Satellite single-mixed signal identification method based on multi-feature fusion

Abstract

The invention relates to the technical field of wireless communication, in particular to a satellite single-mixed signal identification method based on multi-feature fusion, and the method specifically comprises the following steps: 1, carrying out the down-conversion, timing sampling and conjugate difference preprocessing of a signal; 2, calculating a characteristic parameter LNsr of the number of relative prominence spectral lines of a signal square spectrum after preprocessing; step 3, calculating Shannon entropy characteristic parameters H2x and H4x of the power energy spectrum of the preprocessed signal; 4, high-order cumulant characteristic parameters F1x, F2y and F3y of the preprocessed signals are calculated; 5, according to the characteristic parameters LNsr, H2x, H4x, F1x, F2y and F3y, threshold values are set to identify the satellite single-mixed signals; according to the invention, effective identification of satellite signals including PCMA signals under the conditions of low signal-to-noise ratio, certain amplitude ratio range and existence of frequency offset influence is realized.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, and in particular to a satellite single-mixed signal recognition method based on multi-feature fusion. Background Art

[0002] In the field of blind reception research, modulation recognition of communication signals occupies a crucial position. With the rapid development of communication technology, especially with the emergence of PCMA technology, existing modulation recognition methods are facing the challenge of continuous upgrading and innovation.

[0003] In the PCMA system, both communicating parties share the same frequency band and send two uplink signals with the same modulation method. This design not only significantly saves system bandwidth resources, but also effectively enhances the signal's anti-interference and interception capabilities. At present, the technical difficulty faced by PCMA signal reconnaissance is how to efficiently and accurately identify such signals, estimate their characteristic parameters, and achieve blind separation of signals. Accurate identification of PCMA modulation is the basic work for subsequent parameter estimation, signal blind separation and other processing, so an effective solution is urgently needed.

[0004] Existing modulation recognition methods are mainly aimed at single-carrier signals. The algorithms mainly include high-order cumulants, constellation diagrams, instantaneous features, etc. In recent years, multi-dimensional features combined with deep learning methods have also been used to realize signal recognition, but research on modulation recognition based on PCMA signals is relatively limited. Spooner (1995) used high-order cyclic cumulants as classification features to identify mixed signals with PSK and QAM modulation modes, but the algorithm requires more known parameters (such as cyclic frequency, signal power, etc.). Liao Canhui (2010) proposed an algorithm based on likelihood features. The algorithm requires a large signal-to-noise ratio and is sensitive to frequency offset. Zhang Dongling (2011) used the characteristics of the power spectrum to identify BPSK, QPSK, OQPSK, and π / 4-DQPSK mixed signals, but it needs to rely on manual work and requires a high signal-to-noise ratio. Xu Wansong (2015) distinguished single-channel signals from mixed signals by calculating the flatness parameter of the signal amplitude, but because this parameter is seriously affected by the channel, the algorithm is difficult to be practical. Therefore, it is particularly important to propose an algorithm that is resistant to frequency deviation and amplitude ratio and can effectively identify single-mixed signals under low signal-to-noise ratio. Gong Pei (2023) proposed a recognition algorithm based on cumulant and square spectrum characteristics. The algorithm is insensitive to amplitude ratio, but overly sensitive to frequency deviation. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and propose a satellite single-mixed signal recognition method based on multi-feature fusion, which realizes the effective recognition of satellite signals including PCMA signals under low signal-to-noise ratio, a certain amplitude ratio range and the influence of frequency offset.

[0006] The object of the present invention is to achieve the following measures: a satellite single-mixed signal recognition method based on multi-feature fusion, which completes the modulation recognition of the signal by calculating the square spectrum relative prominence spectral line characteristics, square energy spectrum Shannon entropy, fourth power energy spectrum Shannon entropy and high-order cumulant characteristics of the signal, specifically including the following steps:

[0007] Step 1: The signal is preprocessed by down-conversion, timing sampling and conjugate difference;

[0008] Step 2: Calculate the characteristic parameter LN of the number of relative prominence lines of the square spectrum of the preprocessed signal s r ;

[0009] Step 3: Calculate the Shannon entropy characteristic parameter H of the power energy spectrum of the preprocessed signal 2 x and H 4 x ;

[0010] Step 4: Calculate the high-order cumulative characteristic parameter F of the preprocessed signal 1 x , F 2 y , F 3 y ;

[0011] Step 5: According to the characteristic parameter LN s r , H 2 x , H 4 x 、F 1 x 、F 2 y 、F 3 y Set the threshold to identify satellite single-mix signals.

[0012] Preferably, the step 1 specifically includes:

[0013] According to the PCMA system model, the received signal is modulated and identified after down-conversion, timing sampling and conjugate difference preprocessing. The preprocessed signal sequence can be expressed as:

[0014]

[0015] Among them, r k is the full sampling sequence, {h i} i＝1,2 is the amplitude of the two signals, {m i,n}i＝1,2 is the modulation symbol of the two signals, T is the symbol period, {Δf i} i＝1,2 is the frequency deviation of the signal, {θ i} i＝1,2 is the phase deviation of the signal, {g i (t)} i＝1,2 is the role of shaping filter and channel, {τ i} i＝1,2 is the time delay of the two signals, v(t) is additive white Gaussian noise, and the sampling rate is K times the symbol rate, expressed as f s =K / T, the sampling period is T s ;

[0016] When the instantaneous amplitude h 1 or 2 When h is 0, the receiver obtains a conventional single-carrier signal. 1 and h 2 When both are not 0, the receiving end obtains a PCMA signal. For a single carrier signal, the value of the timing position eliminates the inter-code interference and is the best sampling point. For a PCMA signal, the timing estimation position will be located between the two signals. The sequence extracted from the timing position is:

[0017]

[0018] in, To estimate the timing position, when the frequency deviation is Δf 1 , the phase deviation is θ 1 , and when the timing is accurate, the phase information of the kth and k-1th code elements of a single signal can be expressed as:

[0019]

[0020] in represents the phase information of the kth symbol itself. The phase information of the above two symbols is conjugate-differentially calculated to obtain the differential phase:

[0021]

[0022] From the above formula, we can know that the signal after conjugate difference, in addition to the differential phase information carried by the information code element itself, the differential phase introduced by the frequency offset between any two code elements is the same. This operation reduces the influence of the residual frequency offset and converts the timing sampling sequence x k Perform differential processing and record For x k Therefore, after processing the received signal, the full sampling sequence r is obtained. k , timing sampling sequence x k and the difference sequence yk These preprocessed sequences will serve as the basis for various algorithms to extract feature values ​​and subsequent modulation recognition.

[0023] Preferably, the characteristic parameter LN in step 2 s r The calculation methods include:

[0024] For the full sampling sequence r k Take the Fourier transform of the square and take the modulus:

[0025]

[0026] Where k is the frequency point, and k = 1, 2, ..., N, N is the number of Fast Fourier Transformation (FFT) points, W N =e -i(2π / N) ;

[0027] Let the significance factor λ describing the relative prominence of the spectral line be:

[0028]

[0029] Among them, l is twice the oversampling multiple; max 2 for The sum of the two largest points among the l points, after normalizing the λ amplitude, extract the characteristic parameter LN of the number of spectral lines with amplitude greater than th s (th takes the empirical value of 0.7, and the subscript s takes 1 or 0, indicating whether there are discrete prominent spectral lines near the zero frequency).

[0030] Preferably, the characteristic parameter H in step 3 2 x and H 4 x The calculation methods include:

[0031] The sequence x extracted at the timing position k The Mth power (here, M is 2 or 4) is Fourier transformed to obtain calculate The power energy spectrum value of:

[0032]

[0033] Where k = 0, 1, ..., N-1, N is the number of FFT points,

[0034]

[0035] In the formula, It can be regarded as a probability distribution, a measure of the complexity of signal frequency energy, HM is the power energy spectrum entropy, where H 2 represents the squared energy spectrum Shannon entropy, H 4 Represents the Shannon entropy of the fourth power energy spectrum. The Shannon entropy value of the power energy spectrum of a mixed signal is greater than that of a single signal. The system of the power energy spectrum of a mixed signal is more complex, and there are more types of different situations.

[0036] Preferably, the characteristic parameter F in step 4 1 x , F 2 y , F 3 y The calculation methods include:

[0037] For a stationary random process x k , whose mixing moment and p-order cumulant are defined as:

[0038]

[0039]

[0040] Among them, cum(·)) represents the calculation of high-order cumulants, and “*” represents the conjugate operation. High-order cumulants can be obtained through high-order moments, and there is the following conversion relationship:

[0041] C 20 =M 20

[0042] C 21 =M 21

[0043]

[0044] C 41 =M 41 -M 20 M 21

[0045]

[0046] Using the modulation signal to time the sampling sequence x k The higher-order cumulants C 63 and C 42 Constructing feature parameters Using the differential sequence y of each modulation signal k The higher-order cumulants C 21 , C 40 and C 42 Constructing feature parameters and

[0047] Preferably, the step 5 specifically includes:

[0048] Using LN s r , H 2 x , H 4 x 、F 1 x 、F 2 y and F 3 y Feature parameters, construct a decision tree classifier for recognition, where the superscript of the feature parameter represents the sample sequence of feature extraction, LN s r The subscript indicates whether there are prominent shock lines near the zero frequency. r, x, and y represent the full sampling sequence r, k , timing sampling sequence x k and the difference sequence y k ;

[0049] The total signal set is {BPSK, QPSK, OQPSK, π / 4-DQPSK, 8PSK, 16QAM, BPSK mixed, QPSK mixed, OQPSK mixed, π / 4-DQPSK mixed, 8PSK mixed, 16QAM mixed}, which is determined by LN s r Distinguish {BPSK, BPSK mixed}, {OQPSK, OQPSK mixed}, {QPSK, π / 4-DQPSK, 8PSK, 16QAM, QPSK mixed, π / 4-DQPSK mixed, 8PSK mixed, 16QAM mixed}; by H 2 x and H 4 x Identify BPSK, BPSK mixed, OQPSK, OQPSK mixed; by F 1 x In the remaining signal set, 16QAM and {QPSK mixed, π / 4-DQPSK mixed, 8PSK mixed, 16QAM mixed} and {QPSK, π / 4-DQPSK, 8PSK} are identified; 3 y Identify 16QAM mix as {QPSK mix, π / 4-DQPSK mix, 8PSK mix}, and then use H 2 x and H 4 x Identify π / 4-DQPSK mixing, QPSK mixing, and 8PSK mixing; by H 4 x Identify 8PSK and {QPSK, π / 4-DQPSK}; finally, F2 y QPSK and π / 4-DQPSK are identified.

[0050] Beneficial effects of the present invention: The present invention proposes a modulation recognition algorithm for common modulation signals in the field of satellite communications and PCMA-type aliasing signals based on the joint characteristic parameters of the relative prominence of the signal square spectrum, the power energy spectrum Shannon entropy and the high-order cumulant. The algorithm does not need to rely on prior information such as frequency deviation, channel fading and shaping filter parameters. After a series of simulation tests, the feasibility and superiority of this algorithm have been verified. Specifically, in an environment where the signal-to-noise ratio is not less than 9dB, whether it is a single-carrier signal or a PCMA signal, its overall recognition accuracy can be maintained at more than 99%, effectively resisting the interference caused by amplitude ratio fluctuations and frequency offsets, thereby ensuring the wide applicability and practicality of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a decision tree modulation recognition flow chart of the present invention;

[0052] Figure 2 It is a schematic diagram of the square spectrum of some signals of the present invention;

[0053] Figure 3 It is a schematic diagram of relative prominence of square spectra of some signals of the present invention;

[0054] Figure 4 A schematic diagram of Shannon entropy values ​​of square energy spectra of some signals of the present invention;

[0055] Figure 5 This is a schematic diagram of the Shannon entropy value of the fourth power energy spectrum of some signals of the present invention;

[0056] Figure 6 is the high-order cumulant characteristic parameter F of the present invention 1 Value diagram;

[0057] Figure 7 is the high-order cumulant characteristic parameter F of the present invention 2 Value diagram;

[0058] Figure 8 is the high-order cumulant characteristic parameter F of the present invention 3 Value diagram;

[0059] Fig. 9 This is a schematic diagram showing how the recognition rate of each signal varies with the signal-to-noise ratio of the present invention;

[0060] Fig.10 This is a schematic diagram showing how the recognition rates of various signals vary with the amplitude ratios of the present invention;

[0061] Fig.11It is a schematic diagram of the change of each signal recognition rate with the normalized frequency deviation of the present invention. DETAILED DESCRIPTION

[0062] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] Example 1: Figure 1 As shown, a satellite single-mixed signal recognition method based on multi-feature fusion is used to complete the modulation recognition of the signal by calculating the square spectrum relative prominence spectral line characteristics, square energy spectrum Shannon entropy, fourth power energy spectrum Shannon entropy and high-order cumulant characteristics of the signal. Specifically, the following steps are included:

[0064] Step 1: The signal is preprocessed by down-conversion, timing sampling and conjugate difference, that is, a full sampling sequence, a timing sampling sequence and a conjugate difference sequence are obtained;

[0065] Step 2: Calculate the characteristic parameter LN of the number of relative prominence lines of the square spectrum of the preprocessed signal s r ;

[0066] Step 3: Calculate the Shannon entropy characteristic parameter H of the power energy spectrum of the preprocessed signal 2 x and H 4 x ;

[0067] Step 4: Calculate the high-order cumulative characteristic parameter F of the preprocessed signal 1 x , F 2 y , F 3 y ;

[0068] Step 5: According to the characteristic parameter LN s r , H 2 x , H 4 x 、F 1 x 、F 2 y 、F 3 y Set the threshold to identify satellite single-mix signals.

[0069] In step 1, the received satellite single-mixed signal is preprocessed, and the satellite single-mixed signal model is:

[0070]

[0071] Among them, {h i} i＝1,2 is the amplitude of the two signals, {m i,n} i＝1,2 is the modulation symbol of the two signals, T is the symbol period, {Δf i} i＝1,2 is the frequency deviation of the signal, {θ i} i＝1,2 is the phase deviation of the signal, {g i (t)} i＝1,2 is the role of shaping filter and channel, {τ i} i＝1,2 is the time delay of the two signals, v(t) is additive Gaussian white noise. 1 or 2 When h is 0, the receiver gets a conventional single-carrier signal. 1 and h 2 When both are not 0, the receiving end gets the PCMA signal.

[0072] Oversample the received signal and set the sampling rate to be K times the symbol rate, that is, f s =K / T, the sampling period is T s , the oversampled sequence is:

[0073]

[0074] Among them, r k is the full sampling sequence, {h i} i＝1,2 is the amplitude of the two signals, {m i,n} i＝1,2 is the modulation symbol of the two signals, T is the symbol period, {Δf i} i＝1,2 is the frequency deviation of the signal, {θ i} i＝1,2 is the phase deviation of the signal, {g i (t)} i＝1,2 is the role of shaping filter and channel, {τ i} i＝1,2 is the time delay of the two signals, v(t) is additive white Gaussian noise, and the sampling rate is K times the symbol rate, expressed as f s =K / T, the sampling period is T s ;

[0075] When the instantaneous amplitude h1 or 2 When h is 0, the receiver obtains a conventional single-carrier signal. 1 and h 2 When both are not 0, the receiving end obtains a PCMA signal. For a single carrier signal, the value of the timing position eliminates the inter-code interference and is the best sampling point. For a PCMA signal, the timing estimation position will be located between the two signals. The sequence extracted from the timing position is:

[0076]

[0077] in, To estimate the timing position, when the frequency deviation is Δf 1 , the phase deviation is θ 1 , and when the timing is accurate, the phase information of the kth and k-1th code elements of a single signal can be expressed as:

[0078]

[0079] in represents the phase information of the kth symbol itself. The phase information of the above two symbols is conjugate-differentially calculated to obtain the differential phase:

[0080]

[0081] From the above formula, we can know that the signal after conjugate difference, in addition to the differential phase information carried by the information code element itself, the differential phase introduced by the frequency offset between any two code elements is the same. This operation reduces the influence of the residual frequency offset and converts the timing sampling sequence x k Perform differential processing and record For x k Therefore, after processing the received signal, the full sampling sequence r is obtained. k , timing sampling sequence x k and the difference sequence y k These preprocessed sequences will serve as the basis for various algorithms to extract feature values ​​and subsequent modulation recognition.

[0082] The characteristic parameter LN in step 2 s r The calculation methods include:

[0083] For the full sampling sequence r k Take the Fourier transform of the square and take the modulus:

[0084]

[0085] Where k is the frequency point, and k = 1, 2, ..., N, N is the number of Fast Fourier Transformation (FFT) points, W N =e -i(2π / N) ;

[0086] Let the significance factor λ describing the relative prominence of the spectral line be:

[0087]

[0088] Among them, l is twice the oversampling multiple; max 2 for The sum of the two largest points among the l points, after normalizing the λ amplitude, extract the characteristic parameter LN of the number of spectral lines with amplitude greater than th s (th takes the empirical value of 0.7, and the subscript s takes 1 or 0, indicating whether there are discrete prominent spectral lines near the zero frequency).

[0089] The characteristic parameter H in step 3 2 x and H 4 x The calculation methods include:

[0090] The sequence x extracted at the timing position k The Mth power (here, M is 2 or 4) is Fourier transformed to obtain calculate The power energy spectrum value of:

[0091]

[0092] Where k = 0, 1, ..., N-1, N is the number of FFT points,

[0093]

[0094] In the formula, It can be regarded as a probability distribution, a measure of the complexity of signal frequency energy, H M is the power energy spectrum entropy, where H 2 represents the squared energy spectrum Shannon entropy, H 4 Represents the Shannon entropy of the fourth power energy spectrum. The Shannon entropy value of the power energy spectrum of a mixed signal is greater than that of a single signal. The system of the power energy spectrum of a mixed signal is more complex, and there are more types of different situations.

[0095] The characteristic parameter F in step 4 1 x , F 2 y , F 3 y The calculation methods include:

[0096] For a stationary random process x k , whose mixing moment and p-order cumulant are defined as:

[0097]

[0098] Among them, cum(·) represents the calculation of high-order cumulants, and “*” represents the conjugate operation. High-order cumulants can be obtained through high-order moments, and there is the following conversion relationship:

[0099] C 20 =M 20

[0100] C 21 =M 21

[0101]

[0102] C 41 =M 41 -M 20 M 21

[0103]

[0104] Using the modulation signal to time the sampling sequence x k The higher-order cumulants C 63 and C 42 Constructing feature parameters Using the differential sequence y of each modulation signal k The higher-order cumulants C 21 , C 40 and C 42 Constructing feature parameters and

[0105] Step 5 specifically includes:

[0106] Using LN s r , H 2 x , H 4 x 、F 1 x 、F 2 y and F 3 y Feature parameters, construct a decision tree classifier for recognition, where the superscript of the feature parameter represents the sample sequence of feature extraction, LN s rThe subscript indicates whether there are prominent shock lines near the zero frequency. r, x, and y represent the full sampling sequence r, k , timing sampling sequence x k and the difference sequence y k ;

[0107] The total signal set is {BPSK, QPSK, OQPSK, π / 4-DQPSK, 8PSK, 16QAM, BPSK mixed, QPSK mixed, OQPSK mixed, π / 4-DQPSK mixed, 8PSK mixed, 16QAM mixed}, which is determined by LN s r Distinguish {BPSK, BPSK mixed}, {OQPSK, OQPSK mixed}, {QPSK, π / 4-DQPSK, 8PSK, 16QAM, QPSK mixed, π / 4-DQPSK mixed, 8PSK mixed, 16QAM mixed}; by H 2 x and H 4 x Identify BPSK, BPSK mixed, OQPSK, OQPSK mixed; by F 1 x In the remaining signal set, 16QAM and {QPSK mixed, π / 4-DQPSK mixed, 8PSK mixed, 16QAM mixed} and {QPSK, π / 4-DQPSK, 8PSK} are identified; 3 y Identify 16QAM mix as {QPSK mix, π / 4-DQPSK mix, 8PSK mix}, and then use H 2 x and H 4 x Identify π / 4-DQPSK mixing, QPSK mixing, and 8PSK mixing; by H 4 x Identify 8PSK and {QPSK, π / 4-DQPSK}; finally, F 2 y QPSK and π / 4-DQPSK are identified.

[0108] The characteristic parameters of the simulation results are as follows Figure 1-Figure 8 As shown in the figure, the simulation parameters are set as follows: the number of symbols is 8192, the symbol rate is 1000Baud, the sampling rate is 10KHz, the oversampling factor is 10, the roll-off factor of the raised cosine filter is 0.35, the signal-to-noise ratio is in the range of 0 to 25dB, the step is 1dB, the phase deviation is 0 to 2π, the normalized frequency deviation of the single signal is 1e-4, the two-way normalized frequency deviation of the PCMA signal is ±1e-4, and the amplitudes of the two signals are equal. The channel is an additive white Gaussian noise channel, and the Monte Carlo order is 1000.

[0109] like Figure 2 and Figure 3 It can be seen that after the square spectrum is subjected to relative prominence calculation, the discrete spectral line characteristics of the OQPSK single-mixed signal are more obvious, and there is no discrete spectral line at the zero frequency. Based on this feature, the signal set can be divided into {BPSK, BPSK mixed}, {OQPSK, OQPSK mixed}, {QPSK, π / 4-DQPSK, 8PSK, 16QAM, QPSK mixed, π / 4-DQPSK mixed, 8PSK mixed, 16QAM mixed}.

[0110] like Figure 4 It can be seen that according to the different Shannon entropy values ​​of the signal square energy spectrum, BPSK and BPSK mixing and π / 4-DQPSK mixing and {QPSK mixing and 8PSK mixing} can be distinguished.

[0111] like Figure 5 It can be seen that according to the different Shannon entropy values ​​of the signal's fourth power energy spectrum, OQPSK single-mix signals and QPSK mix and 8PSK mix can be distinguished.

[0112] like Figure 6-Figure 8 It can be seen that F 1 x It can identify 16QAM and {QPSK mixed, π / 4-DQPSK mixed, 8PSK mixed, 16QAM mixed} and {QPSK, π / 4-DQPSK, 8PSK}, and finally F 2 y QPSK and π / 4-DQPSK can be identified. 3 y The 16QAM mix can be identified as {QPSK mix, π / 4-DQPSK mix, 8PSK mix}.

[0113] Fig. 9 Is Figure 1 The signal recognition rate diagram obtained from the overall recognition flow chart shows that when the signal-to-noise ratio level is not less than 9dB, the recognition rate of all examined modulated signals can reach 99%, which effectively solves the problem of satellite single-mixed signal recognition under low signal-to-noise ratio conditions.

[0114] Fig.10 It is a schematic diagram of the variation of the recognition rate of each signal with the amplitude ratio of the present invention. When the amplitude ratio is greater than 0.65, the performance of all mixed signals can reach nearly 100%, which effectively solves the recognition problem of PCMA signals under low amplitude ratio conditions.

[0115] Fig.11 It is a schematic diagram of the change of the recognition rate of each signal of the present invention with the normalized frequency deviation. From the diagram, it can be seen that within the normalized frequency deviation range, the recognition rate of various signals of the algorithm in this paper can be close to 100%, which effectively resists the influence of frequency deviation on the recognition of satellite single-mixed signals.

[0116] The present invention discusses the modulation recognition technology of common modulation signals and PCMA-type aliasing signals in the field of satellite communications, and proposes a joint characteristic parameter recognition algorithm based on the relative prominence of the signal square spectrum, the Shannon entropy of the power energy spectrum and the high-order cumulant. The algorithm does not need to rely on prior information such as frequency deviation, channel fading and shaping filter parameters. After a series of simulation tests, the feasibility and superiority of this algorithm have been verified. Specifically, in an environment where the signal-to-noise ratio is not less than 9dB, whether it is a single carrier signal or a PCMA signal, its overall recognition accuracy can be maintained at more than 99%, effectively resisting the interference caused by amplitude ratio fluctuations and frequency offsets, thereby ensuring the wide applicability and practicality of the algorithm.

[0117] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A satellite single-mixed signal recognition method based on multi-feature fusion, characterized in that: The modulation recognition of the signal is completed by calculating the signal's square spectrum relative prominence spectral line characteristics, square energy spectrum Shannon entropy, fourth power energy spectrum Shannon entropy and high-order cumulant characteristics, which specifically includes the following steps: Step 1: The signal is preprocessed by down-conversion, timing sampling and conjugate difference; Step 2: Calculate the characteristic parameter LN of the number of relative prominence lines of the square spectrum of the preprocessed signal s r ; Step 3: Calculate the Shannon entropy characteristic parameter H2 of the power energy spectrum of the preprocessed signal x and H4 x ; Step 4: Calculate the high-order cumulative characteristic parameter F1 of the preprocessed signal x , F2 y , F3 y ; Step 5: According to the characteristic parameter LN s r 、H2 x 、H4 x 、F1 x 、F2 y 、F3 y Set the threshold to identify satellite single-mix signals.

2. The satellite single-mixed signal recognition method based on multi-feature fusion according to claim 1 is characterized in that: The step 1 specifically includes: The signal is modulated and identified after down-conversion, timing sampling and conjugate difference preprocessing. The preprocessed signal sequence can be expressed as: Among them, r k is the full sampling sequence, {h i } i＝1,2 is the amplitude of the two signals, {m i,n } i＝1,2 is the modulation symbol of the two signals, T is the symbol period, {Δf i } i＝1,2 is the frequency deviation of the signal, {θ i } i＝1,2 is the phase deviation of the signal, {g i (t)} i＝1,2 is the role of shaping filter and channel, {τ i } i＝1,2 is the time delay of the two signals, v(t) is additive white Gaussian noise, and the sampling rate is K times the symbol rate, expressed as f s =K / T, the sampling period is T s ; When the instantaneous amplitude h1 or h2 is 0, the receiver obtains a conventional single-carrier signal. When h1 and h2 are not 0, the receiver obtains a PCMA signal. For a single-carrier signal, the value of the timing position eliminates inter-code interference and is the best sampling point. For a PCMA signal, the timing estimation position will be located between the two signals. The sequence extracted by the timing position is: in, To estimate the timing position, when the frequency deviation is Δf1, the phase deviation is θ1, and the timing is accurate, the phase information of the kth and k-1th code elements of the single signal can be expressed as: in represents the phase information of the kth symbol itself. The phase information of the above two symbols is conjugate-differentially calculated to obtain the differential phase: The timing sampling sequence x k Perform differential processing and record For x k The differential sequence of the received signal is processed to obtain the full sampling sequence r k , timing sampling sequence x k and the difference sequence y k .

3. The satellite single-mixed signal recognition method based on multi-feature fusion according to claim 1 is characterized in that: The characteristic parameter LN in step 2 s r The calculation methods include: For the full sampling sequence r k Take the Fourier transform of the square and take the modulus: Where k is the frequency point, and k = 1, 2, ..., N, N is the number of fast Fourier transform points, W N =e -i(2π / N) ; Let the significance factor λ describing the relative prominence of the spectral line be: Among them, l is twice the oversampling multiple; max2 is The sum of the two largest points among the l points, after normalizing the λ amplitude, extract the characteristic parameter LN of the number of spectral lines with amplitude greater than th s .

4. The satellite single-mixed signal recognition method based on multi-feature fusion according to claim 1 is characterized in that: The characteristic parameter H2 in step 3 x and H4 x The calculation methods include: The sequence x extracted at the timing position k M to the power of M, M is 2 or 4, and Fourier transform is performed to obtain calculate The power energy spectrum value of: Where k = 0, 1, ..., N-1, N is the number of FFT points, In the formula, is a probability distribution, a measure of the complexity of signal frequency energy, H M is the power energy spectrum entropy, where H2 represents the square energy spectrum Shannon entropy and H4 represents the fourth power energy spectrum Shannon entropy.

5. The satellite single-mixed signal recognition method based on multi-feature fusion according to claim 1 is characterized in that: The characteristic parameter F1 in step 4 x , F2 y , F3 y The calculation method includes: using the modulation signal timing sampling sequence x k The higher-order cumulants C 63 and C 42 Constructing feature parameters Using the differential sequence y of each modulation signal k The higher-order cumulants C 21 , C 40 and C 42 Constructing feature parameters and 6. The satellite single-mixed signal recognition method based on multi-feature fusion according to claim 1, characterized in that: The step 5 specifically includes: Using LN s r 、H2 x 、H4 x 、F1 x 、F2 y and F3 y Feature parameters, construct a decision tree classifier for recognition, where the superscript of the feature parameter represents the sample sequence of feature extraction, LN s r The subscript indicates whether there are prominent shock lines near the zero frequency. r, x, and y represent the full sampling sequence r, k , timing sampling sequence x k and the difference sequence y k ; The total signal set is {BPSK, QPSK, OQPSK, π / 4-DQPSK, 8PSK, 16QAM, BPSK mixed, QPSK mixed, OQPSK mixed, π / 4-DQPSK mixed, 8PSK mixed, 16QAM mixed}, which is determined by LN s r Distinguish {BPSK, BPSK mixed}, {OQPSK, OQPSK mixed}, {QPSK, π / 4-DQPSK, 8PSK, 16QAM, QPSK mixed, π / 4-DQPSK mixed, 8PSK mixed, 16QAM mixed}; by H2 x and H4 x Identify BPSK, BPSK mixed, OQPSK, OQPSK mixed; by F1 x In the remaining signal set, 16QAM and {QPSK mixed, π / 4-DQPSK mixed, 8PSK mixed, 16QAM mixed} and {QPSK, π / 4-DQPSK, 8PSK} are identified; by F3 y Identify 16QAM mix as {QPSK mix, π / 4-DQPSK mix, 8PSK mix}, and then use H2 x and H4 x Identify π / 4-DQPSK mixing, QPSK mixing, and 8PSK mixing; by H4 x Identify 8PSK and {QPSK, π / 4-DQPSK}; finally, F2 y QPSK and π / 4-DQPSK are identified.

Images