Method for identifying amplitude-phase modulation signals based on combination of multi-stage blind digital receivers

Abstract

The present invention discloses a method for identifying amplitude-phase modulation signals based on the combination of multi-stage blind digital receivers. By using multi-stage receiver loops of different structural types to extract the instantaneous information of the signals, and using the instantaneous information to statistically calculate the corresponding instantaneous features, and distinguishing the signal types one by one according to the differences between the same features of different types of signals, the signal identification ability for the signal set including {single-tone signal, voice signal, QPSK, 8PSK, 16QAM, 32QAM, 16APSK, 32APSK} can be improved. The internal synchronization loop of the receiver of the present invention has a certain frequency offset capture and tracking ability, and can obtain more stable signal instantaneous information; and the method of distinguishing signals according to different features makes full use of the unique characteristics of the signals; when the signal-to-noise ratio is higher than 10 dB, the recognition rate of the recognition model for single-tone signals, voice signals, QPSK, 8PSK, 16QAM, 16APSK, 32QAM, 32APSK can all reach more than 90%. Therefore, the present invention has the characteristics of low complexity, high recognition rate, strong robustness, etc.

Description

Technical Field

[0001] The present invention mainly aims at the recognition problem among quadrature phase shift keying (QPSK), 8-phase shift keying (8PSK), 16-quadrature amplitude modulation (16QAM), 32-quadrature amplitude modulation (32QAM), 16-amplitude phase shift keying (16APSK), 32-amplitude phase shift keying (32APSK), single-tone signals, and voice signals, and proposes a signal recognition algorithm based on the combination of multi-stage blind digital receivers. Background Art

[0002] In the process of non-cooperative communication, the recognition of the modulation type of the captured signal plays a decisive role in subsequent signal processing. At present, multi-level amplitude-phase hybrid modulation such as MAPSK and MQAM has been widely used in the direction of satellite transmission technology due to its advantages such as high spectrum utilization rate.

[0003] The amplitude-phase modulation signal recognition algorithm based on the combination of multi-stage blind digital receivers aims to propose a method that uses a decision tree as the model architecture, and different branches design receivers with different structures to extract stable instantaneous feature information. The corresponding features are statistically calculated using the signal instantaneous information, and the signal types are distinguished one by one according to the corresponding features in a cascaded manner. This method of using receivers with different structural types at multiple levels to extract features still has a stable effect on feature information extraction in non-cooperative environments where the parameters of the captured signal are unknown or the signal may have undergone complex processing such as shaping filtering. At the same time, the method of distinguishing signals one by one using different types of features makes full use of the obvious differences between the same features of different types of signals. The signal types studied mainly include amplitude-phase hybrid modulation signals, and single-tone signals and voice signals existing in the P-band are added. The invented model has the ability to accurately distinguish single-tone signals, voice signals, and multiple types of amplitude-phase hybrid modulation signals. Summary of the Invention

[0004] The object of the present invention is to improve the signal recognition ability of the signal set including {single-tone signal, voice signal, QPSK, 8PSK, 16QAM, 32QAM, 16APSK, 32APSK}, and propose an amplitude-phase modulation signal recognition method based on the combination of multi-stage blind digital receivers. The algorithm model extracts the instantaneous information of the signal by using receivers with different structural types at multiple levels, statistically calculates the corresponding instantaneous features using the instantaneous information, and distinguishes the signal types one by one according to the differences between the same features of different types of signals. In an environment where the signal-to-noise ratio is higher than 10 dB, the recognition rate of the recognition model of the algorithm proposed by the present invention for single-tone signals, voice signals, QPSK, 8PSK, 16QAM, 16APSK, 32QAM, and 32APSK can all reach more than 90%.

[0005] The technical solution adopted by the present invention to solve its technical problems includes the following steps:

[0006] Step 1: The digital signals processed by the ADC enter different receivers and first undergo parameter estimation processing, including center frequency estimation and symbol rate estimation. According to the parameter estimation results, the signals are resampled, matched filtered, and symbol synchronized. The processed signals will pass through different carrier synchronization structures to extract stable instantaneous information;

[0007] Step 2: For the receiver whose carrier synchronization phase discrimination part is the arctangent phase discrimination loop method, the in-phase component information and instantaneous frequency of the signal are extracted. The in-phase polarity matching degree α is statistically calculated using the in-phase component information. By comparing α with the threshold T0, single-tone signals, voice signals, and other signals are distinguished. The instantaneous frequency variance δ is statistically calculated using the instantaneous frequency. By comparing δ with the threshold T1, single-tone signals and voice signals are distinguished;

[0008] Step 3: The other signals whose specific types have not been determined continue to pass through the receiver whose carrier synchronization phase discrimination part is the polarity decision loop method improved for MPSK signals. Template matching is performed on the in-phase and quadrature information after tracking stability and the 8PSK constellation points with a single amplitude and 8 equally divided phases of 2π angle frequency in the local area. By comparing the template matching degree β with a single-mode equally spaced phase distribution and the threshold T2, the signal set is divided into two categories: {QPSK, 8PSK} and {16QAM, 16APSK, 32QAM, 32APSK}. Then, template matching is performed on the in-phase and quadrature information and the local QPSK template. By comparing the template matching degree ε with a single-mode equally spaced phase distribution and the threshold T3, QPSK and 8PSK are distinguished;

[0009] Step 4: The signals whose specific types have not been determined continue to pass through the receiver whose carrier synchronization phase discrimination part is the d-power decision loop method for high-order amplitude-phase modulation signals. Template matching is performed on the in-phase and quadrature information after tracking stability and the local phase distribution template. By comparing the matching degree η with the threshold T4, the signals are divided into two sets: {16QAM, 16APSK} and {32QAM, 32APSK}. The two sets respectively use the extracted signal instantaneous amplitude to statistically calculate the variance μ1 and μ2 of the amplitude trajectory jump interval, and compare them with the thresholds T5 and T6 to distinguish 16QAM from 16APSK and 32QAM from 32APSK.

[0010] The specific implementation of Step 1 is as follows:

[0011] 1-1. The received signal model is as follows:

[0012] The single-tone signal model is:

[0013]

[0014] where f c is the center frequency of the signal, is the initial phase, and n0(t) follows N(0,σ2 ) Gaussian white noise.

[0015] The speech signal model is:

[0016]

[0017] where A(t) represents the instantaneous amplitude of the signal, represents the instantaneous phase of the signal.

[0018] The amplitude-phase modulation signal set includes QPSK, 8PSK, 16QAM, 32QAM, 16APSK, 32APSK, a total of 6 modulation types, and its general signal model is:

[0019]

[0020] where a(n) is the amplitude of the nth symbol, T is the symbol period, representing the duration of one symbol, and g T (t) is the impulse response of the root-mean-square pulse shaping filter, and f c is the center frequency, is the modulation phase, which remains unchanged within the duration of one symbol, is the initial phase.

[0021] 1 - 2. After the received signal is processed by the ADC, the signal sequence s(n) is obtained for parameter estimation. In the signal information extraction process of the present invention, there is a closed-loop carrier synchronization, so the requirement for the estimation accuracy of the center frequency is not high. Therefore, when the signal modulation type is unknown, the power spectrum is obtained by performing a fast Fourier transform on the signal sequence s(n), and the position with the highest peak is selected throughout the signal power spectrum as the estimated signal center frequency to roughly estimate the signal carrier frequency. The symbol rate estimation of the signal adopts the envelope spectrum estimation method. The signal is multiplied by its conjugate to eliminate the phase information and obtain the instantaneous amplitude information. Since the signal envelope contains a DC component, the DC component needs to be eliminated before performing the fast Fourier transform, and the value corresponding to the impulse spectral line position in the envelope spectrum is used as the estimated signal symbol rate

[0022] 1 - 3. In the present invention, the symbol synchronization adopts the Gardner algorithm. The Gardner algorithm can achieve symbol synchronization without requiring a change in the position of the local sampling clock. The algorithm obtains the optimal sampling position through interpolation operations, ensuring that the signal entering the carrier synchronization is approximately one-fold sampled, reducing the computational complexity of the carrier synchronization, and at the same time improving the signal-to-noise ratio of the signal. Using the estimated symbol rate to adjust the loop filter of the symbol synchronization part can ensure the stable loop tracking performance of the symbol synchronization part.

[0023] The sampling rate of the signal after timing synchronization is four times The signal sampling rate after carrier synchronization is approximately equal to Using the estimated symbol rate By adjusting the loop bandwidth parameters in the timing synchronization loop and the carrier synchronization loop, it can ensure that the parameters are approximately normalized during the signal feature extraction process. This processing can ensure that when the intercepted signal parameter fluctuation range is large, the feature extraction method can still achieve a stable effect, and the overall feature extraction process is guaranteed to be fully automatic and without human intervention.

[0024] In the model designed by the present invention, the information extraction part of the hierarchical receiver utilizes the characteristic that there are differences in the signal information extracted by the carrier synchronization loops of different types of signals through the same phase discrimination method, and statistically analyzes the corresponding signal characteristics according to the signal amplitude, phase, and frequency information extracted hierarchically. The carrier synchronization loop is designed with loops of different phase discrimination methods as the key module of the hierarchical receiver in the model.

[0025] The synchronization loop structure of the carrier synchronization part is the Costas loop method. Assuming that the signal before carrier synchronization is not affected by noise, the sampled real signal can be expressed as:

[0026]

[0027] where n represents the nth symbol,

[0028] The input signal is multiplied by the local oscillator respectively, where f NCO represents the frequency of the local oscillator signal, represents the initial phase of the local oscillator signal. Here, it is assumed that the loop tracking has reached a stable state, that is Then the in-phase component s i (n) and the quadrature component s q (n) obtained after passing through the filter are respectively:

[0029]

[0030]

[0031] In the formula, represents the phase difference between the local oscillator signal and the input signal. When , s i (n)≈1 / 2·I(n), s q (n)≈-1 / 2·Q(n). At this time, the output signal is basically free of frequency offset and phase offset, restoring the mapping situation during baseband signal modulation.

[0032] The loop filter used in the design of the carrier synchronization loop is a second-order loop filter, and the noise bandwidth B in the filterL The parameter is related to the stability of the loop carrier capture ability, and at the same time, the noise bandwidth B L is related to the symbol rate. By using adaptive adjustment of the noise bandwidth B L to improve the stability of tracking.

[0033] Step 2 is specifically implemented as follows:

[0034] 2-1. Compared with other types of signals, the single-tone signal and the speech signal have the characteristic of unipolar amplitude information. And compared with the single-tone signal with a constant frequency, the instantaneous frequency of the speech signal is time-varying. Selecting a phase discrimination method with the ability to quickly capture the instantaneous frequency of the signal as shown in Equation (7) can extract the unipolar or bipolar information of the signal, as well as the instantaneous frequency information.

[0035] u d (n)=atan2(s q (n), s i (n)) (7)

[0036] where u d (n) represents the phase discrimination output result, and atan represents a function that can return the arctangent value of a number.

[0037] 2-2. Compared with the amplitude-phase hybrid modulation signal whose baseband information is non-constant envelope bipolar information, the single-tone signal is a sine wave signal with a single carrier, and the speech signal is a signal with a time-varying instantaneous frequency. After removing the carrier, both signals show unipolarity in the time domain. Based on this difference, carrier capture of the intercepted signal is performed using carrier synchronization in a receiver whose phase discrimination part is the arctangent phase discrimination loop method. When the carrier is synchronized, the in-phase branch component of the single-tone signal and the speech signal is unipolar information, while the in-phase branch component of the amplitude-phase hybrid modulation signal is bipolar information due to the phase jump. Assume that x1(n) represents the distribution result of the nth constellation point output after the signal is processed by the receiver. Then:

[0038]

[0039] In Equation (8), and respectively represent the in-phase component and the quadrature component extracted from the signal by the receiver. The ratio of the number of constellation points on the side with more midpoints on both sides of the vertical axis of the constellation diagram coordinate system in the in-phase component to the total number N of statistical data points is the in-phase polarity matching degree α, as shown in Equation (9):

[0040]

[0041] where p represents the number of constellation points in the case of The number of constellation points of the situation. The single-tone signal, speech signal and other signals are distinguished by using the in-phase polarity matching degree α.

[0042] 2-3. The recognition between the single-tone signal and the speech signal mainly utilizes the difference that the single-tone signal is a single carrier and the speech signal is a time-varying instantaneous frequency. The instantaneous frequency of the single-tone signal oscillates gently and is approximately constant, while the oscillation amplitude of the instantaneous frequency of the speech signal is relatively large. The variance δ of the carrier synchronization instantaneous frequency tracking result in the statistical receiver is calculated, and the two signals are distinguished according to the instantaneous frequency variance δ.

[0043] Step 3 is specifically implemented as follows:

[0044] 3-1. For the recognition of single-mode phase modulation signals QPSK, 8PSK and multi-amplitude phase modulation signals MAPSK, MQAM, an improved phase discrimination method that can stably extract the phase information of single-amplitude signals is selected. For example, the phase discrimination method in Equation (10) can stably extract the instantaneous amplitude and instantaneous phase information of single-amplitude signals.

[0045]

[0046] 3-2. In the signal preprocessing stage, the signal amplitude is normalized. Compared with high-order APSK signals and QAM signals, single-amplitude PSK signals are easier to be stable using receiver synchronization. The receiver can still restore the ideal QPSK and 8PSK constellation diagrams when tracking QPSK and 8PSK signals at a lower signal-to-noise ratio. First, the in-phase and quadrature information after stable tracking is template-matched with the 8PSK constellation points of the local single amplitude and 8 equally divided phase angles of 2π angular frequency. The signal set is divided into two categories {QPSK, 8PSK} and {16QAM, 16APSK, 32QAM, 32APSK} by using the template matching degree β of single-mode equally spaced phase distribution. The specific calculation formula is shown in Equation (11):

[0047]

[0048] In Equation (11), and respectively represent the in-phase component and quadrature component extracted by the receiver for the signal. Ξ 8PSK represents the 8PSK constellation diagram template. The constellation diagram template refers to the distribution area with each constellation point in the ideal constellation diagram as the center and r as the radius. The matching degree β is initialized to 0, the radius r = 0.25. Whenever a constellation point is within the range of Ξ 8PSK the statistical result is β + 1 / N. The QPSK, 8PSK and other signals are distinguished by using the matching degree β.

[0049] 3-3. QPSK has 4 fewer modulation phases than 8PSK. This feature can be used to distinguish between the two. QPSK and 8PSK are distinguished using the matching degree ε of the single-mode equally-spaced phase distribution template. The specific calculation formula is shown in Equation (12):

[0050]

[0051] Ξ QPSK represents the QPSK constellation diagram template, which is the same as Ξ 8PSK Similarly, the matching degree ε is initialized to 0, and the radius r = 0.25.

[0052] The specific implementation of Step 4 is as follows:

[0053] 4-1. For the identification of 16QAM, 16APSK, 32QAM, and 32APSK, the following phase discriminators are used:

[0054]

[0055] Among them, and respectively represent the d-th power of the in-phase component s i (n) and the quadrature component s q (n), and the constant d has different meanings when distinguishing between the two sets of {16QAM, 16APSK} and {32QAM, 32APSK}. When distinguishing between 16QAM and 16APSK, d = 3, as shown in Equation (14).

[0056]

[0057] After the constellation diagrams of the two signals of 16APSK and 16QAM are respectively cubed, the distribution of constellation points becomes more divergent, the amplitude difference between the inner and outer circle radii becomes larger, and the distribution points of the outer circle are similar to the traditional QPSK signal, with a relatively higher signal-to-noise ratio. At the same time, the power of the inner circle is close to zero. The phase error can be updated using the distribution points of the outer circle. Compared with the general polarity decision phase discriminator, this method of tracking high signal-to-noise ratio constellation points is more stable.

[0058] When distinguishing between 32APSK signals and 32QAM signals, d = 4, as shown in Equation (15). The phase discrimination method has similar characteristics to those of 16QAM and 16APSK.

[0059]

[0060] 4-2. For the recognition of the signal sets of {16QAM, 16APSK, 32QAM, 32APSK}, due to the complex and dense constellation points, template matching for all constellation points will be greatly affected by noise. The present invention adopts the method of selecting constellation points in some regions for template matching. First, the set is divided into two categories: {16QAM, 16APSK} and {32QAM, 32APSK}. When the symbols are equally probable, the number of symbols distributed at each constellation point of the signal is approximately equal. The ratio of the number of constellation points whose instantaneous phase of the ideal baseband signal is located in the regions of {±π / 4, ±3π / 4} to the total number of points is shown in Table 1:

[0061]

[0062] Due to the influence of noise and phase deviation, the instantaneous phase of the signal will not completely coincide with the ideal situation. The included angle θ is the allowable deviation angle when used as a statistical feature, and {-π / 4, -3π / 4, π / 4, 3π / 4} ± θ is the phase distribution angle range of the phase distribution template.

[0063] Use the receiver to extract the in-phase information of the signal and the quadrature information Statistically calculate the ratio η of the number of points in the specified region to the total number of points N, as shown in Equation (16):

[0064]

[0065] Among them, Ξ phase represents the phase distribution template. In the template designed by the present invention, the allowable deviation phase θ = π / 32 is set, and the phase distribution template matching degree η is initialized to 0. The signal is divided into two sets of {16QAM, 16APSK} and {32QAM, 32APSK} using the matching degree η.

[0066] 4-3. High-order amplitude-phase hybrid modulation signals have the characteristic of multiple amplitudes. The present invention distinguishes signals through the amplitude trajectory difference. Ideally, the constellation points in the constellation diagram are sorted in ascending order of amplitude point by point, and the number of constellation points for each amplitude is k i , i = 1, 2,..., P, where P is the number of amplitude cases, and the signal amplitude trajectory jump position L j , j = 1,..., P - 1 has the following relationship with the number of symbols N and the signal base M:

[0067]

[0068] Substitute the signal base M and the number of amplitudes P in the signal sets of {16QAM, 16APSK, 32QAM, 32APSK} into Equation (17) one by one, and the specific signal amplitude trajectory jump position L j is shown in Table 2.

[0069]

[0070] It can be found from Table 2 that there are differences in the amplitude of the two types of signals {16QAM, 16APSK} and {32QAM, 32APSK}. Different amplitude allocations result in different numbers of constellation points, which bring differences in the signal amplitude trajectory jump positions. Using this feature, the two types of signals can be identified between classes.

[0071] First, sort the information amplitude |x m (n)| output by the phase discriminator to obtain the information sequence D m (n), and calculate the curvature C m (n) of the information sequence D m :

[0072]

[0073]

[0074] Among them, m = 3 and m = 4 respectively represent the data results of the signal passing through two different receivers.

[0075] Select the trajectory jump position L for distinguishing the signal j , and statistically count the logarithm of the variance of P data points near the trajectory jump position L m of the curvature distribution C j as the characteristic parameter μ k for signal recognition. The expression is as follows:

[0076]

[0077] Here, the distinction between 16QAM and 16APSK uses L j = 3N / 4 to obtain the variance μ1 of the amplitude trajectory jump interval; the distinction between 32QAM and 32APSK uses L j = 3N / 8 to obtain the variance μ2 of the amplitude trajectory jump interval. Since the QAM signal has an amplitude jump at the trajectory jump position L j while the APSK signal has a constant amplitude at the trajectory jump position L j , the characteristic parameter μ k of the QAM signal is significantly greater than the characteristic parameter μ k of the APSK signal. Based on this feature, the two sets are identified respectively.

[0078] The beneficial effects of the present invention are as follows:

[0079] 1. The internal synchronization loop of the receiver has a certain frequency offset capture and tracking ability, and can obtain more stable instantaneous signal information compared with the classification model of traditional recognition algorithms based on statistical signal features.

[0080] 2. The method of distinguishing signals according to different features makes full use of the unique characteristics of the signals.

[0081] 3. When the signal-to-noise ratio is higher than 10 dB, the recognition rate of the recognition model for single-tone signals, speech signals, QPSK, 8PSK, 16QAM, 16APSK, 32QAM, and 32APSK can all reach over 90%.

[0082] In summary, the algorithm of the present invention has the characteristics of low complexity, high recognition rate, strong robustness, etc., and still has the advantage of a high recognition rate under low signal-to-noise ratio and with frequency offset. Description of the Drawings

[0083] Figure 1 Design of the signal recognition model;

[0084] Figure 2 Flowchart for extracting amplitude, phase and frequency information of the receiver;

[0085] Figure 3 Curve of the in-phase polarity matching degree α varying with the signal-to-noise ratio;

[0086] Figure 4 Curve of the instantaneous frequency variance δ varying with the signal-to-noise ratio;

[0087] Figure 5 Curve of the 8PSK template matching degree β varying with the signal-to-noise ratio;

[0088] Figure 6 Curve of the QPSK template matching degree ε varying with the signal-to-noise ratio;

[0089] Figure 7 Curve of the phase distribution matching degree η varying with the signal-to-noise ratio;

[0090] Figure 8 Curve of the variance μ1 of the amplitude trajectory jump region varying with the signal-to-noise ratio;

[0091] Figure 9 Curve of the variance μ2 of the amplitude trajectory jump region varying with the signal-to-noise ratio.

[0092] Figure 10 Recognition rates of three algorithms under different signal-to-noise ratios Detailed Embodiments

[0093] The following further describes the specific embodiments of the present invention with reference to the drawings.

[0094] Step 1. The complete amplitude-phase modulation signal recognition algorithm model based on the joint multi-stage blind digital receiver is as shown in Figure 1 wherein, for the signal s(n) entering different receivers, the processes of center frequency estimation, symbol rate estimation, resampling, matched filtering, and symbol synchronization are as shown in Figure 2 ;

[0095] Step 2. Figure 1 In , receiver A is a receiver with an arctangent phase discrimination loop method for the carrier synchronization phase discrimination part. The in-phase polarity matching degree α and the instantaneous frequency variance δ are obtained by processing the signal with receiver A. α is compared with the threshold T0 to distinguish single-tone signals, voice signals, and other signals. At the same time, δ is compared with the threshold T1 to distinguish single-tone signals and voice signals;

[0096] Step 3. Figure 1 In , receiver B is a receiver with a polarity decision loop method improved for MPSK signals for the carrier synchronization phase discrimination part. The 8PSK template matching degree β with a single-mode equally spaced phase distribution and the QPSK template matching degree ε with a single-mode equally spaced phase distribution are obtained by processing the signal with receiver B. β is compared with the threshold T2 to distinguish QPSK, 8PSK, and other signals. At the same time, ε is compared with the threshold T3 to distinguish QPSK and 8PSK;

[0097] Step 4. Figure 1 In , receivers C and D are receivers with a d-th power decision loop method for high-order amplitude-phase modulation signals. d = 3 in receiver C and d = 4 in receiver D. The phase distribution matching degree η is obtained by processing the signal with receiver C, and the variance μ1 of the amplitude trajectory jump region and the variance μ2 of the amplitude trajectory jump region are obtained by processing the signal with receivers C and D respectively. η is compared with the threshold T4 to classify the signal into two categories of {16QAM, 16APSK} and {32QAM, 32APSK}. μ1 is compared with the threshold T5 to distinguish 16QAM and 16APSK, and μ2 is compared with the threshold T6 to distinguish 32QAM and 32APSK.

[0098] The specific implementation of Step 1 is as follows:

[0099] 1-1. The center frequency estimation adopts power spectrum estimation, and the estimated center frequency is used to perform down-conversion on the signal.

[0100] 1-2. The symbol rate estimation adopts envelope spectrum estimation, and the signal is resampled at a sampling rate four times the estimated symbol rate , and the loop bandwidth parameters in the timing synchronization loop and the carrier synchronization loop are adjusted using the estimated symbol rate .

[0101] 1-3. The symbol synchronization design adopts the Gardner algorithm. Among them, the interpolation filter selects the 4th-order cubic Lagrange polynomial structure, and the loop filter selects the second-order loop filter structure.

[0102] Step 2 is specifically implemented as follows:

[0103] 2-1. The curve of the in-phase polarity matching degree α of each signal varying with the signal-to-noise ratio is as Figure 3 shown. Take T0 = 0.7 as the threshold. When α > T0, it is judged as a tone or speech signal; otherwise, it is other signals.

[0104] 2-2. The curve of the instantaneous frequency variance δ of the two signals of tone and speech varying with the signal-to-noise ratio is as Figure 4 shown. Take T1 = 0.6e-5 as the threshold. When δ > T1, it is judged as a speech signal; otherwise, it is a tone signal.

[0105] Step 3 is specifically implemented as follows:

[0106] 3-1. The curve of the template matching degree β of different signals varying with the signal-to-noise ratio is as Figure 5 , in an ideal situation, the matching degree β of QPSK and 8PSK is approximately 1. Due to the influence of noise, the matching degree β of QPSK and 8PSK decreases as the signal-to-noise ratio decreases. Here, take T2 = 0.55 as the threshold. When β > T2, it is judged as a QPSK or 8PSK signal; otherwise, it is other signals.

[0107] 3-2. The curve of the template matching degree ε of 8PSK and QPSK varying with the signal-to-noise ratio is as Figure 6 , in an ideal situation, the matching degree ε of QPSK is approximately 1, and the matching degree ε of 8PSK is approximately 0.5. Due to the influence of noise, the matching degree ε of QPSK and 8PSK decreases as the signal-to-noise ratio decreases. Take T3 = 0.6 as the threshold. When ε > T3, it is judged as a QPSK signal; otherwise, it is an 8PSK signal.

[0108] Step 4 is specifically implemented as follows:

[0109] 4-1. The curve of the template matching degree η of the phase distribution of various signals varying with the signal-to-noise ratio is as Figure 7 , in an ideal situation, the matching degree η of 16APSK and 16QAM is approximately 0.5. Due to the influence of phase noise, the matching degree η of 16APSK and 16QAM decreases as the signal-to-noise ratio decreases. Here, take T4 = 0.26 as the threshold. When η > T4, it is judged as {16QAM, 16APSK} signal; otherwise, it is {32QAM, 32APSK} signal.

[0110] 4-2. The curve of the variance μ1 of the amplitude trajectory jump interval varying with the signal-to-noise ratio is as Figure 8, take T5 = -16 as the threshold. When μ1 > T5, it is judged as a 16QAM signal; otherwise, it is a 16APSK signal.

[0111] 4-3. The curve of the variance μ2 of the amplitude trajectory jump interval changing with the signal-to-noise ratio is as Figure 9 , take T6 = -16 as the threshold. When μ2 > T6, it is judged as a 32QAM signal; otherwise, it is a 32APSK signal.

[0112] Example:

[0113] The single-tone signal and the speech signal in the experimental signal set are the actual satellite channel signals in the P band collected by the antenna. QPSK, 8PSK, 16QAM, 16APSK, 32QAM, and 32APSK are the actual satellite signals provided by a certain research institute of China Electronics Technology Group. Since the signal-to-noise ratio of the signals is relatively high, it is defaulted here as signals without noise. The specific parameters are as follows: the symbol rate range is 2k to 200 kbps, the signal frequency offset range is 0 to 0.01 times the symbol rate, and the number of sampling points is 16384. Among them, the amplitude-phase mixed modulation signal passes through a raised cosine pulse shaping filter, and the shaping coefficient is 0.8. Band-limited Gaussian white noise is added to the collected signals through MATLAB 2020, and the signal-to-noise ratio is 6 to 20 dB.

[0114] The recognition algorithms proposed in this invention and the recognition algorithms in two journal papers, namely "Modulation Recognition Algorithm Based on Constellation Trajectory Diagram in Dense Signal Environment" (Literature 1) and "A Satellite Amplitude-Phase Modulation Signal Recognition Algorithm Against Frequency Offset" (Literature 2), are respectively used to perform recognition simulation tests on the signals under MATLAB 2020 software. Each test conducts 1000 Monte Carlo experiments, and the number of correct recognitions is statistically counted, and the recognition rate is calculated. The recognition rates of the three algorithms are as Figure 10 shown.

[0115] Observe Figure 10 It can be seen that when the signal-to-noise ratio is higher than 10 dB, the recognition rate of the recognition model proposed in this invention for single-tone signals, speech signals, QPSK, 8PSK, 16QAM, 16APSK, 32QAM, and 32APSK can all reach more than 90%. Under different signal-to-noise ratios, for signals of the same modulation type, the recognition effect of the algorithm in this invention is better than that of the other two algorithms; when the signal-to-noise ratio is 8 dB, the other two algorithms no longer have the recognition ability for some high-order modulation signal types, but the algorithm in this invention still has a recognition rate of more than 75%. Moreover, the algorithms proposed in Literature 1 and Literature 2 do not have the ability to distinguish {single-tone signals, speech signals} and {QPSK, 8PSK} with the same amplitude. Therefore, the overall effect of the recognition algorithm in this invention is better than the recognition algorithms proposed in Literature 1 and Literature 2.

[0116] Finally, it should be noted that the purpose of disclosing the embodiments is to help further understand the present invention. However, those skilled in the art can understand that various substitutions and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the present invention should not be limited to the content disclosed in the embodiments, and the scope of protection claimed by the present invention shall be subject to the scope defined by the claims.

Claims

1. A method for identifying amplitude-phase modulation signals based on the combination of multi-stage blind digital receivers, characterized in that This method extracts the instantaneous information of the signal by adopting multi-level receivers with different structural types, uses the instantaneous information to statistically calculate the corresponding instantaneous features, and distinguishes the signal types one by one according to the differences between the same features of different types of signals: Specifically, it includes the following steps: Step 1: The digital signal processed by the ADC enters different receivers and first undergoes parameter estimation processing, including center frequency estimation and symbol rate estimation. According to the parameter estimation results, the signal is resampled, matched filtered, and symbol synchronized. The processed signal will pass through different carrier synchronization structures to extract stable instantaneous information; Step 2: For the receiver whose carrier synchronization phase discrimination part is the arctangent phase discrimination loop method, the in-phase component information and instantaneous frequency of the signal are extracted; the in-phase polarity matching degree α is statistically calculated using the in-phase component information. By comparing α with the threshold T0, single-tone signals, voice signals, and other signals are distinguished; The instantaneous frequency variance δ is statistically calculated using the instantaneous frequency. By comparing δ with the threshold T1, single-tone signals and voice signals are distinguished; Step 3: The other signals whose specific types have not been determined continue to pass through the receiver whose carrier synchronization phase discrimination part is the polarity decision loop method improved for MPSK signals. The in-phase and quadrature information after tracking stability is template-matched with the 8PSK constellation points of the local single amplitude and 8 equally divided phase angles of 2π. By comparing the template matching degree β of the single-mode equally spaced phase distribution with the threshold T2, the signal set is divided into two categories: {QPSK, 8PSK} and {16QAM, 16APSK, 32QAM, 32APSK}; then the in-phase and quadrature information is template-matched with the local QPSK template. By comparing the template matching degree ε of the single-mode equally spaced phase distribution with the threshold T3, QPSK and 8PSK are distinguished; Step 4: The signals whose specific types have not been determined continue to pass through the receiver whose carrier synchronization phase discrimination part is the d-th power decision loop method for high-order amplitude-phase modulation signals. The in-phase and quadrature information after tracking stability is template-matched with the local phase distribution template. By comparing the matching degree η with the threshold T4, the signals are divided into two sets: {16QAM, 16APSK} and {32QAM, 32APSK}; the two sets respectively use the extracted instantaneous amplitude of the signal to statistically calculate the variance μ1, μ2 of the amplitude trajectory jump interval, and compare them with the thresholds T5, T6 to distinguish 16QAM from 16APSK and 32QAM from 32APSK.

2. The method for identifying amplitude-phase modulation signals based on the combination of multi-stage blind digital receivers according to claim 1, characterized in that The specific implementation of Step 1 is as follows: 1-1. The received signal model is as follows: The single-tone signal model is: Among them, f c is the center frequency of the signal, is the initial phase, and n0(t) is Gaussian white noise that follows N(0,σ 2 ); The voice signal model is: where A(t) represents the instantaneous amplitude of the signal, represents the instantaneous phase of the signal; The amplitude-phase modulation signal set includes 6 modulation types: QPSK, 8PSK, 16QAM, 32QAM, 16APSK, 32APSK. Its general signal model is: where a(n) is the amplitude of the n-th symbol, T is the symbol period representing the duration of a symbol, g T (t) is the impulse response of the root-mean-square pulse shaping filter, f c is the center frequency, is the modulation phase, which remains constant within the duration of a symbol, is the initial phase; 1-2. After the received signal is processed by the ADC, the signal sequence s(n) is used for parameter estimation. The fast Fourier transform is performed on the signal sequence s(n) to obtain the power spectrum, and the position with the highest peak is selected by traversing the entire signal power spectrum as the estimated signal center frequency to roughly estimate the signal carrier frequency; the symbol rate estimation of the signal adopts the envelope spectrum estimation method. The signal is multiplied by its own conjugate to eliminate the phase information and obtain the instantaneous amplitude information. Since the signal envelope contains a DC component, the DC component needs to be eliminated before performing the fast Fourier transform, and the value corresponding to the impulse spectral line position in the envelope spectrum is used as the estimated signal symbol rate. 1-3. The Gardner algorithm is used for symbol synchronization, and the estimated symbol rate is utilized. Adjust the loop filter of the symbol synchronization part. The signal sampling rate after timing synchronization is quadrupled The signal sampling rate after carrier synchronization is approximately equal to Using the estimated symbol rate Adjust the loop bandwidth parameters in the timing synchronization loop and the carrier synchronization loop to ensure that the parameters are approximately normalized during the signal feature extraction process; 1-4. The designed hierarchical receiver information extraction part utilizes the characteristic that there are differences in the signal information extracted by the carrier synchronization loops of different types of signals through the same phase discrimination method. According to the amplitude, phase, and frequency information of the signals extracted hierarchically, the corresponding signal features are statistically calculated. The carrier synchronization loop is designed as the key module of the hierarchical receiver in the model with loops of different phase discrimination methods; The synchronization loop structure of the carrier synchronization part is the Costas loop method; assuming that the signal before carrier synchronization is not affected by noise, the sampled real signal can be expressed as: where n represents the nth symbol, The input signal is multiplied by the local oscillator respectively, where f NCO represents the frequency of the local oscillator signal, and represents the initial phase of the local oscillator signal; it is assumed here that the loop tracking has reached a steady state, that is Then the in-phase component s i (n) and the quadrature component s q (n) are respectively: In the formula, represents the phase difference between the local oscillation signal and the input signal. When occurs, s i (n) ≈ 1 / 2·I(n), s q (n) ≈ -1 / 2·Q(n). At this time, the output signal is basically free of frequency offset and phase offset, restoring the mapping situation during baseband signal modulation; The loop filter used in the carrier synchronization loop design is a second-order loop filter, and the noise bandwidth B in the filter L The parameter is related to the stability of the loop carrier acquisition ability. At the same time, the noise bandwidth B L is related to the symbol rate. By using adaptive adjustment of the noise bandwidth B L to improve the stability of tracking.

3. The amplitude-phase modulation signal recognition method based on the combination of multi-stage blind digital receivers according to claim 2, wherein Step 2 is specifically implemented as follows: 2-1. Select a phase discrimination method with the ability to quickly capture the instantaneous frequency of the signal as shown in Equation (7) to extract the unipolar or bipolar information of the signal, as well as the instantaneous frequency information; u d u(n) = atan2(s q (n), s i (n)) (7) where u d (n) represents the phase discrimination output result; 2-2. Compared with the bipolar information of the baseband information of the amplitude-phase hybrid modulation signal being non-constant envelope, the single-tone signal is a sine wave signal of a single carrier, and the voice signal is a signal with time-varying instantaneous frequency. After removing the carrier, both signals are unipolar in the time domain. Based on this difference, use the carrier synchronization in the receiver with the arctangent phase discrimination loop method as the phase discrimination part to perform carrier capture on the intercepted signal; when the carrier is synchronized, the in-phase branch component of the single-tone signal and the voice signal is unipolar information, while for the amplitude-phase hybrid modulation signal, due to the phase jump, its in-phase branch component is bipolar information; assume that x1(n) represents the distribution result of the nth constellation point output after the signal is processed by the receiver, then: In Equation (8), and respectively represent the in-phase component and the quadrature component extracted by the receiver. The ratio of the number of constellation points on the side with more midpoints on both sides of the vertical axis of the constellation diagram coordinate system of the in-phase component to the total number of statistical data points N is the in-phase polarity matching degree α, as shown in Equation (9): where p represents the number of constellation points in the case, and q represents the number of constellation points in the case; the single-tone signal, voice signal and other signals are distinguished by using the in-phase polarity matching degree α; 2-3. The identification between the single-tone signal and the voice signal mainly uses the difference that the single-tone signal is a single carrier and the voice signal has time-varying instantaneous frequency. The instantaneous frequency oscillation of the single-tone signal is gentle and approximately constant, and the instantaneous frequency oscillation amplitude of the voice signal is large. Statistically calculate the variance δ of the instantaneous frequency tracking result of the carrier synchronization in the receiver, and distinguish the two signals according to the instantaneous frequency variance δ.

4. The method for identifying amplitude-phase modulation signals based on the combination of multi-stage blind digital receivers according to claim 1 or 3, characterized in that Step 3 is specifically implemented as follows: 3-1. For the identification of single-mode phase modulation signals QPSK, 8PSK and multi-amplitude phase modulation signals MAPSK, MQAM, select a phase discrimination method as shown in Equation (10) to stably extract the instantaneous amplitude and instantaneous phase information of the single-amplitude signal; 3-2. First, perform template matching on the in-phase and quadrature information after stable tracking and the 8PSK constellation points of the local single amplitude and 8 equally spaced phases dividing the 2π angular frequency. Use the template matching degree β of the single-mode equally spaced phase distribution to divide the signal set into two categories: {QPSK, 8PSK} and {16QAM, 16APSK, 32QAM, 32APSK}. The specific calculation formula is shown in Equation (11): In formula (11), and respectively represent the in-phase component and the quadrature component of the signal extracted by the receiver. Ξ 8PSK represents the 8PSK constellation diagram template. The constellation diagram template refers to the distribution area centered on each constellation point in the ideal constellation diagram with a radius of r; the matching degree β is initialized to 0, the radius r = 0.25, and every time a constellation point is located within Ξ 8PSK the range, the statistical result is β + 1 / N; the matching degree β is used to distinguish QPSK, 8PSK and other signals; 3-3. QPSK has 4 fewer modulation phases than 8PSK. Use this feature to distinguish the two; use the template matching degree ε of the single-mode equally spaced phase distribution to distinguish QPSK and 8PSK. The specific calculation formula is shown in Equation (12): Ξ QPSK represents the QPSK constellation diagram template, and is the same as Ξ 8PSK Similarly, the matching degree ε is initialized to 0, and the radius r = 0.

25.

5. The method for identifying amplitude-phase modulation signals based on the combination of multi-stage blind digital receivers according to claim 4, characterized in that Step 4 is specifically implemented as follows: 4-1. For the identification of 16QAM, 16APSK, 32QAM, 32APSK, the phase discriminator used is as follows: Among them, and respectively represent the d-th power of the in-phase component s i (n) and the quadrature component s q (n), and the constant d has different meanings when distinguishing within the two sets of {16QAM, 16APSK} and {32QAM, 32APSK}; when distinguishing 16QAM and 16APSK, d = 3, as shown in Equation (14); When distinguishing the 32APSK signal and the 32QAM signal, d = 4, as shown in Equation (15), and the phase discrimination method has characteristics similar to those of 16QAM and 16APSK; 4-2. For the recognition of the signal sets of {16QAM, 16APSK, 32QAM, 32APSK}, template matching is performed by selecting constellation points in some regions. First, the set is divided into two categories: {16QAM, 16APSK} and {32QAM, 32APSK}; when the symbols are equally probable, the number of symbols distributed at each constellation point of the signal is approximately equal. The proportion of the number of constellation points of the ideal baseband signal whose instantaneous phase is located in the region of {±π / 4, ±3π / 4} is counted as follows: The proportion of the 16QAM class located at the coordinate diagonal points is 1 / 2; The proportion of the 16APSK class located at the coordinate diagonal points is 1 / 2; The proportion of the 32QAM class located at the coordinate diagonal points is 1 / 4; The proportion of the 32APSK class located at the coordinate diagonal points is 1 / 4; Due to the influence of noise and phase offset, the instantaneous phase of the signal will not completely coincide with the ideal situation. The included angle θ is the allowable offset angle when used as a statistical feature, and {-π / 4, -3π / 4, π / 4, 3π / 4} ± θ is the phase distribution angle range of the phase distribution template; Using a receiver to extract the in-phase information of a signal and the quadrature information Statistically calculate the ratio η of the number of points in a specified area to the total number of points N, as shown in Equation (16): Among them, Ξ phase represents a phase distribution template. In the template, the allowed offset phase θ = π / 32 is set, and the matching degree η of the phase distribution template is initialized to 0; the signals are divided into two sets {16QAM, 16APSK} and {32QAM, 32APSK} using the matching degree η; 4-3. Signal discrimination by amplitude trajectory difference; ideally, the constellation points in the constellation diagram are sorted in ascending order of amplitude point by point, and the number of constellation points for each amplitude is k i , i = 1, 2, …, P, where P is the number of amplitude cases, and the signal amplitude trajectory jump position L j , j = 1, …, P - 1, and the relationships with the number of symbols N and the signal radix M are as follows: Substitute the signal order M and the number of amplitudes P in the signal set {16QAM, 16APSK, 32QAM, 32APSK} into Equation (17) one by one, and the specific signal amplitude trajectory jump position L j is as follows; The number of amplitudes of 16QAM is 3, and the amplitude trajectory jump position is L j , j = 1, …, P - 1; The number of amplitudes of the 16APSK class is 2, and the amplitude trajectory jump positions are N / 4 and 3N / 4; The number of amplitudes of the 32QAM class is 5, and the amplitude trajectory jump positions are N / 8, 3N / 8, 4N / 8, and 6N / 8; The number of amplitudes of the 32APSK class is 3, and the amplitude trajectory jump positions are N / 8 and 4N / 8; Therefore, it can be found that there are differences in the amplitude situations within both the two signal categories of {16QAM, 16APSK} and {32QAM, 32APSK}. The different numbers of constellation points with different amplitude distributions bring differences in the amplitude trajectory jump positions of the signals. Using this feature, the two signal categories can be recognized; First, sort the information amplitude |x m (n)| output by the phase discriminator to obtain the information sequence D m (n), and calculate the curvature C m (n) of the information sequence D m (n): Among them, m = 3 and m = 4 respectively represent the data results of the signal passing through two different receivers; Select the trajectory jump position L for distinguishing signals j , and statistically calculate the curvature distribution C m (n) Take the logarithm of the variance of P data points near the trajectory jump position L j as the characteristic parameter μ for signal recognition k , and the expression is as follows: The discrimination between 16QAM and 16APSK uses L j = 3N / 4 to obtain the variance μ1 of the amplitude trajectory jump interval; the discrimination between 32QAM and 32APSK uses L j = 3N / 8 to obtain the variance μ2 of the amplitude trajectory jump interval. Since the QAM signal has an amplitude jump at the trajectory jump position L j while the APSK signal has a constant amplitude at the trajectory jump position L j the characteristic parameter μ of the QAM signal k is significantly greater than the characteristic parameter μ of the APSK signal k , and the two sets are identified respectively according to this characteristic.

Images