupies a time slot of 0.48 ms in total, and the data occupies a time slot of 17.28 ms. The pilot is an unmodulated signal, the unique word is modulated by Binary Phase Shift Keying (BPSK), and the data is modulated by Quadrature Phase Shift Keying (QPSK).

[0049] The simplex channel time frame can be taken as the target time frame signal of the Iridium signal, so that the pilot and the unique word of the Iridium signal are included in the target time frame signal.

[0050] In some embodiments of the present application, the target time frame signal can be segmented and each segment of the target time frame signal can be subjected to a Fast Fourier Transform (FFT) to obtain a frequency domain peak value of each segment. In an example, the target time frame signal of each signal block can be further divided into several segments of equal length, and each segment can be subjected to an FFT to obtain a frequency spectrum of the segment, which includes at least the frequency domain peak value.

[0051] The segment whose frequency domain peak value meets a preset peak value condition can be determined as a target segment, and a Doppler frequency coarse estimation value of the pilot can be determined based on the frequency domain peak value in the target segment. The preset peak value condition can include that the frequency domain peak value of the segment is greater than a preset peak threshold value, and the frequency domain peak value of the segment is among the top N values after sorting the frequency domain peak values of all segments from large to small, where N is a positive integer.

[0052] In an example, the top N peak values greater than the preset peak threshold value among the frequency domain peak values of the segments can be determined as the frequency domain peak values meeting the preset peak value condition. If the number of the frequency domain peak values greater than the preset peak threshold value is less than N, all the frequency domain peak values greater than the preset peak threshold value can be determined as the frequency domain peak values meeting the preset peak value condition. The specific values of the preset peak threshold value and N can be set according to actual needs, for example, the preset peak threshold value can be set as -15 dB and N can be equal to 9.

[0053] In some embodiments of the present application, prior information can also be obtained, which includes at least prior pilot information and prior unique word information. Further, the prior information can be used to generate a local unique word signal, and a peak value of a cross-correlation function between the local unique word signal and a general pilot rough position signal can be determined as a pilot starting position. The general pilot rough position signal is a signal whose distance from the pilot rough position is less than a preset distance threshold value.

[0054] In some embodiments of the present application, a pilot signal can be determined based on the pilot starting position, and the pilot signal can be subjected to an FFT again to obtain a pilot signal frequency spectrum. That is, after the pilot starting position is determined, the pilot signal can be stripped from the pilot starting position based on the time slot length of the pilot signal in the low-orbit opportunity signal. The number of points subjected to the FFT again is the same as the time domain sampling point number of the pilot signal.

[0055] Then, a frequency offset factor estimation value and an instantaneous signal-to-noise ratio estimation value can be determined based on the pilot signal frequency spectrum. The specific implementation methods of determining the frequency offset factor estimation value and the instantaneous signal-to-noise ratio estimation value are described in detail below, and will not be described here again.

[0056] In some embodiments of the present application, a KNN decision model can be pre-trained for determining an algorithm for fine estimation of the pilot based on the frequency offset factor and the instantaneous signal-to-noise ratio. The calculated frequency offset factor estimate and the instantaneous signal-to-noise ratio estimate can be input into the pre-trained KNN decision model to obtain a first algorithm weight and a second algorithm weight, wherein the first algorithm can be a frequency estimation Rife algorithm based on a phase difference of a discrete-time signal, and the second algorithm can be a maximum likelihood estimation (MLE) algorithm.

[0057] The first algorithm weight and the second algorithm weight are compared, and if the first algorithm weight is greater than the second algorithm weight, the first algorithm is determined as the target algorithm, and otherwise if the second algorithm weight is greater than the first algorithm weight, the second algorithm is determined as the target algorithm. The Doppler frequency compensation value of the pilot is determined using the determined target algorithm, and finally the Doppler frequency accurate value of the pilot is determined based on the Doppler frequency coarse estimate value of the pilot and the Doppler frequency compensation value of the pilot.

[0058] By using the technical solution provided by the embodiments of the present application, through the adaptive algorithm matching mechanism, the problem of insufficient accuracy of the traditional method in the complex scene of large signal-to-noise ratio fluctuation and large frequency change is solved, the robustness of Doppler estimation is significantly improved, and the technical solution is suitable for non-cooperative signal processing of low-orbit satellite navigation systems.

[0059] In some embodiments of the present application, when the target time frame signal is acquired, the received low-orbit opportunistic signal can be divided into fixed-length sub-block signals according to the beam polling period of the low-orbit opportunistic signal. Then, each sub-block signal is band-pass filtered to obtain the target time frame signal.

[0060] Still taking the low-orbit opportunistic signal as the Iridium signal as an example, since the bandwidth allocated to the Iridium is 1616.0 megahertz (MHz) to 1626.5 MHz. Among them, 1616.0 MHz to 1626.0 MHz is a duplex channel, which is used as a service channel, and 1626.0 MHz to 1626.5 MHz is a simplex channel, which is used as a signaling channel. Therefore, the lower limit cutoff frequency of the band-pass filter can be set to 1626.0 MHz or a nearby value less than 1626.0 MHz, and the upper limit cutoff frequency can be set to 1626.5 MHz or a nearby value greater than 1626.5 MHz, so as to filter out the simplex channel containing the pilot and the unique word from the Iridium signal as the target time frame signal.

[0061] Figure 3 is a flowchart of a method for determining a Doppler frequency coarse estimate value of a pilot provided by an embodiment of the present application. As shown in Figure 3 , the method comprises the following steps:

[0062] In step 201, each block signal is divided into multiple equal-length signal segments, and an FFT transform is independently performed on each signal segment to obtain the frequency-domain peak value of each segmented signal.

[0063] In step 202, a target segmented signal is determined, and the frequency-domain peak value of each target segmented signal is obtained.

[0064] In step 203, a Doppler frequency coarse estimation value of the pilot is determined based on the pilot rough position.

[0065] In some embodiments of the present application, each block signal can be first divided into multiple equal-length signal segments, and an FFT transform is independently performed on each signal segment to obtain the frequency-domain peak value of each segmented signal. Then, a target segmented signal is determined in the multiple equal-length signal segments based on the frequency-domain peak value of each segmented signal, and the frequency-domain peak value of each target segmented signal is obtained. The specific implementation of determining the target segmented signal is described in detail above, and will not be described here.

[0066] Finally, the Doppler frequency coarse estimation value of the pilot can be determined based on the pilot rough position, where, the Doppler frequency coarse estimation value of the pilot is determined based on the pilot rough position, where, is the Doppler frequency coarse estimation value of the pilot, is the spectral line serial number corresponding to the maximum peak value of each target segmented signal, is the sampling rate of the FFT transform on each signal segment, is the point number of the FFT transform on each signal segment.

[0067] Figure 4 is a flowchart of a method for determining a pilot starting position provided by an embodiment of the present application. As shown in Figure 4 , the method includes the following steps:

[0068] In step 401, a local independent word signal is generated using prior information.

[0069] In step 402, a cross-correlation function of the local independent word signal and the pilot rough position signal is constructed, and the peak value of the cross-correlation function is determined as the pilot starting position.

[0070] In some embodiments of the present application, a local independent word signal can be first generated using prior information , where, is the local independent word signal, is the prior pilot amplitude, is the prior unique word amplitude, is the imaginary symbol, is the Doppler frequency coarse estimation value of the pilot, is the sampling rate for FFT transform of each signal segment, n is the time domain sampling point index of pilot signal, each n value represents a sampling point, is the sampling point number of each independent character of each low orbit opportunity signal, is the prior unique character information, is the rounding up symbol.

[0071] Then, the cross-correlation function of the local independent word signal and the pilot rough position signal is constructed, and the peak value of the cross-correlation function is determined as the pilot start position , wherein is the pilot start position, is the function with the independent variable k, is the maximum value of the function with the independent variable point set, is the cross-correlation function, is the point index of the cross-correlation function, is the point number of , and is the pilot rough position signal.

[0072] As described above, after the pilot start position is determined, the pilot signal can be stripped from the pilot start position. Figure 5 is a schematic diagram of the pilot signal determined by taking Iridium as an example. As shown in Figure 5 , in each 4.32s signal block, at least one 90ms Iridium signal time frame is included, for example, two 90ms Iridium signal time frames are included in the signal block m, and one 90ms Iridium signal time frame is included in the signal block m+1. The time frames can be numbered in time sequence as time frame j, time frame j+1 and time frame j+2. After the pilot start position is determined, a pilot signal can be stripped from each pilot start position backward 2.56ms, and the pilot j, the pilot j+1 and the pilot j+2 are obtained, respectively.

[0073] Each determined pilot signal can be subjected to FFT processing again to obtain the spectrum of each pilot signal. Further, the pilot signal spectrum can be used to determine the frequency offset factor estimate and the instantaneous signal-to-noise ratio estimate.

[0074] In some embodiments, the frequency offset factor estimate can be determined in the following manner: first, the pilot signal spectrum peak value is determined, then the adjacent two spectral lines and of the spectrum peak value are extracted, and finally the formula:

[0075]

[0076] is used.

[0077] The frequency offset factor estimation value is calculated by The frequency offset factor estimation value is denoted as The frequency estimation direction is denoted as

[0078] In some other embodiments, the instantaneous SNR estimation value can be determined in the following way: first, the pilot signal spectrum peak value is determined Then, the instantaneous SNR estimation value is calculated by the formula The instantaneous SNR estimation value is denoted as The number of points of the FFT processing is denoted as The pilot signal spectrum line index is denoted as The pilot signal spectrum line is denoted as The pilot signal spectrum line is denoted as

[0079] Figure 6 is a flowchart of a method for training a KNN decision model provided by an embodiment of the present application. As shown in Figure 6 The method comprises the following steps:

[0080] In step 11, Monte Carlo traversal simulation is performed in the preset frequency offset factor and preset instantaneous SNR interval based on real iridium satellite signal statistics, to obtain the root mean square error of Doppler frequency estimation accuracy when different frequency offset factor and instantaneous SNR combinations use the first algorithm and the second algorithm respectively.

[0081] In step 12, the algorithm label of each frequency offset factor and instantaneous SNR combination is determined based on the accuracy error, to obtain the label points in the pre-trained KNN decision model.

[0082] Each label point comprises a pair of frequency offset factor and instantaneous SNR combination, and the algorithm label of the label point, and the algorithm label is the first algorithm or the second algorithm.

[0083] In some embodiments of the present application, Monte Carlo traversal simulation is performed in the preset frequency offset factor and preset instantaneous SNR interval based on real iridium satellite signal statistics, to obtain the root mean square error of Doppler frequency estimation accuracy when different frequency offset factor and instantaneous SNR combinations use the first algorithm and the second algorithm respectively. Then, the algorithm label of each frequency offset factor and instantaneous SNR combination is determined based on the accuracy error, to obtain the label points in the pre-trained KNN decision model.

[0084] Figure 7 is a schematic diagram of a label point set in a pre-trained KNN decision model provided by an embodiment of the present application. As shown in Figure 7 Each label point comprises a frequency offset factor parameter value and an instantaneous SNR parameter value, and further comprises an algorithm label, which can be the Rife algorithm or the MLE algorithm.​

[0085] Figure 8 is a flowchart of a method for determining a first algorithm weight and a second algorithm weight using a pre-trained KNN decision model according to an embodiment of the present application. As shown in the figure, the method comprises the following steps: Figure 8

[0086] In step 501, the frequency offset factor estimate and the instantaneous signal-to-noise ratio estimate are input into the pre-trained decision model, and the Euclidean distance between the combination of the frequency offset factor estimate and the instantaneous signal-to-noise ratio estimate and each label point is calculated.

[0087] In step 502, the K label points with the smallest Euclidean distance are selected as the decision sample space.

[0088] In step 503, the first algorithm weight and the second algorithm weight are calculated based at least on the decision sample space.

[0089] In some embodiments of the present application, when determining the first algorithm weight and the second algorithm weight, the frequency offset factor estimate and the instantaneous signal-to-noise ratio estimate can be first input into the pre-trained decision model, and the Euclidean distance between the combination of the frequency offset factor estimate and the instantaneous signal-to-noise ratio estimate and each label point is calculated.

[0090] Then the K label points with the smallest Euclidean distance are selected as the decision sample space , wherein is the decision sample space, and K is a positive integer.

[0091] Next, the first algorithm weight can be calculated using the formula , wherein is the first algorithm weight, is a logical judgment symbol, and if is true, the value of is 1, otherwise it is 0, denotes the algorithm label in the corresponding label point of the sample in the decision sample space is the first algorithm, is a non-zero minimum number used to avoid a zero denominator in the fraction.

[0092] That is, when calculating the first algorithm weight, all label points with the algorithm label as the first algorithm in the decision sample space can be first selected, then the reciprocal of the sum of the Euclidean distance from each label point to the decision point and the non-zero minimum number is calculated, and finally the reciprocals calculated for each label point are all added together to obtain the first algorithm weight. The decision point is determined by the frequency offset factor estimate and the instantaneous signal-to-noise ratio estimate.

[0093] ​On the other hand, the second algorithm weight can also be calculated by using the formula wherein is the first algorithm weight, is a logical judgment symbol, and if is true, then is 1, otherwise, is 0, represents the label point corresponding to the sample in the decision sample space with the algorithm label being the second algorithm.

[0094] That is, in the calculation of the second algorithm weight, all label points with the algorithm label being the second algorithm in the decision sample space can be selected first, then the reciprocal of the sum of the Euclidean distance from each label point to the to-be-decided point and the non-zero minimum number is calculated for each label point, and finally the reciprocals calculated for all label points are added together to obtain the second algorithm weight.

[0095] After the first algorithm weight and the second algorithm weight are determined, the target weight can be determined based on the first algorithm weight and the second algorithm weight.

[0096] In an example, if the first algorithm weight is greater than or equal to the second algorithm weight, the target algorithm can be determined as the first algorithm; otherwise, if the second weight is greater than the first weight, the target algorithm can be determined as the second algorithm.

[0097] If the target algorithm is determined as the first algorithm, the Doppler frequency compensation value of the pilot can be determined in the following manner: first, the adjacent two spectral lines and of the pilot signal spectrum peak and are obtained, and then the Doppler frequency compensation value of the pilot is determined by using the formula wherein is the Doppler frequency compensation value calculated using the first algorithm, is the sampling rate of the FFT transformation of each signal segment, is the number of points of the FFT transformation of each signal segment, is the frequency estimation direction, is 1 or -1, is the frequency offset factor estimation value.

[0098] On the other hand, if the target algorithm is determined as the second algorithm, the Doppler frequency compensation value of the pilot can be determined in the following manner: first, the pilot signal is down-converted using the Doppler frequency coarse estimation value of the pilot to obtain a low-frequency signal wherein is the low-frequency signal, is the pilot signal, a Doppler frequency coarse estimation value of the pilot, a sampling rate for performing FFT transformation on each signal segment, n is a time domain sampling point index of the pilot signal, is an imaginary symbol.

[0099] Then, a frequency search is performed on the low-frequency signal near zero frequency, and a frequency value that minimizes a likelihood function of the low-frequency signal is determined as a Doppler frequency compensation value of the pilot, and the Doppler frequency compensation value of the pilot is:

[0100]

[0101] wherein, is an independent variable is a function of a minimum independent variable point set function, is a likelihood function of the low-frequency signal, is a point number of the low-frequency signal.

[0102] After the Doppler frequency compensation value of the pilot is determined, a sum of the Doppler frequency coarse estimation value of the pilot and the Doppler frequency compensation value of the pilot is determined as a Doppler frequency accurate value of the pilot, so as to complete Doppler estimation of the low-orbit opportunity signal.

[0103] Figure 9 is a flowchart of another low-orbit opportunity signal Doppler estimation method provided by the present application, which is oriented to multi-beam dynamic switching. As shown in Figure 9 , the method comprises the following steps: firstly, performing segmented FFT processing on the received low-orbit opportunity signal to realize Doppler frequency coarse estimation of the pilot; then, combining prior independent word information generated locally to accurately position a starting position of the pilot signal, and then stripping the pilot signal. Then, performing feature parameter extraction on the stripped pilot signal, inputting the extracted feature parameter into a pre-trained N decision model, if the decision conclusion is Rife, using Rife algorithm to perform Doppler frequency fine estimation on the pilot signal, if the decision conclusion is MLE, using MLE algorithm to perform Doppler fine estimation on the pilot signal. Finally, combining the Doppler frequency coarse estimation result and the fine estimation result of the pilot to obtain Doppler frequency estimation of the low-orbit opportunity signal.

[0104] The technical scheme provided by the embodiment of the present application can realize high-precision and robust estimation in a large dynamic Doppler scenario. The estimation precision of the traditional frequency refinement estimation algorithm is restricted by the realization of large parameter changes. For example, the performance of the Rife algorithm deteriorates or even fails when the frequency offset factor is small, and the estimation precision of the MLE significantly decreases when the signal-to-noise ratio is low. The technical scheme provided by the embodiment of the present application adaptively selects the optimal accurate frequency estimation algorithm of the low-orbit opportunity signal frame under different parameters according to the feature parameters of each low-orbit opportunity signal frame, realizes the complementary of the advantage parameter interval between different frequency refinement algorithms, and can obtain higher estimation precision under the condition of large-range Doppler frequency and signal-to-noise ratio fluctuation of the measured low-orbit opportunity signal.

[0105] Meanwhile, the technical scheme provided by the embodiment of the present application can accurately identify the position of the pilot signal from the complex scenario of the low-orbit opportunity signal, and extract the feature parameters such as the frequency offset factor and the signal-to-noise ratio. By establishing a feature parameter-algorithm mapping database, the optimal path of the refinement algorithm is adaptively matched, and the optimal estimator matching of different signal characteristics is realized.

[0106] The above-described specific embodiments further illustrate the purpose, technical scheme and beneficial effects of the present application. It should be understood that the above-described specific embodiments are only examples of the present application and are not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for Doppler estimation of low-Earth orbit opportunistic signals for multi-beam dynamic switching, characterized in that, The method comprises: Step 1, block the received low-orbit opportunity signal according to a preset period, and pre-process each block signal to obtain a target time frame signal, wherein the target time frame signal at least includes a pilot and a unique word of the low-orbit opportunity signal; wherein the preset period matches a beam polling period of the low-orbit opportunity signal; Step 2, segment process the target time frame signal, and perform fast Fourier transform (FFT) on each segment signal of the target time frame signal to obtain a frequency domain peak value of each segment signal, determine a segment signal satisfying a preset peak value condition as a target segment signal, and determine a Doppler frequency coarse estimation value of the pilot based on the frequency domain peak value in the target segment signal; wherein the preset peak value condition includes that the frequency domain peak value of the segment signal is greater than a preset peak value threshold, and the frequency domain peak value of the segment signal is located in the first N positions after sorting the frequency domain peak values of all segment signals from large to small, and N is a positive integer; Step 3, generate a local unique word signal using prior information, and determine a peak value of a cross-correlation function between the local unique word signal and a general pilot rough position signal as a pilot starting position, wherein the general pilot rough position signal is a signal with a distance less than a preset distance threshold from the pilot rough position; wherein the prior information at least includes prior pilot information and prior unique word information; Step 4, determine a pilot signal based on the pilot starting position, perform FFT processing on the pilot signal again to obtain a pilot signal spectrum, and determine a frequency offset factor estimation value and an instantaneous signal-to-noise ratio estimation value based on the pilot signal spectrum; wherein the number of points of the FFT processing is the same as the time domain sampling point number of the pilot signal; Step 5, input the frequency offset factor estimation value and the instantaneous signal-to-noise ratio estimation value into a pre-trained K-nearest neighbor (KNN) decision model to obtain a first algorithm weight and a second algorithm weight; wherein the first algorithm is a frequency estimation Rife algorithm based on a phase difference of a discrete time signal, and the second algorithm is a maximum likelihood estimation (MLE) algorithm; Step 6, determine a Doppler frequency compensation value of the pilot using a target algorithm, and determine a Doppler frequency accurate value of the pilot based on the Doppler frequency coarse estimation value of the pilot and the Doppler frequency compensation value of the pilot; wherein when the first algorithm weight is greater than the second algorithm weight, the target algorithm is the first algorithm; and when the second algorithm weight is greater than the first algorithm weight, the target algorithm is the second algorithm.

2. The method of claim 1, wherein, The step 1 comprises: dividing the received low-orbit opportunity signal into block signals with a fixed time length according to a beam polling period of the low-orbit opportunity signal; performing band-pass filtering on each block signal to obtain a target time frame signal.

3. The method of claim 1, wherein, The step 2 comprises: dividing each block signal into a plurality of equal-length signal segments, independently performing FFT transformation on each signal segment to obtain a frequency domain peak value of each segment signal; determining a target segment signal and obtaining a frequency domain peak value of each target segment signal; using the formula determining a Doppler frequency rough estimation value of the pilot based on the pilot rough position; wherein, is a Doppler frequency rough estimation value of the pilot, is a spectral line serial number corresponding to a maximum peak value of each target segmented signal, is a sampling rate of FFT transformation on each signal segment, is a point number of FFT transformation on each signal segment.

4. The method of claim 1, wherein, The step 3 comprises: Generating local independent word signals using a priori information wherein is the local independent word signal, is a priori pilot amplitude, is a priori unique word amplitude, is a virtual sign, is a coarse Doppler frequency estimate of the pilot, is a sampling rate for FFT transform of each signal segment, n is a time domain sampling point index of the pilot signal, each n value represents a sampling point, is the number of sampling points of each independent word signal of each low orbit opportunity signal, is a priori unique word information, is a rounding up sign; constructing a cross-correlation function of the local pilot signal and the omni pilot coarse position signal, determining a peak of the cross-correlation function as a pilot start position ; The pilot start position is wherein is a function of the argument k is a set of points of the argument for which is the cross-correlation function is a point index of the cross-correlation function is is the number of points is the coarse pilot position signal 5. The low earth opportunity signal Doppler estimation method for multi-beam dynamic switching according to claim 1, wherein, The step 4 comprises: Determining pilot signal spectral peaks ; extracting the spectral peak of the two adjacent spectral lines and ; using the formula: ； ； The frequency offset factor estimate is calculated, wherein is the frequency offset factor estimate, is the frequency estimation direction.

6. The low earth opportunity signal Doppler estimation method for multi-beam dynamic switching according to claim 1, wherein, The step 4 further comprises: Determining pilot signal spectral peaks ; The instantaneous signal-to-noise ratio estimate is calculated using the formula wherein is the instantaneous signal-to-noise ratio estimate, is the number of points of the re- performed FFT processing, is the pilot signal spectrum line index, is the first spectrum line in the pilot signal spectrum.

7. The low earth opportunity signal Doppler estimation method for multi-beam dynamic switching according to claim 1, wherein, The pre-trained KNN decision model in the step 5 is obtained by training in the following manner: The Monte Carlo traversal simulation is performed in a preset frequency offset factor and a preset instantaneous signal-to-noise ratio interval based on real iridium star signal statistics, to obtain root mean square errors of Doppler frequency estimation accuracy when the first algorithm and the second algorithm are respectively used for different combinations of frequency offset factors and instantaneous signal-to-noise ratios; An algorithm label of each combination of frequency offset factor and instantaneous signal-to-noise ratio is determined based on the root mean square error, to obtain a label point in the pre-trained KNN decision model; Each label point includes a pair of frequency offset factor and instantaneous signal-to-noise ratio, and an algorithm label of the label point, and the algorithm label is the first algorithm or the second algorithm.

8. The method of claim 7, wherein, The step 5 includes: The frequency offset factor estimation value and the instantaneous signal-to-noise ratio estimation value are input into the pre-trained KNN decision model, and a Euclidean distance between the combination of the frequency offset factor estimation value and the instantaneous signal-to-noise ratio estimation value and each label point is calculated; Selecting K label points with minimum Euclidean distance as decision sample space wherein, is the decision sample space, and K is a positive integer. using the formula to calculate the first algorithm weight, wherein is the first algorithm weight, is a logical judgment symbol, if is true, then the value of is 1, otherwise 0, denotes the decision sample space , wherein the algorithm label in the label point corresponding to the sample is the first algorithm, is a non-zero minimum number, used to avoid the denominator of the fraction being zero; using the formula wherein the second algorithm weight is calculated, is the first algorithm weight, is a logical decision symbol, if is true then the value of is 1, otherwise 0, denotes the decision sample space wherein the algorithm label in the label point corresponding to the sample is the second algorithm.

9. The low earth opportunity signal Doppler estimation method for multi-beam dynamic switching according to claim 1, wherein, The step 6 includes: In response to determining that the target algorithm is a first algorithm, obtaining a pilot signal spectrum peak and adjacent two spectral lines and ; using the formula determining a Doppler frequency compensation value for the pilot, wherein the Doppler frequency compensation value is calculated using a first algorithm, is a sampling rate for the FFT transform of the signal segment, is a number of points for the FFT transform of the signal segment, is a frequency estimation direction, has a value of 1 or -1, is the frequency offset factor estimate; The sum of the coarse Doppler frequency estimation value of the pilot and the Doppler frequency compensation value of the pilot is determined as the accurate Doppler frequency value of the pilot.

10. The low earth opportunity signal Doppler estimation method for multi-beam dynamic switching according to claim 1, wherein, The step 6 also includes: in response to determining that the target algorithm is a second algorithm, down-converting the pilot signal using a Doppler frequency coarse estimation value of the pilot to obtain a low frequency signal wherein is the low frequency signal, is the pilot signal, is the Doppler frequency coarse estimation value of the pilot, is a sampling rate for FFT transforming each signal segment, and n is a time domain sampling point index of the pilot signal, is an imaginary symbol; The low frequency signal is searched in the vicinity of zero frequency to determine the frequency value that minimizes its likelihood function as the Doppler frequency compensation value of the pilot, and the Doppler frequency compensation value of the pilot is is: ； wherein, is a function of is a function of is a function of is a likelihood function of the low frequency signal, is a number of points of the low frequency signal; The sum of the coarse Doppler frequency estimation value of the pilot and the Doppler frequency compensation value of the pilot is determined as the accurate Doppler frequency value of the pilot.

Images