CN112800863A - Time-frequency feature extraction method and system for communication signal modulation pattern recognition - Google Patents
Time-frequency feature extraction method and system for communication signal modulation pattern recognition Download PDFInfo
- Publication number
- CN112800863A CN112800863A CN202110028826.9A CN202110028826A CN112800863A CN 112800863 A CN112800863 A CN 112800863A CN 202110028826 A CN202110028826 A CN 202110028826A CN 112800863 A CN112800863 A CN 112800863A
- Authority
- CN
- China
- Prior art keywords
- time
- frequency
- signal
- decomposition
- components
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Noise Elimination (AREA)
Abstract
The invention relates to a time-frequency characteristic extraction method and a time-frequency characteristic extraction system for communication signal modulation mode identification, which are used for acquiring an original communication signal and carrying out zero equalization; acquiring a three-order cumulant slice spectrum of the communication signal after zero equalization; judging the number of main frequency components of the communication signal after zero equalization according to the third-order cumulant slicing spectrum, thereby determining the number of decomposition layers and the optimal decomposition layer of VMD decomposition; performing VMD decomposition on the original communication signal according to the optimal decomposition layer number to obtain intrinsic mode function components, and calculating zero delay fourth-order cumulant for each intrinsic mode function component; judging whether each intrinsic mode function component is a useful signal component according to a judgment criterion based on the high-order cumulant; and respectively carrying out ZAM time-frequency transformation on all useful signal components, and overlapping the transformed time-frequency signal components to obtain the time-frequency characteristics of the original communication signals. The invention combines VMD decomposition and ZAM time frequency analysis method, and improves the accuracy of time frequency feature extraction.
Description
Technical Field
The invention relates to the technical field of signal time-frequency feature extraction, in particular to a time-frequency feature extraction method and a time-frequency feature extraction system for communication signal modulation mode identification.
Background
Modulation identification techniques are one of the important issues to be addressed in the field of communications. With the rapid development of wireless communication technology, especially with the application of 5G mobile communication and its next generation communication system, the problem of insufficient wireless spectrum resources is increasingly prominent. The method enhances the detection and management of wireless communication signals and improves the utilization rate of frequency spectrum, and is an important work of national radio management departments, and the modulation identification of the communication signals is indispensable. In the military field, the key problems of communication electronic warfare, electronic information interception, radar reconnaissance and electronic interference are all the premise of identifying the modulation mode of an intercepted signal. Therefore, modulation recognition technology plays an important role in both the civilian and military fields.
Currently, the most widely used modulation recognition technology is a pattern recognition method based on feature extraction. The recognition rate of the method depends mainly on the noise immunity of the extracted signal features. Compared with other classification characteristics, the communication signal time-frequency characteristics have stronger noise robustness and are important classification characteristics in modulation identification. Under the condition of low signal-to-noise ratio, most of the existing time-frequency analysis methods face the problems of time-frequency graph blurring and cross term interference, and the time-frequency feature extraction effect is seriously influenced. Smooth pseudo Wigner-Ville distribution (SPWVD), pseudo Wigner-Ville distribution (PWVD), Choi-Williams distribution (CWD) and the like solve the problem of cross term interference by designing different kernel functions, but reduce time-frequency aggregation while inhibiting cross terms. The method of frequency analysis when Empirical Mode Decomposition (EMD) is combined can inhibit cross terms and simultaneously keep time-frequency aggregation, but the EMD algorithm has defects of mode aliasing, end effect, pseudo components and the like. Therefore, under the condition of low signal-to-noise ratio, how to effectively inhibit cross term interference while maintaining time-frequency aggregation performance is an urgent problem to be solved, so that the extracted time-frequency features are more accurate.
Aiming at the problems, the invention provides a time-frequency feature extraction method and a time-frequency feature extraction system for communication signal modulation mode identification, so as to solve the technical problems of noise interference of communication signals under the condition of low signal-to-noise ratio and cross term interference in time-frequency analysis.
Disclosure of Invention
The invention aims to provide a time-frequency feature extraction method and a time-frequency feature extraction system for communication signal modulation pattern recognition. The VMD decomposition algorithm can inhibit the generation of cross terms in the time-frequency analysis process, solve the problem that cross term inhibition and time-frequency aggregation are contradictory, and introduce ZAM time-frequency transformation to further inhibit noise and cross terms, so that time-frequency features with strong noise robustness can be extracted, the accuracy of time-frequency feature extraction is improved, and the method is applied to the identification of a communication signal modulation mode and can also improve the accuracy of modulation identification.
In order to achieve the purpose, the invention provides the following scheme:
a time-frequency feature extraction method for communication signal modulation pattern recognition comprises the following steps:
acquiring an original communication signal, and carrying out zero equalization on the original communication signal;
acquiring a three-order cumulant slice spectrum of the communication signal after zero equalization;
judging the number of main frequency components of the communication signal after zero equalization according to the third-order cumulant slice spectrum, and determining the number of decomposition layers of VMD decomposition based on the number of the main frequency components; determining an optimal number of decomposition layers k from the number of decomposition layers0；
According to the optimal decomposition layer number k0Performing VMD decomposition on the original communication signal to obtain k0The intrinsic mode function components are respectively calculated to obtain the zero delay fourth-order cumulant of each intrinsic mode function component;
according to a judgment criterion based on high-order cumulant, judging useful signal components and noise components for each intrinsic mode function component to obtain intrinsic mode function components judged as useful signal components, and recording the intrinsic mode function components as the useful signal components;
respectively carrying out ZAM time-frequency transformation on all the useful signal components to obtain transformed time-frequency signal components;
and superposing all the transformed time-frequency signal components to obtain the time-frequency characteristics of the original communication signals.
The invention also provides a time-frequency feature extraction system for communication signal modulation mode identification, which comprises the following steps:
the signal acquisition module is used for acquiring an original communication signal and carrying out zero equalization on the original communication signal;
the third-order cumulant slice spectrum acquisition module is used for acquiring a third-order cumulant slice spectrum of the communication signal after zero equalization;
the optimal decomposition layer number determining module is used for judging the number of main frequency components of the communication signal after zero equalization according to the third-order cumulant slice spectrum and determining the decomposition layer number of VMD decomposition based on the number of the main frequency components; determining an optimal number of decomposition layers k from the number of decomposition layers0；
A zero time delay fourth-order cumulant calculation module for calculating the optimal decomposition layer number k0Performing VMD decomposition on the original communication signal to obtain k0The intrinsic mode function components are respectively calculated to obtain the zero delay fourth-order cumulant of each intrinsic mode function component;
the useful signal component judging module is used for judging a useful signal component and a noise component for each intrinsic mode function component according to a judgment criterion based on the high-order cumulant to obtain the intrinsic mode function component which is judged as the useful signal component and is marked as the useful signal component;
the ZAM time-frequency transformation module is used for respectively carrying out ZAM time-frequency transformation on all the useful signal components to obtain transformed time-frequency signal components;
and the time-frequency characteristic acquisition module is used for superposing all the transformed time-frequency signal components to obtain the time-frequency characteristics of the original communication signals.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
the invention provides a time-frequency feature extraction method and a time-frequency feature extraction system for communication signal modulation mode identification, which are used for decomposing a signal into a plurality of single-frequency signals by adopting VMD (virtual machine format) decomposition, inhibiting the generation of cross terms in the time-frequency analysis process and solving the problem that the cross term inhibition and the time-frequency aggregation are contradictory. And then, the property that the high-order cumulant can inhibit Gaussian noise is utilized, the high-order cumulant is combined with the VMD decomposition, a judgment criterion of a useful signal component and a noise component is provided, and the problem of how to distinguish a signal mode from a noise mode in the VMD decomposition is effectively solved. The noise component is screened out, and the effect of denoising the signal under the condition of low signal-to-noise ratio is achieved. In the determination of the VMD decomposition layer number, the cross-correlation function of the reconstructed signal and the original signal with noise is solved by utilizing the characteristic that white noise is irrelevant to the signal, and the relevance between the reconstructed signal and the signal without noise is indirectly measured, so that the decomposition layer number when the relevance between the reconstructed signal and the signal without noise is maximum is selected as the optimal decomposition layer number of the VMD, and the problems that the VMD decomposition is blind when the decomposition layer number is preset and the theoretical basis is lacked are effectively solved. Finally, the VMD decomposition and the ZAM time-frequency analysis method are combined, the ZAM time-frequency analysis method is high in cross item inhibition capacity, and the cross item inhibition and time-frequency aggregation compromise performance is good. And performing ZAM time-frequency transformation on each useful signal component, thereby achieving the purpose of further suppressing noise and cross terms. The method effectively solves the technical problems of noise interference of non-stationary signals and cross term interference suppression under the condition of low signal-to-noise ratio, can extract the time-frequency characteristics with stronger noise robustness, lays a good foundation for the next step of identification by adopting a classifier, and ensures the accuracy of modulation identification better.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without inventive exercise.
Fig. 1 is a flowchart of a time-frequency feature extraction method for communication signal modulation pattern recognition according to embodiment 1 of the present invention;
fig. 2 is a time domain waveform diagram of a quaternary frequency shift keying signal without noise according to embodiment 1 of the present invention;
FIG. 3 is a time domain waveform diagram of a quaternary frequency shift keying signal with a signal-to-noise ratio of-7 dB according to embodiment 1 of the present invention;
FIG. 4 is a third order cumulant slice spectrogram of a quaternary frequency shift keying signal with a signal-to-noise ratio of-7 dB, provided in embodiment 1 of the present invention;
fig. 5 is a line graph of a relationship between a cross-correlation function value and a decomposition level of a reconstructed signal and an original signal obtained by decomposing the levels of different VMDs according to embodiment 1 of the present invention;
fig. 6 is a time domain waveform diagram of a reconstructed signal after denoising processing according to embodiment 1 of the present invention;
fig. 7 is a time domain waveform diagram of each signal component of the quaternary frequency shift keying signal with the signal-to-noise ratio of-7 dB obtained by VMD decomposition in the case of the optimal decomposition level number of 5 according to embodiment 1 of the present invention;
fig. 8 is a frequency domain waveform diagram of each signal component of the quaternary frequency shift keying signal with the signal-to-noise ratio of-7 dB obtained by VMD decomposition in the case of the optimal number of decomposition layers 5 according to embodiment 1 of the present invention;
fig. 9 is a ZAM transform time-frequency analysis diagram of a quaternary frequency shift keying signal with a signal-to-noise ratio of-7 dB, which is obtained without using the method of the present invention in embodiment 1 of the present invention;
fig. 10 is a time-frequency analysis diagram of a quaternary frequency shift keying signal with a signal-to-noise ratio of-7 dB obtained by the method according to embodiment 1 of the present invention;
fig. 11 is a block diagram of a time-frequency feature extraction system for communication signal modulation pattern recognition according to embodiment 2 of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention applies a Variational Modal Decomposition (VMD) algorithm, which is a new signal processing technology and is a completely non-recursive, self-adaptive and quasi-orthogonal signal decomposition method. Generally, the communication signal is a non-stationary signal composed of single or multiple fixed frequency modes, so the VMD algorithm is well suited for the decomposition and processing of the communication signal. The VMD determines the frequency center of each Intrinsic Mode Function (IMF) component and its bandwidth by iteratively searching for an optimal solution to the variational model. The method has the essence of wiener filtering, has good noise robustness, and effectively solves the problems of modal aliasing, endpoint effect, pseudo component and the like in the EMD algorithm. However, when the VMD is used to decompose the signal, the number of decomposition layers needs to be preset, and the selection of the decomposition layers has a problem; in addition, how to distinguish the signal modal component from the noise modal component in the modal components obtained by VMD decomposition is also a difficult problem to be solved.
The invention aims to provide a time-frequency feature extraction method and a time-frequency feature extraction system for communication signal modulation pattern recognition. The VMD decomposition algorithm can inhibit the generation of cross terms in the time-frequency analysis process, solve the problem that cross term inhibition and time-frequency aggregation are contradictory, and introduce ZAM time-frequency transformation to further inhibit noise and cross terms, so that time-frequency features with strong noise robustness can be extracted, the accuracy of time-frequency feature extraction is improved, and the method is applied to the identification of a communication signal modulation mode and can also improve the accuracy of modulation identification.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Example 1
Referring to fig. 1, the present embodiment provides a time-frequency feature extraction method for communication signal modulation pattern recognition, including:
in order to make the technical scheme of the invention more clear to those skilled in the art, the simulation signal is selected as a quaternary frequency shift keying signal (4fsk) in the communication signal, the sampling frequency is 1024Khz, the number of sampling points is 2048, and the frequency f of the carrier signal is selected1＝40KHz，f2＝80KHz，f3＝200KHz，f4The timing characteristic extraction method of the present invention is illustrated by taking 300KHz with gaussian white noise and signal-to-noise ratio of-7 dB as an example, which does not have any limiting effect on the scheme of the present invention. FIG. 2 is a time domain waveform diagram of a quad frequency shift keying signal without noise; fig. 3 is a time domain waveform diagram of a quaternary frequency shift keyed signal with a signal to noise ratio of-7 dB.
Step S1: acquiring an original communication signal, and carrying out zero equalization on the original communication signal;
the expression for zero-averaging the original communication signal x (t) (t ═ 1, 2., 2048) is as follows:
wherein x (t) is the original communication signal,
x' (t) is the original communication signal mean value, and is the communication signal after zero equalization.
Step S2: acquiring a three-order cumulant slice spectrum of the communication signal after zero equalization;
wherein, the expression of the third-order cumulant slice spectrum is as follows:
wherein S (f) is a third-order cumulant slice spectrum, f is frequency, tau is time delay, and c (tau ) is x' (t);
c(τ,τ)＝E[x'(t)x'(t+τ)x'(t+τ)]
in the formula, x' (t) is a communication signal after zero equalization, t is a sampling time, τ is a time delay, and E (.) is a mathematical expectation symbol.
Step S3: judging the number of main frequency components of the communication signal after zero equalization according to the third-order cumulant slice spectrum, and determining the number of decomposition layers of VMD decomposition based on the number of the main frequency components; determining an optimal number of decomposition layers k from the number of decomposition layers0；
As shown in fig. 4, a third order cumulant slice spectrum s (f) of a quaternary frequency shift keying signal with a signal-to-noise ratio of-7 dB is shown;
step S301: according to the third-order cumulant slice spectrum, the number of main frequency components of the signal can be preliminarily judged to be 4, and the number of decomposition layers is 5, 6, 7, 8, 9, 10, 11 and 12; it is necessary to determine an optimum number of decomposition layers k from the plurality of decomposition layers0；
Step S302: randomly selecting a decomposition layer number, which is marked as k1Performing VMD decomposition on the original communication signal to generate k1Intrinsic mode function component, denoted
Calculating zero-delay fourth-order cumulant of each intrinsic mode function component;
wherein, calculating the zero-delay fourth-order cumulant for each eigenmode function component respectively, specifically comprises:
carrying out zero equalization processing on each intrinsic mode function component to obtain a zero equalized signal component which is recorded as
Carrying out maximum amplitude normalization processing on each equalized intrinsic mode function component, and recording the normalized intrinsic mode function component as
Wherein the maximum amplitude valueThe normalized computational expression is:
imfi'(t)＝imfi(t)/max(|imfi(t)|) i＝1,2,...,k1
in the formula, imfi' is the normalized eigenmode function component; imfiA zero mean signal component;
and calculating the zero time delay fourth-order cumulant of each normalized intrinsic mode function component according to the zero time delay fourth-order cumulant expression.
The zero-delay fourth-order cumulant expression is as follows:
wherein the content of the first and second substances,
representing the fourth order cumulant of zero delay;
is imfiThe zero-delay fourth-order moment of',
the zero-time-delay second-order moment of (c),
i＝1,2,...,k1and E (.) is the mathematically expected symbol.
Step S303: based on a judgment criterion of high-order cumulant, judging useful signal components and noise components of each intrinsic mode function component to obtain intrinsic mode function components which are judged as useful signal components and marked as useful signals;
the judgment criterion based on the high-order cumulant is specifically as follows:
setting each of said normalized eigenmode function componentsThe zero-delay fourth-order cumulant is respectively
If it is
The ith eigenmode function component is determined to be a noise component.
Step S304: adding all the useful signal components to obtain a reconstructed signal, and calculating a cross-correlation function value of the reconstructed signal and the original communication signal at the time delay of 0;
suppose k1Of the IMF components, m useful signal components are set as xmf1,ximf2,...,ximfmIf the reconstructed signal x "(t) is xmf1+ximf2+...+ximfm；
The calculating a cross-correlation function value of the reconstructed signal and the original communication signal at a time delay of 0 specifically includes:
the cross-correlation function expression of the reconstructed signal and the original communication signal is as follows:
rxx”(τ)＝E[x(t)x”(t+τ)]
wherein r isxx”(τ) is a cross-correlation function value, x (t) is an original communication signal, x' (t) is a communication signal after zero equalization, E (.) is a mathematical expectation symbol, and τ is a time delay;
let τ equal to 0, then rxx”(0) The value of the cross-correlation function at a time delay of 0 for the reconstructed signal and the original communication signal is obtained.
Step S305: returning to step 302 until all the decomposition layer numbers are traversed; obtaining a corresponding cross-correlation function value for each decomposition layer number;
as shown in fig. 5, a line graph of the relationship between the cross-correlation function value of the reconstructed signal and the original signal obtained by different decomposition levels and the decomposition level;
step S306: selecting the decomposition layer number corresponding to the maximum cross-correlation function value as the optimal decomposition layer number k0。
As can be seen from fig. 5, when the number of decomposition levels is 5, the value of the obtained cross-correlation function is the largest, and therefore the optimal number of decomposition levels k can be obtained0＝5。
In addition, k is0And k is1The relationship of (1) is: number of decomposition levels k corresponding to maximum value of cross-correlation function1For the optimum number of decomposition layers k0Here if k1When 5 is taken, then k0And k is1Represent the same value when k1When not taking 5, k0And k is1Representing different values.
Table 1 shows the fourth-order cumulative magnitude of each eigenmode function component IMF when different decomposition levels are taken;
TABLE 1 fourth order cumulative magnitude of each eigenmode function component for different decomposition levels
As can be seen from the contents in table 1, when the number of decomposition layers is 5, the calculated zero-delay fourth-order cumulative quantity distribution of each normalized eigenmode function component is 0.0071, 0.0189, 0.0082, 0.0119, or 0.0004; and according to a judgment criterion based on the high-order cumulant, judging that the useful signal components are the first 4 intrinsic mode function components, and judging that the last intrinsic mode function component is a noise component. Then, all the useful signal components are added to obtain a reconstructed signal, i.e., the signal after the denoising process, for example, fig. 6 shows a time domain waveform diagram of the reconstructed signal after the denoising process, and compared with a time domain waveform diagram of a 4fsk signal when the signal-to-noise ratio in fig. 3 is-7 dB, the noise of the reconstructed signal is greatly reduced.
Step S4: according to the optimal decomposition layer number k0Performing VMD decomposition on the original communication signal to obtain k0The intrinsic mode function components are respectively calculated to obtain the zero delay fourth-order cumulant of each intrinsic mode function component;
according to the number k of decomposition layers in step 315, 6, 7, 8, 9, 10, 11, 12, the optimum number of decomposition layers k was obtained0Then according toOptimum number of decomposition layers k0Performing VMD decomposition on the original communication signal to obtain k0The eigenmode function components, as shown in fig. 7, are time domain waveforms of the signal components of the quaternary frequency shift keying signal with the signal-to-noise ratio of-7 dB decomposed by the VMD at the optimal decomposition level of 5; fig. 8 shows a frequency domain waveform diagram of each signal component of the quaternary frequency shift keying signal with the signal-to-noise ratio of-7 dB decomposed by the VMD at the optimal number of decomposition layers of 5.
Wherein, calculating the zero-delay fourth-order cumulant for each eigenmode function component respectively, specifically comprises: (Note that the specific contents of this part are the same as the corresponding part in step 302, only that k is added1Is replaced by k0)
Carrying out zero equalization processing on each intrinsic mode function component;
carrying out maximum amplitude normalization processing on each zero-equalized intrinsic mode function component;
and calculating the zero time delay fourth-order cumulant of each normalized intrinsic mode function component according to the zero time delay fourth-order cumulant expression.
The zero-delay fourth-order cumulant expression is as follows:
wherein the content of the first and second substances,
representing the fourth order cumulant of zero delay; imfi' is the normalized eigenmode function component;
is imfiThe zero-delay fourth-order moment of',
is imfiThe zero-time-delay second-order moment of',
i＝1,2,...,k0and E (.) is the mathematically expected symbol.
Step S5: according to a judgment criterion based on high-order cumulant, judging useful signal components and noise components for each intrinsic mode function component to obtain intrinsic mode function components judged as useful signal components, and recording the intrinsic mode function components as the useful signal components;
the judgment criterion based on the high-order cumulant is specifically as follows: (Again, the specific contents of this section are the same as the corresponding contents in step 303, except that k is set to1Is replaced by k0)
Setting the zero time delay fourth-order cumulant of each normalized eigenmode function component as
If it is
The ith eigenmode function component is determined to be a noise component.
Step S6: respectively carrying out ZAM time-frequency transformation on all the useful signal components to obtain transformed time-frequency signal components;
wherein, the ZAM time-frequency transformation expression is as follows:
where C (t, ω) is ZAM time-frequency transform of the useful signal s (t), t is time, ω is frequency, s is conjugate of s, u is an integral variable, τ is time delay, θ is frequency offset, and α is a constant controlling the shape of the kernel function.
Step S7: and superposing all the transformed time-frequency signal components to obtain the time-frequency characteristics of the original communication signals.
FIG. 9 is a time-frequency analysis diagram of a quaternary frequency-shift keying signal obtained without the method of the present invention when the signal-to-noise ratio is-7 dB. Fig. 10 is a time-frequency analysis diagram of the quaternary frequency shift keying signal obtained by the method of the present invention. The two characteristics are compared to see that under the condition of low signal-to-noise ratio, the noise reduction effect of the method is obvious, the cross term inhibition performance is greatly improved, and the time-frequency characteristic extraction effect of the signal is greatly improved.
In the embodiment, the VMD decomposition is firstly adopted to decompose the signal into a plurality of single-frequency signals, then the property that the high-order cumulant can inhibit gaussian noise is utilized, the high-order cumulant is combined with the VMD decomposition, a judgment criterion of a useful signal component and a noise component is provided, and the problem of how to distinguish the signal mode from the noise mode in the VMD decomposition is effectively solved. In the determination of VMD decomposition layer number, the cross-correlation function of the reconstructed signal and the original signal with noise is solved, and the correlation between the reconstructed signal and the signal without noise is indirectly measured, so that the decomposition layer number when the correlation between the reconstructed signal and the signal without noise is maximum is selected as the optimal VMD decomposition layer number, and the problems of blindness and lack of theoretical basis when the VMD decomposition preset decomposition layer number are effectively solved. And finally, VMD decomposition is combined with a ZAM time-frequency analysis method, the ZAM time-frequency analysis method has strong capability of inhibiting cross terms, and has good performance on compromise of cross term inhibition and time-frequency aggregation. And performing ZAM time-frequency transformation on each useful signal component, thereby achieving the purpose of further suppressing noise and cross terms. Therefore, according to the method provided by the embodiment, the accuracy of time-frequency feature extraction is greatly improved.
Example 2
As shown in fig. 11, the present embodiment provides a time-frequency feature extraction system for communication signal modulation pattern recognition, including:
a signal obtaining module M1, configured to obtain an original communication signal, and perform zero averaging on the original communication signal;
a third-order cumulant slice spectrum acquisition module M2, configured to acquire a third-order cumulant slice spectrum of the communication signal after zero equalization;
an optimal decomposition layer number determining module M3 for judging the third-order cumulant slice spectrumDetermining the number of decomposition layers of VMD decomposition based on the number of main frequency components of the communication signal after zero equalization; determining an optimal number of decomposition layers k from the number of decomposition layers0；
A zero-delay fourth-order cumulant calculation module M4 for calculating the number of the optimal decomposition layers k0Performing VMD decomposition on the original communication signal to obtain k0The intrinsic mode function components are respectively calculated to obtain the zero delay fourth-order cumulant of each intrinsic mode function component;
a useful signal component judging module M5, configured to perform, according to a judgment criterion based on a high-order cumulant, judgment on a useful signal component and a noise component for each eigenmode function component, to obtain an eigenmode function component judged as a useful signal component, and record the eigenmode function component as a useful signal component;
a ZAM time-frequency transformation module M6, configured to perform ZAM time-frequency transformation on all the useful signal components, respectively, to obtain transformed time-frequency signal components;
and the time-frequency characteristic obtaining module M7 is configured to superimpose all the transformed time-frequency signal components to obtain the time-frequency characteristic of the original communication signal.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (10)
1. A time-frequency feature extraction method for communication signal modulation mode identification is characterized by comprising the following steps:
acquiring an original communication signal, and carrying out zero equalization on the original communication signal;
acquiring a three-order cumulant slice spectrum of the communication signal after zero equalization;
judging the number of main frequency components of the communication signal after zero equalization according to the third-order cumulant slice spectrum, and determining the number of decomposition layers of VMD decomposition based on the number of the main frequency components; determining an optimal number of decomposition layers k from the number of decomposition layers0；
According to the optimal decomposition layer number k0Performing VMD decomposition on the original communication signal to obtain k0The intrinsic mode function components are respectively calculated to obtain the zero delay fourth-order cumulant of each intrinsic mode function component;
according to a judgment criterion based on high-order cumulant, judging useful signal components and noise components for each intrinsic mode function component to obtain intrinsic mode function components judged as useful signal components, and recording the intrinsic mode function components as the useful signal components;
respectively carrying out ZAM time-frequency transformation on all the useful signal components to obtain transformed time-frequency signal components;
and superposing all the transformed time-frequency signal components to obtain the time-frequency characteristics of the original communication signals.
2. The method of claim 1, wherein the expression for zero averaging the original communication signal is:
wherein x (t) is the original communication signal,
the original communication signal mean value is the communication signal after x' (t) is zero equalization。
3. The method of claim 1, wherein the third order cumulant slice spectrum is expressed as:
wherein S (f) is a third-order cumulant slice spectrum, f is frequency, tau is time delay, and c (tau ) is x' (t);
c(τ,τ)＝E[x'(t)x'(t+τ)x'(t+τ)]
in the formula, x' (t) is a communication signal after zero equalization, t is a sampling time, τ is a time delay, and E (.) is a mathematical expectation symbol.
4. The method of claim 1, wherein determining the number of decomposition levels for the VMD decomposition based on the number of dominant frequency components; determining an optimal number of decomposition layers k from the number of decomposition layers0The method specifically comprises the following steps:
if the number of the main frequency components is k, the number of decomposition layers of VMD decomposition is k +1, k +2, k +3,. and k + 8;
randomly selecting one decomposition layer number, performing VMD decomposition on the original communication signal to generate k1The method comprises the following steps of calculating zero delay fourth-order cumulant for each intrinsic mode function component;
based on a judgment criterion of high-order cumulant, judging useful signal components and noise components of each intrinsic mode function component to obtain intrinsic mode function components which are judged as useful signal components and marked as the useful signal components;
adding all the useful signal components to obtain a reconstructed signal, and calculating a cross-correlation function value of the reconstructed signal and the original communication signal at the time delay of 0;
returning to the step of randomly selecting one decomposition layer number until all the decomposition layer numbers are traversed; obtaining the corresponding cross-correlation function value for each decomposition layer number;
and selecting the decomposition layer number corresponding to the maximum cross-correlation function value as the optimal decomposition layer number.
5. The method according to claim 4, wherein said calculating a cross-correlation function value of the reconstructed signal and the original communication signal at a time delay of 0 comprises:
the cross-correlation function expression of the reconstructed signal and the original communication signal is as follows:
rxx”(τ)＝E[x(t)x”(t+τ)]
wherein r isxx”(τ) is a cross-correlation function value, x (t) is an original communication signal, x' (t) is a communication signal after zero equalization, E (.) is a mathematical expectation symbol, and τ is a time delay;
let τ equal to 0, then rxx”(0) The value of the cross-correlation function at a time delay of 0 for the reconstructed signal and the original communication signal is obtained.
6. The method according to claim 1 or 4, wherein the calculating the zero-delay fourth-order cumulant for each eigenmode function component comprises:
carrying out zero equalization processing on each intrinsic mode function component;
carrying out maximum amplitude normalization processing on each zero-equalized intrinsic mode function component;
and calculating the zero time delay fourth-order cumulant of each normalized intrinsic mode function component according to the zero time delay fourth-order cumulant expression.
7. The method of claim 6, wherein the zero-delay fourth-order cumulant expression is:
wherein the content of the first and second substances,
representing the fourth order cumulant of zero delay; imfi' is the normalized eigenmode function component;
is imfiThe zero-delay fourth-order moment of',
is imfiThe zero-time-delay second-order moment of',
e (.) is the mathematically expected symbol.
8. The method according to claim 6, wherein the decision criterion based on the high order cumulant is specifically:
setting the zero time delay fourth-order cumulant of each normalized eigenmode function component as
If it is
The ith eigenmode function component is determined to be a noise component.
9. The method of claim 1, where the ZAM time-frequency transform expression is:
where C (t, ω) is ZAM time-frequency transform of the useful signal s (t), t is time, ω is frequency, s is conjugate of s, u is an integral variable, τ is time delay, θ is frequency offset, and α is a constant controlling the shape of the kernel function.
10. A time-frequency feature extraction system based on the method of any one of claims 1 to 9, comprising:
the signal acquisition module is used for acquiring an original communication signal and carrying out zero equalization on the original communication signal;
the third-order cumulant slice spectrum acquisition module is used for acquiring a third-order cumulant slice spectrum of the communication signal after zero equalization;
the optimal decomposition layer number determining module is used for judging the number of main frequency components of the communication signal after zero equalization according to the third-order cumulant slice spectrum and determining the decomposition layer number of VMD decomposition based on the number of the main frequency components; determining an optimal number of decomposition layers k from the number of decomposition layers0；
A zero time delay fourth-order cumulant calculation module for calculating the optimal decomposition layer number k0Performing VMD decomposition on the original communication signal to obtain k0The intrinsic mode function components are respectively calculated to obtain the zero delay fourth-order cumulant of each intrinsic mode function component;
the useful signal component judging module is used for judging a useful signal component and a noise component for each intrinsic mode function component according to a judgment criterion based on the high-order cumulant to obtain the intrinsic mode function component which is judged as the useful signal component and is marked as the useful signal component;
the ZAM time-frequency transformation module is used for respectively carrying out ZAM time-frequency transformation on all the useful signal components to obtain transformed time-frequency signal components;
and the time-frequency characteristic acquisition module is used for superposing all the transformed time-frequency signal components to obtain the time-frequency characteristics of the original communication signals.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110028826.9A CN112800863B (en) | 2021-01-11 | 2021-01-11 | Time-frequency feature extraction method and system for communication signal modulation pattern recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110028826.9A CN112800863B (en) | 2021-01-11 | 2021-01-11 | Time-frequency feature extraction method and system for communication signal modulation pattern recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112800863A true CN112800863A (en) | 2021-05-14 |
| CN112800863B CN112800863B (en) | 2022-07-01 |
Family
ID=75809659
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110028826.9A Active CN112800863B (en) | 2021-01-11 | 2021-01-11 | Time-frequency feature extraction method and system for communication signal modulation pattern recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112800863B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115022915A (en) * | 2022-04-12 | 2022-09-06 | 中国电子科技集团公司第十研究所 | Communication signal visual expression method based on high-order cumulant time-frequency diagram |
| CN115987731A (en) * | 2022-12-23 | 2023-04-18 | 中南民族大学 | Frequency modulation adaptive VMD iterative noise suppression system and method thereof |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE2202659A1 (en) * | 1972-01-20 | 1973-07-26 | North American Rockwell | ELECTRONIC MUSICAL INSTRUMENT |
| CN109307862A (en) * | 2018-07-05 | 2019-02-05 | 西安电子科技大学 | A kind of target radiation source individual discrimination method |
| CN110855374A (en) * | 2019-10-31 | 2020-02-28 | 西安交通大学 | Underwater sound target radiation noise modulation feature extraction method |
| CN112073345A (en) * | 2020-07-28 | 2020-12-11 | 中国科学院信息工程研究所 | Modulation mode identification method and device, electronic equipment and storage medium |
-
2021
- 2021-01-11 CN CN202110028826.9A patent/CN112800863B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE2202659A1 (en) * | 1972-01-20 | 1973-07-26 | North American Rockwell | ELECTRONIC MUSICAL INSTRUMENT |
| CN109307862A (en) * | 2018-07-05 | 2019-02-05 | 西安电子科技大学 | A kind of target radiation source individual discrimination method |
| CN110855374A (en) * | 2019-10-31 | 2020-02-28 | 西安交通大学 | Underwater sound target radiation noise modulation feature extraction method |
| CN112073345A (en) * | 2020-07-28 | 2020-12-11 | 中国科学院信息工程研究所 | Modulation mode identification method and device, electronic equipment and storage medium |
Non-Patent Citations (3)
| Title |
|---|
| XIAODI TIAN等: "《Modulation Pattern Recognition of Communication Signals Based on Fractional Low-Order Choi-Williams Distribution and Convolutional Neural Network in Impulsive Noise Environment》", 《IEEE》 * |
| 刘景良等: "《变分模态分解和同步挤压小波变换识别时变结构瞬时频率》", 《振动与冲击》 * |
| 孙晓东等: "《基于驱动响应同步的混沌背景下谐波信号频率估计方法》", 《长春理工大学学报(自然科学版)》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115022915A (en) * | 2022-04-12 | 2022-09-06 | 中国电子科技集团公司第十研究所 | Communication signal visual expression method based on high-order cumulant time-frequency diagram |
| CN115987731A (en) * | 2022-12-23 | 2023-04-18 | 中南民族大学 | Frequency modulation adaptive VMD iterative noise suppression system and method thereof |
| CN115987731B (en) * | 2022-12-23 | 2024-04-16 | 中南民族大学 | VMD iterative noise suppression system adapting to frequency modulation and method thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112800863B (en) | 2022-07-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112800863B (en) | Time-frequency feature extraction method and system for communication signal modulation pattern recognition | |
| CN105785324B (en) | Linear frequency-modulated parameter estimating method based on MGCSTFT | |
| CN108880604B (en) | Multi-component frequency modulation interference suppression method and device in spread spectrum communication system | |
| Li et al. | Blind detection of underwater acoustic communication signals based on deep learning | |
| CN114785379B (en) | Method and system for estimating parameters of underwater sound JANUS signals | |
| Fu et al. | A circular zone partition method for identifying Duffing oscillator state transition and its application to BPSK signal demodulation | |
| Sisi et al. | Carrier frequency and symbol rate estimation based on cyclic spectrum | |
| CN110289926B (en) | Spectrum sensing method based on symmetric peak values of cyclic autocorrelation function of modulation signal | |
| CN101588191A (en) | Method and device for radio signal recognition | |
| CN109004996B (en) | Signal detection method based on multi-sine-window power spectrum peak value | |
| CN103997475A (en) | Method for recognizing digital modulation signals under Alpha stable distribution noise | |
| CN116032709B (en) | Method and device for blind demodulation and modulation feature analysis of FSK signal without priori knowledge | |
| CN110190917B (en) | Frequency spectrum hole sensing method, device and equipment for LTE230MHz power wireless private network | |
| CN109709581B (en) | Method for rapidly detecting and acquiring cyclic characteristic parameters of satellite navigation signals with strong interference | |
| Shi et al. | Fourth order cumulants in distinguishing single carrier from OFDM signals | |
| CN115510924B (en) | Radio frequency fingerprint identification method based on improved variational modal decomposition | |
| CN111490956A (en) | MFSK modulation identification method based on first-order cyclostationarity | |
| CN115378776A (en) | MFSK modulation identification method based on cyclic spectrum parameters | |
| Imani et al. | Using weighted multilevel wavelet decomposition for wideband spectrum sensing in cognitive radios | |
| CN107493147B (en) | Polarization similarity cognitive signal learning method for full-duplex authorized user | |
| CN115941084A (en) | Underwater acoustic communication preamble signal detection method and device based on time-frequency graph template matching | |
| CN114201990A (en) | Communication interference identification method based on decision tree | |
| CN113162879B (en) | Modulation signal identification method combining feature extraction | |
| Molchanov et al. | A novel bispectrum-based frequency encoding technique developed for noisy and multipath fading radio links | |
| CN113271273A (en) | Modulation identification method based on wiener filtering preprocessing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |