Decision tree-based modulation identification method, storage medium and equipment

Abstract

The embodiment of the invention provides a modulation recognition method based on a decision tree, a storage medium and equipment, and the method comprises the steps: carrying out the signal preprocessing, including down-conversion, filtering, sampling and normalization, of input data, carrying out the spectrum feature extraction and statistical feature extraction of the preprocessed data, and carrying out the recognition of the data. And realizing modulation signal type identification based on multi-stage logic judgment of a preset threshold value. According to the embodiment of the invention, the high-order cumulant is combined with other characteristic parameters to perform advantage complementation among the characteristic parameters, and rapid and efficient modulation identification of various signals under a low signal-to-noise ratio is realized.

Description

Technical Field

[0001] This invention relates to the field of signal modulation recognition technology, and in particular to a modulation recognition method, storage medium, and device based on decision trees. Background Technology

[0002] Extracting time-domain features to identify modulated signals was an early method. However, since this method processes signals in the time domain, the signals are less resistant to noise and are easily interfered with. When there are many types of signals to be identified, the signal features may overlap, which will greatly reduce the recognition rate.

[0003] Identifying modulated signals from the frequency spectrum utilizes the signal's modulation characteristics in the frequency domain (such as spectral peaks and spectral lines near the carrier wave). Compared to time-domain characteristics, frequency-domain characteristics offer better stability, and Fourier transform is typically used to convert the signal from the time domain to the frequency domain. There are many types of frequency-domain spectra, including cyclic spectra, spectral frequencies, higher-order spectra, and power spectra. Different spectral characteristics can effectively distinguish different modulation schemes. However, because signal processing occurs in the frequency domain, the conversion speed is slow, the complexity is high, and real-time performance is poor. Modulation identification based on higher-order cumulant features achieves good results in Gaussian white noise environments, but the accuracy of signal identification drops significantly in other types of noise environments; that is, identification algorithms based on higher-order cumulant features have certain limitations. Therefore, there is an urgent need for an identification method that can accurately and quickly identify various types of modulated signals. Summary of the Invention

[0004] In view of the above-mentioned problems in the prior art, the present invention provides a modulation recognition method, storage medium and device based on decision tree to solve the technical problems of the limitations of existing recognition methods and insufficient recognition efficiency and accuracy.

[0005] This invention provides a modulation recognition method based on a decision tree, comprising the following steps:

[0006] Step S1: Calculate the envelope square spectrum of the input data and determine whether there is a symbol rate spectral line in the envelope square spectrum. If yes, proceed to step S2; otherwise, proceed to step S11.

[0007] Step S2: Calculate the spectrum of the input data and determine whether there is a single discrete spectral line in the spectrum. If so, the input data is an ASK modulation type signal; otherwise, proceed to step S3.

[0008] Step S3: Calculate the octave spectrum of the input data and determine whether there is a single discrete spectral line in the octave spectrum. If so, proceed to step S4; otherwise, proceed to step S5.

[0009] Step S4: Calculate the quadruple spectrum and double spectrum of the input data, and determine whether the peak value of the quadruple spectrum is greater than the peak value of the double spectrum than the set threshold. If it is greater, the input data is a UQPSK modulation type signal; otherwise, the input data is a BPSK modulation type signal.

[0010] Step S5: Determine whether there are two discrete spectral lines in the octave spectrum of the input data. If so, the input data is a Pi / 2BPSK modulation type signal; otherwise, proceed to step S6.

[0011] Step S6: Calculate the quadruple spectrum of the input data and determine whether there are discrete spectral lines in the quadruple spectrum. If yes, proceed to step S7; otherwise, proceed to step S10.

[0012] Step S7: Determine whether there is a single discrete spectral line in the quadruple spectrum of the input data. If yes, proceed to step S8; otherwise, the input data is a Pi / 4QPSK modulated signal.

[0013] Step S8: Calculate the higher-order cumulative amount of the input data and determine whether its higher-order cumulative amount is greater than the set threshold. If so, the input data is a QPSK modulation type signal; otherwise, proceed to step S9.

[0014] Step S9: Calculate the number of constellations in the input data. If the number of constellations is 16, the input data is a 16QAM modulation signal. If the number of constellations is 32, the input data is a 32QAM modulation signal. If the number of constellations is 64, the input data is a 64QAM modulation signal.

[0015] Step S10: Calculate the higher-order cumulative amount of the input data. If the higher-order cumulative amount is higher than the high-value threshold, the input data is an 8PSK modulation type signal. If the higher-order cumulative amount is between the high-value threshold and the low-value threshold, the input data is a 16PSK modulation type signal. If the higher-order cumulative amount is lower than the low-value threshold, the input data is a 32PSK modulation type signal.

[0016] Step S11: Calculate the double spectrum and quadruple spectrum of the input data. If the double spectrum is double-peaked and the quadruple spectrum is single-peaked, then the input data is an OQPSK modulated signal; otherwise, proceed to step S12.

[0017] Step S12: Calculate the spectrum of the input data and determine whether the spectrum has symmetry. If the symmetry coefficient is close to 1, the input data is an LSB modulation type signal. If the symmetry coefficient is close to -1, the input data is a USB modulation type signal. Otherwise, proceed to step S13.

[0018] Step S13: Calculate the maximum value of the normalized amplitude spectral density of the input data. If the maximum value of the normalized amplitude spectral density is greater than the set threshold, then the input data is an AM modulation type signal; otherwise, proceed to step S14.

[0019] Step S14: Determine whether there is a single discrete spectral line in the spectrum of the input data. If it exists, the input data is a CW modulation type signal; otherwise, proceed to step S15.

[0020] Step S15: Determine whether there are discrete spectral lines in the spectrum of the input data. If there are 2 discrete spectral lines, the input data is a 2FSK modulation type signal. If there are 4 discrete spectral lines, the input data is a 4FSK modulation type signal. Otherwise, proceed to step S16.

[0021] Step S16: Calculate the cyclic spectrum of the input data and determine whether there is a symbol rate spectral line in the cyclic spectrum. If not, the input data is an FM modulation type signal; otherwise, proceed to step S17.

[0022] Step S17: Calculate the octave spectrum of the input data and determine whether there are two discrete spectral lines in the octave spectrum. If not, the input data is a signal of other CPM modulation type; otherwise, proceed to step S18.

[0023] Step S18: Calculate the result of the input data after delay multiplication, and determine whether there is a symbol rate spectral line after delay multiplication. If there is, the input data is an MSK modulation type signal; otherwise, the input data is a GMSK modulation type signal.

[0024] In one embodiment, in step S1, the input data is preprocessed data.

[0025] In one embodiment, the preprocessing operations include downconversion, filtering, sampling, and normalization.

[0026] In one embodiment, the threshold is set based on the signal-to-noise ratio and statistical feature distribution of the input data.

[0027] In addition, embodiments of the present invention also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the modulation recognition method based on decision tree as described in any embodiment of the present invention.

[0028] In addition, embodiments of the present invention also provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the modulation recognition method based on decision tree as described in any embodiment of the present invention.

[0029] Compared with the prior art, the beneficial effects of the modulation recognition method, storage medium and device based on decision tree provided by the embodiments of the present invention are as follows: the embodiments of the present invention combine high-order cumulants with other feature parameters to achieve complementary advantages between feature parameters, thereby realizing fast and efficient modulation recognition of multiple signals under low signal-to-noise ratio. Attached Figure Description

[0030] Figure 1 Part 1 of a flowchart illustrating a modulation recognition method based on a decision tree, provided as an embodiment of the present invention;

[0031] Figure 2 Part II is a flowchart illustrating a modulation recognition method based on a decision tree, provided as an embodiment of the present invention.

[0032] Figure 3 A schematic diagram of the envelope spectrum of a BPSK signal involved in a modulation recognition method based on a decision tree provided in an embodiment of the present invention;

[0033] Figure 4 A schematic diagram of the OQPSK signal envelope spectrum involved in a modulation recognition method based on a decision tree provided in an embodiment of the present invention;

[0034] Figure 5 A schematic diagram of the octave spectrum of a BPSK signal involved in a modulation identification method based on a decision tree provided in an embodiment of the present invention;

[0035] Figure 6 A schematic diagram of the 4-fold spectrum of a PI / 4QPSK signal involved in a modulation recognition method based on a decision tree provided in an embodiment of the present invention;

[0036] Figure 7 A schematic diagram of the 4-fold spectrum of an 8PSK signal involved in a modulation recognition method based on a decision tree provided in an embodiment of the present invention;

[0037] Figure 8 The following is a schematic diagram of the data flow for a modulation recognition method based on a decision tree provided in an embodiment of the present invention. Detailed Implementation

[0038] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0039] Various embodiments and features of this application are described herein with reference to the accompanying drawings.

[0040] These and other features of this application will become apparent from the following description of preferred forms of embodiments given as non-limiting examples, with reference to the accompanying drawings.

[0041] It should also be understood that although this application has been described with reference to some specific examples, those skilled in the art can certainly implement many other equivalent forms of this application, which have the features described in the claims and are therefore all within the scope of protection defined herein.

[0042] The above and other aspects, features and advantages of this application will become more apparent when taken in conjunction with the accompanying drawings and in view of the following detailed description.

[0043] Specific embodiments of this application are described below with reference to the accompanying drawings; however, it should be understood that the claimed embodiments are merely examples of this application, which can be implemented in various ways. Well-known and / or repeated functions and structures are not described in detail to ascertain the true intent based on the user's historical operations, and to avoid unnecessary or redundant details that would obscure this application. Therefore, the specific structural and functional details claimed herein are not intended to be limiting, but merely serve as the basis and representative basis for the claims to teach those skilled in the art to use this application in various ways with substantially any suitable detailed structure.

[0044] This specification may use the phrases “in one embodiment,” “in another embodiment,” “in yet another embodiment,” or “in other embodiments,” all of which may refer to one or more of the same or different embodiments according to this application.

[0045] The principles and features of the present invention are described below with reference to the accompanying drawings. The embodiments described are for illustrative purposes only and are not intended to limit the scope of the invention. The following description, in conjunction with... Figure 1-8 The preferred embodiments of the present invention will be described in further detail below:

[0046] like Figure 1-8 As shown, this embodiment of the invention provides a modulation recognition method based on a decision tree, comprising the following steps:

[0047] Step S1: Calculate the envelope square spectrum of the input data and determine whether there is a symbol rate spectral line in the envelope square spectrum. If yes, proceed to step S2; otherwise, proceed to step S11.

[0048] Step S2: Calculate the spectrum of the input data and determine whether there is a single discrete spectral line in the spectrum. If so, the input data is an ASK modulation type signal; otherwise, proceed to step S3.

[0049] Step S3: Calculate the octave spectrum of the input data and determine whether there is a single discrete spectral line in the octave spectrum. If so, proceed to step S4; otherwise, proceed to step S5.

[0050] Step S4: Calculate the quadruple spectrum and double spectrum of the input data, and determine whether the peak value of the quadruple spectrum is greater than the peak value of the double spectrum than the set threshold. If it is greater, the input data is a UQPSK modulation type signal; otherwise, the input data is a BPSK modulation type signal.

[0051] Step S5: Determine whether there are two discrete spectral lines in the octave spectrum of the input data. If so, the input data is a Pi / 2BPSK (π / 2-BPSK) modulation type signal; otherwise, proceed to step S6.

[0052] Step S6: Calculate the quadruple spectrum of the input data and determine whether there are discrete spectral lines in the quadruple spectrum. If yes, proceed to step S7; otherwise, proceed to step S10.

[0053] Step S7: Determine whether there is a single discrete spectral line in the quadruple spectrum of the input data. If yes, proceed to step S8; otherwise, the input data is a Pi / 4QPSK (π / 4-QPSK) modulation type signal.

[0054] Step S8: Calculate the higher-order cumulative amount of the input data and determine whether its higher-order cumulative amount is greater than the set threshold. If so, the input data is a QPSK modulation type signal; otherwise, proceed to step S9.

[0055] Step S9: Calculate the number of constellations in the input data. If the number of constellations is 16, the input data is a 16QAM modulation signal. If the number of constellations is 32, the input data is a 32QAM modulation signal. If the number of constellations is 64, the input data is a 64QAM modulation signal.

[0056] Step S10: Calculate the higher-order cumulative amount of the input data. If the higher-order cumulative amount is higher than the high-value threshold, the input data is an 8PSK modulation type signal. If the higher-order cumulative amount is between the high-value threshold and the low-value threshold, the input data is a 16PSK modulation type signal. If the higher-order cumulative amount is lower than the low-value threshold, the input data is a 32PSK modulation type signal.

[0057] Step S11: Calculate the double spectrum and quadruple spectrum of the input data. If the double spectrum is double-peaked and the quadruple spectrum is single-peaked, then the input data is an OQPSK modulated signal; otherwise, proceed to step S12.

[0058] Step S12: Calculate the spectrum of the input data and determine whether the spectrum has symmetry. If the symmetry coefficient is close to 1, the input data is an LSB modulation type signal. If the symmetry coefficient is close to -1, the input data is a USB modulation type signal. Generally, if the difference between the symmetry coefficient and 1 or -1 is 20% of the standard value, it is considered close. Otherwise, proceed to step S13.

[0059] Step S13: Calculate the maximum value of the normalized amplitude spectral density of the input data. If the maximum value of the normalized amplitude spectral density is greater than the set threshold, then the input data is an AM modulation type signal; otherwise, proceed to step S14.

[0060] Step S14: Determine whether there is a single discrete spectral line in the spectrum of the input data. If it exists, the input data is a CW modulation type signal; otherwise, proceed to step S15.

[0061] Step S15: Determine whether there are discrete spectral lines in the spectrum of the input data. If there are 2 discrete spectral lines, the input data is a 2FSK modulation type signal. If there are 4 discrete spectral lines, the input data is a 4FSK modulation type signal. Otherwise, proceed to step S16.

[0062] Step S16: Calculate the cyclic spectrum of the input data and determine whether there is a symbol rate spectral line in the cyclic spectrum. If not, the input data is an FM modulation type signal; otherwise, proceed to step S17.

[0063] Step S17: Calculate the octave spectrum of the input data and determine whether there are two discrete spectral lines in the octave spectrum. If not, the input data is a signal of other CPM modulation type (CPM is continuous phase modulation, MSK and GMSK are two types of CPM, and other CPM here refers to CPM modulation signals other than MSK and GMSK). Otherwise, proceed to step S18.

[0064] Step S18: Calculate the result of the input data after delay multiplication, and determine whether there is a symbol rate spectral line after delay multiplication. If there is, the input data is an MSK modulation type signal; otherwise, the input data is a GMSK modulation type signal.

[0065] In one embodiment, in step S1, the input data is preprocessed data.

[0066] In one embodiment, the preprocessing operations include downconversion, filtering, sampling, and normalization.

[0067] In one embodiment, the threshold is set based on the signal-to-noise ratio and statistical feature distribution of the input data.

[0068] In addition, embodiments of the present invention also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the modulation recognition method based on decision tree as described in any embodiment of the present invention.

[0069] In addition, embodiments of the present invention also provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the modulation recognition method based on decision tree as described in any embodiment of the present invention.

[0070] To correctly identify modulation types such as AM, FM, USB, LSB, CW, ASK, 2FSK, 4FSK, CPM, BPSK, QPSK, OQPSK, pi / 4QPSK (π / 4-QPSK), 8PSK, 16APSK, 32APSK, MSK, GMSK, UQPSK, pi / 2BPSK (π / 2-BPSK), 16QAM, 32QAM, and 64QAM under low signal-to-noise ratio conditions, this invention employs a modulation identification algorithm based on parameter extraction and pattern recognition. The algorithm is optimized in two ways: firstly, by optimizing the identification parameters, and secondly, by optimizing the identification strategy.

[0071] Decision parameter design

[0072] The modulation recognition algorithm used in this embodiment of the invention is a parameter extraction algorithm selected and optimized based on previous research, considering the following parameters for identifying the modulation signal:

[0073] a) R parameters

[0074] The R parameter reflects the degree of change in the signal envelope. This parameter can be used to better distinguish between modulated signals with amplitude changes and modulated signals with small amplitude changes.

[0075] b)F 2CW parameter

[0076] F 2CW The parameter reflects the energy of twice the carrier component contained in the square spectrum of the signal, and is mainly used to distinguish BPSK signals from other digital modulation signals.

[0077] c) Peak energy of four times the carrier wave F 4CW

[0078] F 4CW The parameter reflects the energy of the four-fold carrier component contained in the fourth power spectrum of the signal, and is mainly used to distinguish QPSK and OQPSK signals from 8PSK and QAM signals.

[0079] d) Symbol rate energy peak F r

[0080] F r The parameter reflects the energy of the symbol rate contained in the square spectrum of the signal and is mainly used to distinguish OQPSK signals from other digital modulation signals.

[0081] e) Analyzing the higher-order cumulant Cum of the signal 21 Cum 42 Cum 40 Cum 60 Cum 63 .

[0082] Since the higher-order cumulants of a Gaussian random process are always zero, if a non-Gaussian signal is observed in an independent additive Gaussian noise environment, the higher-order cumulants of the observed process will be identical to those of the non-Gaussian signal process. Therefore, using higher-order cumulants as an analytical tool can theoretically completely suppress the influence of Gaussian noise, making them an effective tool for extracting non-Gaussian signals against a Gaussian noise background.

[0083] Decision-making process design

[0084] The conventional processing flow of decision tree-based signal modulation technology is as follows: Figure 1-2 As shown.

[0085] Decision algorithm design

[0086] OQPSK signal recognition

[0087] like Figure 3-4 As shown, the envelope square spectrum of a digitally modulated signal generally has a distinct symbol rate line, while the envelope fluctuation of an OQPSK signal is not obvious and does not have this characteristic.

[0088] Therefore, setting parameters

[0089]

[0090] Among them, max1 and max2 are the maximum and second largest values ​​in the envelope square spectrum, respectively, thus identifying the OQPSK signal.

[0091] BPSK signal recognition

[0092] like Figure 5 As shown, unlike other signals, the BPSK signal has a single spectral line in its octave spectrum, and this parameter characteristic can be used to identify BPSK.

[0093] Design parameters

[0094]

[0095] Where max1 and max2 are the maximum and second maximum values ​​of the signal's octave spectrum, respectively. These values ​​are relatively small for BPSK signals and can be identified.

[0096] PI / 4QPSK (π / 4-QPSK) signal identification

[0097] Similar to the spectral characteristics of signals, the fourth-order spectrum of a QPSK signal has a single discrete spectral line, the fourth-order spectrum of a PI / 4QPSK (π / 4-QPSK) signal has two discrete spectral lines, and other signals do not have discrete spectral lines, such as... Figure 6-7 As shown.

[0098] Design parameters

[0099]

[0100] Where max1, max2, max3, and max4 are the first, second, third, and fourth largest values ​​of the fourfold spectrum of the signal, respectively.

[0101] This parameter is smaller for QPSK and PI / 4QPSK (π / 4-QPSK) signals, and larger for other signals.

[0102] Similarly, parameters can be designed.

[0103]

[0104] Where max1 and max2 are the maximum and second maximum values ​​of the signal's four-fold spectrum, respectively. This parameter is relatively large for PI / 4QPSK (π / 4-QPSK) signals, while it is relatively small for QPSK and 16QAM signals, which have only a single spectral line.

[0105] QPSK and 16QAM signal identification

[0106] Since the higher-order cumulants of a Gaussian random process are always zero, if a non-Gaussian signal is observed in an independent additive Gaussian noise environment, the higher-order cumulants of the observed process will be identical to those of the non-Gaussian signal process. Therefore, using higher-order cumulants as an analytical tool can theoretically completely suppress the influence of Gaussian noise, making them an effective tool for extracting non-Gaussian signals against a Gaussian noise background.

[0107] For a complex-valued stationary signal X(t), its second-order, fourth-order, and sixth-order cumulants are defined as follows, depending on the position of the conjugate term:

[0108] C 20 =cum(X,X)=M 20

[0109] C 21 =cum(X,X) * ) = M 21

[0110]

[0111] C 41 =cum(X,X,X,X) * ) = M41 -3M 21 M 20

[0112]

[0113] Where: M pq It is the p-th order mixing moment of X(t), defined as M pq =E[X(t)] p-q ·(X * (t)) q ]

[0114] The higher-order cumulative quantities of commonly used signals are represented in Table 1.

[0115] To differentiate between QPSK and 16QAM, design parameters

[0116]

[0117] QPSK has a larger p7 value, while 16QAM has a smaller p7 value. With proper design of the discrimination threshold, the two signals can be identified and distinguished.

[0118] Table 1 Commonly Used Signal Higher-Order Cumulatives

[0119] signal type <![CDATA[|C 20 |]]> <![CDATA[|C 21 |]]> <![CDATA[|C 40 |]]> <![CDATA[|C 41 |]]> <![CDATA[|C 42 |]]> <![CDATA[|C 60 |]]> <![CDATA[|C 61 |]]> <![CDATA[|C 80 |]]> 2ASK E E <![CDATA[2E 2 ]]> <![CDATA[2E 2 ]]> <![CDATA[2E 2 ]]> <![CDATA[16E 3 ]]> <![CDATA[13E 3 ]]> <![CDATA[272E 4 ]]> 4ASK E E <![CDATA[1.36E 2 ]]> <![CDATA[1.36E 2 ]]> <![CDATA[1.36E 2 ]]> <![CDATA[8.32E 3 ]]> <![CDATA[9.16E 3 ]]> <![CDATA[111.85E 4 ]]> 8ASK E E <![CDATA[1.24E 2 ]]> <![CDATA[1.24E 2 ]]> <![CDATA[1.24E 2 ]]> <![CDATA[7.19E 3 ]]> <![CDATA[8.76E 3 ]]> <![CDATA[92.02E 4 ]]> 2FSK 0 E 0 0 <![CDATA[E 2 ]]> 0 <![CDATA[4E 3 ]]> 0 4FSK 0 E 0 0 <![CDATA[E 2 ]]> 0 <![CDATA[4E 3 ]]> 0 8FSK 0 E 0 0 <![CDATA[E 2 ]]> 0 <![CDATA[4E 3 ]]> 0 BPSK E E <![CDATA[2E 2 ]]> <![CDATA[2E 2 ]]> <![CDATA[2E 2 ]]> <![CDATA[16E 3 ]]> <![CDATA[13E 3 ]]> <![CDATA[272E 4 ]]> QPSK 0 E <![CDATA[E 2 ]]> 0 <![CDATA[E 2 ]]> 0 <![CDATA[4E 3 ]]> <![CDATA[34E 4 ]]> 8PSK 0 E 0 0 <![CDATA[E 2 ]]> 0 <![CDATA[4E 3 ]]> <![CDATA[E 4 ]]> 16QAM 0 E <![CDATA[0.68E 2 ]]> 0 <![CDATA[0.68E 2 ]]> 0 <![CDATA[2.08E 3 ]]> <![CDATA[13.98E 4 ]]> 32QAM 0 E <![CDATA[0.19E 2 ]]> 0 <![CDATA[0.69E 2 ]]> 0 <![CDATA[2.11E 3 ]]> <![CDATA[1.99E 4 ]]> 64QAM 0 E <![CDATA[0.619E 2 ]]> 0 <![CDATA[0.619E 2 ]]> 0 <![CDATA[1.797E 3 ]]> <![CDATA[11.50E 4 ]]>

[0120] 8PSK, 16APSK, and 32APSK signal recognition

[0121] Similarly, these three types of signals are identified using higher-order cumulants, and the higher-order cumulants C are calculated for each. 42 C 21 C 63 Design parameters

[0122]

[0123] The p8 parameter is the largest for 8PSK, followed by 16APSK, and the smallest for 32APSK. This parameter can be used to distinguish them.

[0124] This invention aims to describe in detail the system architecture and algorithm implementation details of "decision tree-based modulation recognition technology". This system is mainly used to automatically identify the modulation type of received communication signals in low signal-to-noise ratio (Low SNR) environments.

[0125] Identify target

[0126] This invention supports automatic classification of the following modulation types:

[0127] a) PSK type: BPSK, QPSK, OQPSK, π / 4-QPSK, 8PSK

[0128] b) QAM type: 16QAM, 32QAM

[0129] c) APSK type: 16 APSK, 32 APSK

[0130] Technical route overview

[0131] This invention employs a hierarchical decision tree architecture. Unlike traditional single-feature classification, this invention combines spectral correlation features (for distinguishing BPSK, OQPSK, and π / 4-QPSK) and higher-order cumulant features (for distinguishing higher-order modulation QPSK, 8PSK, QAM, and APSK) to improve the recognition rate while ensuring computational efficiency.

[0132] System Architecture of the Embodiments of the Invention

[0133] Logical architecture diagram

[0134] The system processing flow is divided into four main stages:

[0135] a) Signal preprocessing module: down-conversion, filtering, sampling, and normalization.

[0136] b) Spectral feature extraction module: FFT transform, calculate envelope spectrum, square spectrum, fourth power spectrum, double spectrum, second spectrum and fourth spectrum.

[0137] c) Statistical feature extraction module: Calculates higher-order moments and higher-order cumulants.

[0138] d) Decision tree classifier: multi-level logical decision based on preset thresholds.

[0139] Detailed Algorithm Design

[0140] 1 Preprocessing Module

[0141] a) Input: Analog / digital intermediate frequency signal r(t)

[0142] b) Processing:

[0143] 1) Quadrature downconversion: shifts the signal to zero intermediate frequency.

[0144] 2) Low-pass filter: Filters out harmonic components and out-of-band noise.

[0145] 3) Power normalization: Normalize the signal power to 1 to eliminate the influence of signal strength on the amplitude of characteristic values.

[0146]

[0147] c) Output: Normalized complex baseband sequence y(n)

[0148] 2 Feature Extraction Module

[0149] 2.1 Spectral Feature Extraction

[0150] This section mainly utilizes the spectral characteristics of different modulation signals after nonlinear transformation.

[0151] 2.1.1 Envelope Spectrum Features (for OQPSK)

[0152] a) Principle: OQPSK has smaller envelope fluctuations and a specific pattern due to the interleaving of I / Q branches, which is different from other PSK / QAM.

[0153] b) Calculation:

[0154] 1) Calculate the envelope E(n) = |y(n)|.

[0155] 2) Perform FFT to obtain S env (f).

[0156] 3) Extract the feature parameter R (specifically, the energy percentage of a specific frequency component of the envelope spectrum).

[0157] 4.2.1.2 Squared Spectrum Characteristics (for BPSK)

[0158] a) Principle: After the BPSK signal is squared, e j(ωt+φ) →e j(2ωt+2φ) Phase doubling eliminates modulation information at 2f c A strong single-frequency component is formed at this location.

[0159] b) Calculation:

[0160] 1) Calculate y 2 (n).

[0161] 2) Perform FFT to obtain the squared spectrum S sq (f).

[0162] 3) Calculation characteristic: the ratio of the maximum spectral peak to the average spectral base.

[0163]

[0164] 4.2.1.3 Fourth-order spectral characteristics (for PI / 4-QPSK)

[0165] c) Principle: After the fourth power of ordinary M-PSK, at 4f... c The fourth-order spectrum is a single peak at 4f; due to the special phase rotation, the fourth-order spectrum of π / 4-QPSK has a single peak at 4f. c It splits into two peaks.

[0166] d) Calculation:

[0167] 1) Calculate y 4 (n).

[0168] 2) Perform FFT to obtain S quad (f).

[0169] 3) Find the largest peak P1 and the second largest peak P2.

[0170] 4) Calculate the ratio Ar = P2 / P1.

[0171] 4.2.2 Characteristics of higher-order cumulants (for higher-order modulation)

[0172] When spectral features cannot distinguish between different types (such as QPSK, 8PSK, 16QAM, etc.), statistical features are used.

[0173] a) Definition of mixing moment: M pq =E[y(n)] p-q (y * (n)) q ]

[0174] 1)M 20 =E[y 2 ],M 21 =E[|y| 2 ]

[0175] 2)M 40 =E[y 4 ],M 41 =E[y 3 y * ],M 42 =E[|y| 4 ]

[0176] 3)M 63 =E[|y| 6 ]

[0177] b) Cumulative amount calculation formula:

[0178] 1) Fourth-order cumulant C 42 :C 42 =M 42 -|M 20 | 2 -2(M 21 ) 2

[0179] 2) Sixth-order cumulant C 63 :C 63 =M 63 -9C 42 M 21 -6(M 21 ) 3

[0180] c) Constructing classification features:

[0181] 1) Feature I (Cr): Used to distinguish between QPSK and 16QAM. Note: In theory, QPSK is a constant envelope, so this value is relatively large; 16QAM is a multi-level signal, so this value is relatively small.

[0182] 2) Feature II (Cum): Used to distinguish between 8PSK, 16APSK, and 32APSK. A combination feature needs to be selected based on the simulation statistical distribution (e.g., directly using |Cum|). 42 |or|C 63 | normalized value).

[0183] 4.3 Decision Tree Classifier Design

[0184] The classifier consists of a series of logical decision units. A threshold set T = T needs to be pre-determined through training / statistical analysis on a large number of samples. R ,T F2 ,T Ar ,T Cr ,T Cum1 ,T Cum2 .

[0185] 4.3.1 Executing pseudocode:

[0186]

[0187] 5. Module Interface Definition

[0188] 5.1 Input Interface

[0189] a) Data Stream: complex double[](I / Q data streams)

[0190] b) Parameters:

[0191] 1) fs (sampling rate)

[0192] 2) N (data block length, recommended 2048 or 4096 points)

[0193] 4.2 Threshold Configuration Interface

[0194] The system should provide a config.json file with a loading threshold so that it can be fine-tuned based on the actual channel environment (SNR level).

[0195]

[0196]

[0197] 5.2 Output Interface

[0198] c)Result Code:Enum ModulationType

[0199] d) Confidence: float (Calculates confidence based on the distance between the feature value and the threshold)

[0200] 6 Performance Requirements and Testing

[0201] 6.1 Performance Indicators

[0202] a) SNR adaptability: When SNR≥5dB, the overall recognition rate should be ≥90%.

[0203] b) Real-time performance: Single recognition and processing latency <100ms (based on a general PC platform).

[0204] 6.2 Key Test Points

[0205] a) Confusion matrix test: Focus on the false positive rate of QPSK and 16QAM, and the ability of 16APSK and 16QAM to distinguish between them.

[0206] b) Frequency offset robustness: The stability of the squared spectrum characteristics is tested in the presence of carrier frequency offset (Δf).

[0207] c) Threshold calibration: Statistical feature distribution needs to be performed under different SNRs (0dB, 5dB, 10dB, 15dB) to determine the optimal threshold boundary.

[0208] The above embodiments are merely exemplary embodiments of the present invention and are not intended to limit the present invention. The scope of protection of the present invention is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to the present invention within its spirit and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of the present invention.

Claims

1. A modulation recognition method based on decision trees, characterized in that, Includes the following steps: Step S1: Calculate the envelope square spectrum of the input data and determine whether there is a symbol rate spectral line in the envelope square spectrum. If yes, proceed to step S2; otherwise, proceed to step S11. Step S2: Calculate the spectrum of the input data and determine whether there is a single discrete spectral line in the spectrum. If so, the input data is an ASK modulation type signal; otherwise, proceed to step S3. Step S3: Calculate the octave spectrum of the input data and determine whether there is a single discrete spectral line in the octave spectrum. If so, proceed to step S4; otherwise, proceed to step S5. Step S4: Calculate the quadruple spectrum and double spectrum of the input data, and determine whether the peak value of the quadruple spectrum is greater than the peak value of the double spectrum than the set threshold. If it is greater, the input data is a UQPSK modulation type signal; otherwise, the input data is a BPSK modulation type signal. Step S5: Determine whether there are two discrete spectral lines in the octave spectrum of the input data. If so, the input data is a Pi / 2BPSK modulation type signal; otherwise, proceed to step S6. Step S6: Calculate the quadruple spectrum of the input data and determine whether there are discrete spectral lines in the quadruple spectrum. If yes, proceed to step S7; otherwise, proceed to step S10. Step S7: Determine whether there is a single discrete spectral line in the quadruple spectrum of the input data. If yes, proceed to step S8; otherwise, the input data is a Pi / 4QPSK modulated signal. Step S8: Calculate the higher-order cumulative amount of the input data and determine whether its higher-order cumulative amount is greater than the set threshold. If so, the input data is a QPSK modulation type signal; otherwise, proceed to step S9. Step S9: Calculate the number of constellations in the input data. If the number of constellations is 16, the input data is a 16QAM modulation signal. If the number of constellations is 32, the input data is a 32QAM modulation signal. If the number of constellations is 64, the input data is a 64QAM modulation signal. Step S10: Calculate the higher-order cumulative amount of the input data. If the higher-order cumulative amount is higher than the high-value threshold, the input data is an 8PSK modulation type signal. If the higher-order cumulative amount is between the high-value threshold and the low-value threshold, the input data is a 16PSK modulation type signal. If the higher-order cumulative amount is lower than the low-value threshold, the input data is a 32PSK modulation type signal. Step S11: Calculate the double spectrum and quadruple spectrum of the input data. If the double spectrum is double-peaked and the quadruple spectrum is single-peaked, then the input data is an OQPSK modulated signal; otherwise, proceed to step S12. Step S12: Calculate the spectrum of the input data and determine whether the spectrum has symmetry. If the symmetry coefficient is close to 1, the input data is an LSB modulation type signal. If the symmetry coefficient is close to -1, the input data is a USB modulation type signal. Otherwise, proceed to step S13. Step S13: Calculate the maximum value of the normalized amplitude spectral density of the input data. If the maximum value of the normalized amplitude spectral density is greater than the set threshold, then the input data is an AM modulation type signal; otherwise, proceed to step S14. Step S14: Determine whether there is a single discrete spectral line in the spectrum of the input data. If it exists, the input data is a CW modulation type signal; otherwise, proceed to step S15. Step S15: Determine whether there are discrete spectral lines in the spectrum of the input data. If there are 2 discrete spectral lines, the input data is a 2FSK modulation type signal. If there are 4 discrete spectral lines, the input data is a 4FSK modulation type signal. Otherwise, proceed to step S16. Step S16: Calculate the cyclic spectrum of the input data and determine whether there is a symbol rate spectral line in the cyclic spectrum. If not, the input data is an FM modulation type signal; otherwise, proceed to step S17. Step S17: Calculate the octave spectrum of the input data and determine whether there are two discrete spectral lines in the octave spectrum. If not, the input data is a signal of other CPM modulation type; otherwise, proceed to step S18. Step S18: Calculate the result of the input data after delay multiplication, and determine whether there is a symbol rate spectral line after delay multiplication. If there is, the input data is an MSK modulation type signal; otherwise, the input data is a GMSK modulation type signal.

2. The modulation recognition method based on decision tree according to claim 1, characterized in that: In step S1, the input data is the preprocessed data.

3. The modulation recognition method based on decision tree according to claim 2, characterized in that: The preprocessing operations include down-conversion, filtering, sampling, and normalization.

4. The modulation recognition method based on decision tree according to claim 2, characterized in that: The threshold is set based on the signal-to-noise ratio and statistical feature distribution of the input data.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by the processor, it implements the modulation recognition method based on the decision tree as described in any one of claims 1-4.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the modulation recognition method based on decision tree as described in any one of claims 1-4.

Images