CN111010356A - Underwater acoustic communication signal modulation mode identification method based on support vector machine - Google Patents
Underwater acoustic communication signal modulation mode identification method based on support vector machine Download PDFInfo
- Publication number
- CN111010356A CN111010356A CN201911084879.1A CN201911084879A CN111010356A CN 111010356 A CN111010356 A CN 111010356A CN 201911084879 A CN201911084879 A CN 201911084879A CN 111010356 A CN111010356 A CN 111010356A
- Authority
- CN
- China
- Prior art keywords
- signal
- underwater acoustic
- time
- frequency
- acoustic communication
- 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.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/0012—Modulated-carrier systems arrangements for identifying the type of modulation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B11/00—Transmission systems employing sonic, ultrasonic or infrasonic waves
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B13/00—Transmission systems characterised by the medium used for transmission, not provided for in groups H04B3/00 - H04B11/00
- H04B13/02—Transmission systems in which the medium consists of the earth or a large mass of water thereon, e.g. earth telegraphy
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B15/00—Suppression or limitation of noise or interference
-
- 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
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
The invention provides an underwater acoustic communication signal modulation mode identification method based on a support vector machine, which comprises the steps of preprocessing an underwater acoustic signal containing complex ocean background noise through a nonlinear piecewise exponential function, extracting time domain L-Z complexity characteristics, frequency domain fractal box dimension characteristics and time-frequency domain Renyi entropy characteristics of the underwater acoustic signal as the input of an SVM classifier, and then selecting proper kernel functions and parameters to complete the modulation mode identification of the underwater acoustic communication signal. According to the method, the influence of complex ocean background noise is eliminated through a nonlinear transformation preprocessing method, three features in a time domain, a frequency domain and a time-frequency domain are extracted to form the feature vector, and compared with a single feature method, the performance is obviously improved. The method can be used for monitoring and identifying the underwater acoustic communication signals in shallow sea, so that the underwater acoustic communication behaviors of enemies can be sensed in advance, and the marine defense strength of China is improved.
Description
Technical Field
The invention relates to the field of information signal processing, in particular to an underwater acoustic communication signal identification method, relating to theories of underwater signal processing, feature identification, machine learning and the like.
Background
The twenty-first century is an ocean development era, and with the rapid development of science and technology, the ocean gradually becomes an important field of human development, wherein the identification of underwater acoustic communication signal modulation modes under non-cooperative conditions is an important research direction in the ocean safety field.
In recent years, defense and monitoring in the marine field are gradually strengthened in all military and major countries in the world. Currently, the most advanced underwater monitoring network is the Seaweb in the united states, which can perform high-quality data transmission in a wide water area under the severe shallow sea conditions by using an underwater acoustic network, and can support data packets with the length of 2 kbytes and the communication rate of 2400bits/s at most. No matter the underwater vehicle or the underwater monitoring network works normally, information exchange and data transmission among equipment are required, and the underwater communication can only select acoustic signals as energy carriers to meet normal requirements. Therefore, the underwater acoustic communication technology is a core technology of all underwater operations, and the realization of the identification of the underwater acoustic communication signal modulation mode under the non-cooperative condition has important significance on underwater acoustic countermeasure.
Currently, there are two main methods for identification of modulated signals: a decision theory method based on likelihood function and a statistical pattern recognition method based on feature extraction. The statistical pattern recognition process using feature extraction is mainly divided into two modules: firstly, extracting features, and secondly, carrying out pattern matching on the extracted features. Because the underwater acoustic channel has various characteristics of strong multi-path, narrow bandwidth, long time delay, non-Gaussian background noise and the like, the traditional modulation mode identification method cannot meet the requirement of identification of the underwater acoustic signal modulation mode. Therefore, the research on the suitable underwater acoustic communication signal modulation mode identification method has important significance in military and civil aspects.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides an underwater acoustic communication signal modulation mode identification method based on a support vector machine. Aiming at various characteristics of strong multi-path, long time delay, short-time pulse interference and the like of an underwater sound channel, a Support Vector Machine (SVM) -based underwater sound communication signal modulation mode identification method is provided. The underwater acoustic communication signal modulation method based on the nonlinear piecewise exponential function preprocesses an underwater acoustic signal containing complex ocean background noise, extracts time domain L-Z complexity characteristics, frequency domain fractal box dimension characteristics and time frequency domain Renyi entropy characteristics of the underwater acoustic signal as input of an SVM classifier, and then selects proper kernel functions and parameters to complete modulation mode identification of the underwater acoustic communication signal.
The technical scheme adopted by the invention for solving the technical problem comprises the following steps:
the first step is as follows: the sensor receives an underwater acoustic communication signal;
receiving an underwater acoustic communication signal S containing complex marine environment noise through an underwater acoustic sensor;
the second step is that: carrying out nonlinear transformation pretreatment;
preprocessing the underwater acoustic communication signal S through nonlinear transformation to eliminate the influence of part of impact noise;
selecting piecewise linear piecewise exponential to do nonlinear transformation, wherein the expression is as follows:
in the formula PSF (S) represents that the underwater sound signal S is subjected to nonlinear transformation, and x is a signal subjected to the nonlinear transformation;
the third step: extracting time domain L-Z complexity characteristics;
L-Z complexity characterizes a sequence by two operations, copy and add; firstly, carrying out difference sequence transformation on a signal x after nonlinear transformation, then carrying out quantization coding on the transformed sequence to obtain a reconstructed sequence { r (k) }, and solving L-Z complexity C (L) of the reconstructed sequence;
the fourth step: extracting frequency domain fractal box dimension characteristics;
in the second step, the preprocessed underwater acoustic signal x is a real signal and is converted by HilbertObtaining the frequency spectrum of the complex signal, and extracting the frequency domain fractal box dimension characteristic D by using a box dimension calculation methodb(ii) a Calculating the box dimension by adopting a two-dimensional box method, namely, regarding the frequency spectrum sequence as a plane of a two-dimensional space, and then calculating the number of covered grids;
the fifth step: extracting time-frequency domain Renyi entropy characteristics;
normalizing the energy of the preprocessed underwater sound signal x, and then carrying out smooth pseudo-wigner-Weirr time-frequency transformation to obtain a time-frequency distribution matrix P (t, f) of the signal, wherein t is a time axis, f is a frequency axis, and then solving a Renyi entropy for the time-frequency distribution matrix P (t, f) to obtain the Renyi entropy of the time-frequency distribution of the signal, wherein the formula is as follows:
in the formula RαRenyi entropy of order α, order α, log2Represents logarithm of base 2, and ^ integral ^ Pα(t, f) dtdf represents the double integration of the α -order time-frequency distribution matrix with respect to time t and frequency f;
and a sixth step: constructing a signal feature vector;
extracting three characteristics of the known sample signal according to the steps from the third step to the fifth step, and dividing the time domain L-Z complexity characteristic C (L) and the frequency domain into box-dimension characteristic DbAnd the time-frequency domain Renyi entropy R3Normalizing the features to synthesize a multi-dimensional feature vector
The seventh step: SVM classifier obtained through sample training
When training the SVM, inputting the feature vector of the sample data into a classifier for training;
feature vector from known sample signal
Inputting the data into SVM for training to obtain a multi-feature fusion modulation mode recognition classifier model of multi-class SVM, and then utilizing the classifier model to carry out classification on test samplesAnd (4) line classification, wherein the result is subjected to a maximum voting method, and the class with the most votes is determined as a sample class.
The detailed steps for extracting the time domain L-Z complexity feature in the third step are as follows:
a. assuming that the signal time series { x (i) }, i ═ 1,2, …, L +1, and the length is L +1, a new series formed by { s (k) } being the absolute value of the signal difference at the adjacent time is defined, i.e., s (k) ═ x (k) |, k ═ 1,2, …, L, and the length is k;
b. quantizing and coding s (k), assuming the quantization level is N, and making it
At (0, a)]Dividing s (k) into N layers, there are:
c. obtaining an L-Z complexity characteristic c (L) for the obtained reconstructed sequence { r (k) }, k ═ 1,2, …, L containing N symbols;
the detailed steps of the L-Z complexity feature C (L) are as follows:
initially adding r (1) to an empty generation pool, setting existing character strings r (1) r (2) … r (L) in the generation pool, wherein L is less than L, and r (L) is completed by adding operation, and making P (r (1) r (2) … r (L) and Q (r +1) both represent character strings, and PQ represents a character string obtained by splicing the P character string and the Q character string together and deleting the last character;
judging whether Q can be copied from PQ, wherein the copying indicates that the PQ character string contains the same character sequence as the Q character string, if the Q can be copied, P is kept unchanged, Q continuously supplements one character, namely Q is r (l +1) r (l +2), and then judging whether Q can be copied from PQ; if the generation pool cannot be copied, adding Q to the generation pool, wherein P is r (1) r (2) … r (l) r (l +1) and Q is r (l +2), judging whether Q can be copied from PQ, repeating the steps until the generation pool contains all reconstruction sequences, and counting the times c (L) of the adding operation;
if the last step before the generation pool contains all the reconstruction sequences, judging whether the operation of Q being the substring in the PQ is copying, and if the reconstruction sequences are judged completely and Q cannot supplement characters any more, c (L) plus 1;
the resulting normalized L-Z complexity C (L) is:
log in formulaNRepresenting the logarithm based on the quantization series N.
The detailed steps of extracting the frequency domain fractal box dimension characteristics in the fourth step are as follows:
a. regarding the frequency spectrum signal, regarding the frequency spectrum signal as a plane of a two-dimensional space, the length of the plane is a frequency value of the frequency spectrum signal represented by an abscissa, the width of the plane is an ordinate representing an amplitude value of the frequency spectrum signal, and the length is recorded as M and the width is recorded as W;
b. dividing a plane where the frequency spectrum signal is located into R multiplied by R grids;
c. finding out a maximum envelope u and a minimum envelope value b in each scale of which the frequency value of the frequency spectrum signal, namely the abscissa is divided, wherein the grid numbers of the ordinate corresponding to the maximum envelope u and the minimum envelope b are k and j respectively, n boxes are needed to cover the section of envelope signal, and n is k-j + 1;
d. the number of the covering boxes corresponding to the scales divided by each abscissa is summed, and the sum is SN；
e. Changing the value of R and solving a set of SNThen, linear fitting is applied, and the slope of the obtained straight line is the box dimension Db；
The detailed steps of the standardization processing in the sixth step are as follows:
a. and (3) standardization:
after standardization, the mean value of each characteristic parameter is 0, the variance is 1 and no dimension exists; wherein, Ck,iFor the ith feature value in the kth class of features,
is the mean, σ, of the class eigenvalueskIs standard deviation, C'k,iThe normalized characteristic value is obtained;
b. poor conversion:
converting all feature values to [0,1]Within interval, wherein C'k,minAnd C'k,maxRespectively a minimum characteristic value and a maximum characteristic value in the kth class of characteristics; c ″)k,iThe characteristic value after the extremely poor conversion is obtained.
The underwater acoustic communication signal modulation mode recognition method based on the Support Vector Machine (SVM) has the advantages that underwater acoustic signals containing complex ocean background noise are preprocessed through a nonlinear piecewise exponential function, time domain L-Z complexity characteristics, frequency domain fractal box dimension characteristics and time frequency domain Renyi entropy characteristics of the underwater acoustic signals are extracted and serve as input of the SVM classifier, then proper kernel functions and parameters are selected, and the underwater acoustic communication signal modulation mode recognition is completed. Aiming at the problem of identification of an underwater acoustic communication signal modulation mode with short-time impulse noise in a shallow sea environment, the influence of complex ocean background noise is eliminated by a nonlinear transformation preprocessing method, three features in a time domain, a frequency domain and a time-frequency domain are extracted to form a feature vector, and compared with a single feature method, the performance is obviously improved. The method can be used for monitoring and identifying the underwater acoustic communication signals in shallow sea, so that the underwater acoustic communication behaviors of enemies can be sensed in advance, and the marine defense strength of China is improved.
Drawings
Fig. 1 is a block diagram of an underwater acoustic communication signal modulation mode identification method according to the present invention.
Detailed Description
The invention is further illustrated with reference to the following figures and examples.
As shown in fig. 1, the detailed steps of the technical scheme of the invention are as follows:
the first step is as follows: sensor receiving underwater acoustic communication signal
The underwater acoustic communication signal S is received and extracted by the sensor.
The second step is that: performing nonlinear transformation preprocessing
The received signal with impact noise is directly used for next detection and identification, so that a large error occurs, and therefore, the received signal needs to be preprocessed to eliminate the influence of part of the impact noise.
Selecting piecewise linear piecewise exponential to do nonlinear transformation, wherein the expression is as follows:
in the formula PSE is an index with a natural constant as the base for the average power of the received underwater acoustic signal S. And f (S) shows that the underwater sound signal S is subjected to nonlinear transformation, and x is a signal subjected to nonlinear transformation. For a received signal having an amplitude greater than twice the square of the average power of the signal, the received signal is multiplied by an exponential fading factor associated with the signal, and when the amplitude is close to twice the square of the average power of the signal, the fading is small, thus not only distorting the signal but also eliminating large impulsive noise.
The third step: extracting time domain L-Z complexity features
Lempel-Ziv complexity (hereinafter abbreviated as L-Z complexity), describes the behavior of a sequence by two simple operations, copy and add. Firstly, reconstructing a signal x (i) after nonlinear transformation to obtain a reconstruction sequence { r (k) }, and solving the L-Z complexity C (L) of the reconstruction sequence.
The L-Z complexity C (L) is obtained by the following steps:
a. assuming that the signal time series { x (i) }, i ═ 1,2, …, L +1 is of length L +1, a new series formed by { s (k) } being the absolute values of the signal differences at adjacent time instants is defined, i.e., s (k) | x (k) — x (k +1) |, k ═ 1,2, …, L, of length k;
b. quantizing and coding s (k), assuming the quantization level is N, and making it
At (0, a)]Dividing s (k) into N layers, there are:
c. obtaining an L-Z complexity characteristic c (L) for the obtained reconstructed sequence { r (k) }, k ═ 1,2, …, L containing N symbols;
the L-Z complexity feature c (L) is obtained for the sequence r (k) }, k ═ 1,2, …, L as follows:
initially adding r (1) to an empty generation pool, setting existing character strings r (1) r (2) … r (L) in the generation pool, wherein L is less than L, and r (L) is completed by adding operation, and making P (r (1) r (2) … r (L) and Q (r +1) both represent character strings, and PQ represents a character string obtained by splicing the P character string and the Q character string together and deleting the last character;
judging whether Q can be copied from PQ, wherein the copying indicates that the PQ character string contains the same character sequence as the Q character string, if the Q can be copied, P is kept unchanged, Q continuously supplements one character, namely Q is r (l +1) r (l +2), and then judging whether Q can be copied from PQ; if the generation pool cannot be copied, adding Q to the generation pool, wherein P is r (1) r (2) … r (l) r (l +1) and Q is r (l +2), judging whether Q can be copied from PQ, repeating the steps until the generation pool contains all reconstruction sequences, and counting the times c (L) of the adding operation;
if the last step before the generation pool contains all the reconstruction sequences, judging whether the operation of Q being the substring in the PQ is copying, and if the reconstruction sequences are judged completely and Q cannot supplement characters any more, c (L) needs to be added with 1;
finally obtaining the normalized L-Z complexity C (L):
log in formulaNRepresenting the logarithm based on the quantization series N;
the fourth step: extracting frequency domain fractal box dimension features
The obtained preprocessed underwater sound signal is a real signal, a complex signal is obtained by using a Hilbert transform method, and a frequency spectrum is calculated.
The real signal is represented as x (t), and Hilbert transform is carried out on the x (t) to obtain an analytic form
For analytic signals
And performing FFT, transforming the time domain signal to a frequency domain, and extracting a frequency domain fractal box dimension characteristic by using the following box dimension calculation method:
a. regarding the frequency spectrum signal, regarding the frequency spectrum signal as a plane of a two-dimensional space, the length of the plane is a frequency value of the frequency spectrum signal represented by an abscissa, the width of the plane is an ordinate representing an amplitude of the frequency spectrum signal, and the length is recorded as M and the width is recorded as W;
b. dividing a plane where the frequency spectrum signal is located into R multiplied by R grids;
c. finding out a maximum envelope u and a minimum envelope value b in each scale of which the frequency value of the frequency spectrum signal, namely the abscissa is divided, wherein the grid numbers of the ordinate corresponding to the maximum envelope u and the minimum envelope value b are k and j respectively, n boxes are needed to cover the section of envelope signal, and n is k-j + 1;
d. the number of the covering boxes corresponding to the scales divided by each abscissa is summed, and the sum is SN；
e. Changing the value of R to obtain a set of SNFrom that, using a linear fit, the slope of the resulting line, i.e. the box dimension Db；
The fifth step: extracting time-frequency domain Renyi entropy characteristics of the preprocessed underwater sound signals;
and normalizing the energy of the preprocessed underwater acoustic signal, and then carrying out smooth pseudo-Wegener-Weirr time-frequency transformation to obtain a time-frequency distribution matrix P (t, f) of the signal, wherein t is a time axis and f is a frequency axis. The wigner-well distribution is a quadratic transformation of the signal, with cross terms present, which are suppressed by window smoothing since they are oscillatory.
The smooth pseudo-wigner-well distribution is defined as:
wherein h (τ) is a rectangular window;
and then solving Renyi entropy for the time-frequency distribution matrix P (t, f) to obtain the Renyi entropy of the time-frequency distribution of the preprocessed signal:
in the formula Rαα th Renyi entropy, α th order, log2Represents logarithm of base 2, and ^ integral ^ Pα(t, f) dtdf represents the double integration of the α th order time-frequency distribution matrix over time t and frequency f.
For signals with strong time-frequency distribution regularity and low complexity, the information content is less, the corresponding entropy value is small, and for signals composed of a large number of randomly scattered signal components, the corresponding entropy value is larger, the non-integer order α generates negative entropy value, the odd order condition with better stability is considered, α is 3, and the time-frequency domain three-order Renyi entropy characteristic R is obtained3。
And a sixth step: constructing signal feature vectors
Extracting the three characteristics of the known sample signal, namely a time domain L-Z complexity characteristic C (L) and a frequency domain fractal box dimension characteristic DbTime-frequency domain Renyi entropy R3The features are normalized to synthesize multi-dimensional feature vectors
The steps of the feature normalization method (taking the time domain L-Z complexity feature C as an example) are as follows:
a. and (3) standardization:
after standardization, the mean value of each characteristic parameter is 0, the variance is 1 and no dimension exists; wherein, Ck,iFor the ith feature value in the kth class of features,
for the value of the mean, σ, of the class characteristickIs standard deviation, C'k,iThe normalized characteristic value is obtained;
b. poor conversion:
converting all feature values to [0,1]Within interval, wherein C'k,minAnd C'k,maxRespectively a minimum characteristic value and a maximum characteristic value in the kth class of characteristics; c ″)k,iThe characteristic value after the range conversion is obtained;
the seventh step: SVM classifier obtained through sample training
For the multi-class classification condition, the multi-class problem is converted into a plurality of two-class problems, and a one-to-one method is adopted. And (3) respectively processing every two classes in all classes K of the training sample, and needing K (K-1)/2 classifiers.
Feature vector from known sample signal
And inputting the data to an SVM for training to obtain a multi-feature fusion modulation mode recognition classifier model of the multi-class SVM. And classifying the test samples by using all classifier models, adopting a maximum voting method for the result, and determining the class with the maximum number of votes as the sample class.
Claims (4)
1. An underwater acoustic communication signal modulation mode identification method based on a support vector machine is characterized by comprising the following steps:
the first step is as follows: the sensor receives an underwater acoustic communication signal;
receiving an underwater acoustic communication signal S containing complex marine environment noise through an underwater acoustic sensor;
the second step is that: carrying out nonlinear transformation pretreatment;
preprocessing the underwater acoustic communication signal S through nonlinear transformation to eliminate the influence of part of impact noise;
selecting piecewise linear piecewise exponential to do nonlinear transformation, wherein the expression is as follows:
in the formula PSF (S) represents that the underwater sound signal S is subjected to nonlinear transformation, and x is a signal subjected to the nonlinear transformation;
the third step: extracting time domain L-Z complexity characteristics;
L-Z complexity characterizes a sequence by two operations, copy and add; firstly, carrying out difference sequence transformation on a signal x after nonlinear transformation, then carrying out quantization coding on the transformed sequence to obtain a reconstructed sequence { r (k) }, and solving L-Z complexity C (L) of the reconstructed sequence;
the fourth step: extracting frequency domain fractal box dimension characteristics;
in the second step, the preprocessed underwater acoustic signal x is a real signal, a complex signal is obtained by Hilbert transformation, a frequency spectrum is obtained, and a frequency domain fractal box dimension characteristic D is extracted by a box dimension calculation methodb(ii) a Calculating the box dimension by adopting a two-dimensional box method, namely, regarding the frequency spectrum sequence as a plane of a two-dimensional space, and then calculating the number of covered grids;
the fifth step: extracting time-frequency domain Renyi entropy characteristics;
normalizing the energy of the preprocessed underwater sound signal x, and then carrying out smooth pseudo-wigner-Weirr time-frequency transformation to obtain a time-frequency distribution matrix P (t, f) of the signal, wherein t is a time axis, f is a frequency axis, and then solving a Renyi entropy for the time-frequency distribution matrix P (t, f) to obtain the Renyi entropy of the time-frequency distribution of the signal, wherein the formula is as follows:
in the formula RαRenyi entropy of order α, order α, log2Represents logarithm of base 2, and ^ integral ^ Pα(t, f) dtdf represents the double integration of the α -order time-frequency distribution matrix with respect to time t and frequency f;
and a sixth step: constructing a signal feature vector;
extracting three characteristics of the known sample signal according to the steps from the third step to the fifth step, and dividing the time domain L-Z complexity characteristic C (L) and the frequency domain into box-dimension characteristic DbAnd the time-frequency domain Renyi entropy R3Normalizing the features to synthesize a multi-dimensional feature vector
The seventh step: SVM classifier obtained through sample training
When training the SVM, inputting the feature vector of the sample data into a classifier for training;
feature vector from known sample signal
Inputting the training data to the SVM, obtaining a multi-feature fusion modulation mode recognition classifier model of a multi-class SVM, classifying the test samples by using the classifier model, and determining the class with the maximum number of votes as the sample class by adopting a maximum voting method for the result.
2. The underwater acoustic communication signal modulation mode identification method based on the support vector machine according to claim 1, characterized in that:
the detailed steps for extracting the time domain L-Z complexity feature in the third step are as follows:
a. assuming that the signal time series { x (i) }, i ═ 1,2, …, L +1, and the length is L +1, a new series formed by { s (k) } being the absolute value of the signal difference at the adjacent time is defined, i.e., s (k) ═ x (k) |, k ═ 1,2, …, L, and the length is k;
b. quantizing and coding s (k), assuming the quantization level is N, and making it
At (0, a)]Dividing s (k) into N layers, there are:
c. obtaining an L-Z complexity characteristic c (L) for the obtained reconstructed sequence { r (k) }, k ═ 1,2, …, L containing N symbols;
the detailed steps of the L-Z complexity feature C (L) are as follows:
initially adding r (1) to an empty generation pool, setting existing character strings r (1) r (2) … r (L) in the generation pool, wherein L is less than L, and r (L) is completed by adding operation, and making P (r (1) r (2) … r (L) and Q (r +1) both represent character strings, and PQ represents a character string obtained by splicing the P character string and the Q character string together and deleting the last character;
judging whether Q can be copied from PQ, wherein the copying indicates that the PQ character string contains the same character sequence as the Q character string, if the Q can be copied, P is kept unchanged, Q continuously supplements one character, namely Q is r (l +1) r (l +2), and then judging whether Q can be copied from PQ; if the generation pool cannot be copied, adding Q to the generation pool, wherein P is r (1) r (2) … r (l) r (l +1) and Q is r (l +2), judging whether Q can be copied from PQ, repeating the steps until the generation pool contains all reconstruction sequences, and counting the times c (L) of the adding operation;
if the last step before the generation pool contains all the reconstruction sequences, judging whether the operation of Q being the substring in the PQ is copying, and if the reconstruction sequences are judged completely and Q cannot supplement characters any more, c (L) plus 1;
the resulting normalized L-Z complexity C (L) is:
log in formulaNRepresenting the logarithm based on the quantization series N.
3. The underwater acoustic communication signal modulation mode identification method based on the support vector machine according to claim 1, characterized in that:
the detailed steps of extracting the frequency domain fractal box dimension characteristics in the fourth step are as follows:
a. regarding the frequency spectrum signal, regarding the frequency spectrum signal as a plane of a two-dimensional space, the length of the plane is a frequency value of the frequency spectrum signal represented by an abscissa, the width of the plane is an ordinate representing an amplitude value of the frequency spectrum signal, and the length is recorded as M and the width is recorded as W;
b. dividing a plane where the frequency spectrum signal is located into R multiplied by R grids;
c. finding out a maximum envelope u and a minimum envelope value b in each scale of which the frequency value of the frequency spectrum signal, namely the abscissa is divided, wherein the grid numbers of the ordinate corresponding to the maximum envelope u and the minimum envelope b are k and j respectively, n boxes are needed to cover the section of envelope signal, and n is k-j + 1;
d. the number of the covering boxes corresponding to the scales divided by each abscissa is summed, and the sum is SN；
e. Changing the value of R and solving a set of SNThen, linear fitting is applied, and the slope of the obtained straight line is the box dimension Db。
4. The underwater acoustic communication signal modulation mode identification method based on the support vector machine according to claim 1, characterized in that:
the detailed steps of the standardization processing in the sixth step are as follows:
a. and (3) standardization:
after standardization, the mean value of each characteristic parameter is 0, the variance is 1 and no dimension exists; wherein, Ck,iFor the ith feature value in the kth class of features,
is the mean, σ, of the class eigenvalueskIs standard deviation, C'k,iThe normalized characteristic value is obtained;
b. poor conversion:
converting all feature values to [0,1]Within interval, wherein C'k,minAnd C'k,maxRespectively a minimum characteristic value and a maximum characteristic value in the kth class of characteristics; c'k,iThe characteristic value after the extremely poor conversion is obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911084879.1A CN111010356A (en) | 2019-11-08 | 2019-11-08 | Underwater acoustic communication signal modulation mode identification method based on support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911084879.1A CN111010356A (en) | 2019-11-08 | 2019-11-08 | Underwater acoustic communication signal modulation mode identification method based on support vector machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111010356A true CN111010356A (en) | 2020-04-14 |
Family
ID=70111782
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911084879.1A Pending CN111010356A (en) | 2019-11-08 | 2019-11-08 | Underwater acoustic communication signal modulation mode identification method based on support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111010356A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111612130A (en) * | 2020-05-18 | 2020-09-01 | 吉林大学 | Frequency shift keying communication signal modulation mode identification method |
| CN112737992A (en) * | 2020-09-23 | 2021-04-30 | 青岛科技大学 | Underwater sound signal modulation mode self-adaptive in-class identification method |
| CN112800862A (en) * | 2021-01-11 | 2021-05-14 | 吉林大学 | Non-stationary signal time-frequency matrix reconstruction method and system |
| CN113452637A (en) * | 2021-09-01 | 2021-09-28 | 中国海洋大学 | Underwater acoustic communication signal modulation identification method based on feature selection and support vector machine |
| CN114004982A (en) * | 2021-10-27 | 2022-02-01 | 中国科学院声学研究所 | Acoustic Haar feature extraction method and system for underwater target recognition |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107577999A (en) * | 2017-08-22 | 2018-01-12 | 哈尔滨工程大学 | A kind of radar emitter signal intra-pulse modulation mode recognition methods based on singular value and fractal dimension |
| US20180314974A1 (en) * | 2017-04-27 | 2018-11-01 | Raytheon Company | Machine learning algorithm with binary pruning technique for automatic interpulse modulation recognition |
| CN109299697A (en) * | 2018-09-30 | 2019-02-01 | 泰山学院 | Deep neural network system and method based on underwater sound communication Modulation Mode Recognition |
| CN110166389A (en) * | 2019-06-12 | 2019-08-23 | 西安电子科技大学 | Modulation Identification method based on least square method supporting vector machine |
-
2019
- 2019-11-08 CN CN201911084879.1A patent/CN111010356A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180314974A1 (en) * | 2017-04-27 | 2018-11-01 | Raytheon Company | Machine learning algorithm with binary pruning technique for automatic interpulse modulation recognition |
| CN107577999A (en) * | 2017-08-22 | 2018-01-12 | 哈尔滨工程大学 | A kind of radar emitter signal intra-pulse modulation mode recognition methods based on singular value and fractal dimension |
| CN109299697A (en) * | 2018-09-30 | 2019-02-01 | 泰山学院 | Deep neural network system and method based on underwater sound communication Modulation Mode Recognition |
| CN110166389A (en) * | 2019-06-12 | 2019-08-23 | 西安电子科技大学 | Modulation Identification method based on least square method supporting vector machine |
Non-Patent Citations (2)
| Title |
|---|
| 樊伟: "通信信号自动调制识别中的分类器设计", 《中国优秀博硕士学位论文全文数据库（硕士） 信息科技辑》 * |
| 赵自璐: "水声通信信号调制方式识别技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111612130A (en) * | 2020-05-18 | 2020-09-01 | 吉林大学 | Frequency shift keying communication signal modulation mode identification method |
| CN111612130B (en) * | 2020-05-18 | 2022-12-23 | 吉林大学 | Frequency shift keying communication signal modulation mode identification method |
| CN112737992A (en) * | 2020-09-23 | 2021-04-30 | 青岛科技大学 | Underwater sound signal modulation mode self-adaptive in-class identification method |
| CN112737992B (en) * | 2020-09-23 | 2023-02-03 | 青岛科技大学 | Underwater sound signal modulation mode self-adaptive in-class identification method |
| CN112800862A (en) * | 2021-01-11 | 2021-05-14 | 吉林大学 | Non-stationary signal time-frequency matrix reconstruction method and system |
| CN112800862B (en) * | 2021-01-11 | 2022-08-02 | 吉林大学 | Non-stationary signal time-frequency matrix reconstruction method and system |
| CN113452637A (en) * | 2021-09-01 | 2021-09-28 | 中国海洋大学 | Underwater acoustic communication signal modulation identification method based on feature selection and support vector machine |
| CN114004982A (en) * | 2021-10-27 | 2022-02-01 | 中国科学院声学研究所 | Acoustic Haar feature extraction method and system for underwater target recognition |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111010356A (en) | Underwater acoustic communication signal modulation mode identification method based on support vector machine | |
| CN109446877B (en) | Radar radiation source signal modulation identification method combined with multi-dimensional feature migration fusion | |
| CN107220606B (en) | Radar radiation source signal identification method based on one-dimensional convolutional neural network | |
| CN110443293B (en) | Zero sample image classification method for generating confrontation network text reconstruction based on double discrimination | |
| CN109495214B (en) | Channel coding type identification method based on one-dimensional inclusion structure | |
| CN109890043B (en) | Wireless signal noise reduction method based on generative countermeasure network | |
| CN111723701B (en) | Underwater target identification method | |
| CN108600135A (en) | A kind of recognition methods of signal modulation mode | |
| CN106529428A (en) | Underwater target recognition method based on deep learning | |
| CN113259288B (en) | Underwater sound modulation mode identification method based on feature fusion and lightweight hybrid model | |
| CN111222442A (en) | Electromagnetic signal classification method and device | |
| CN111680737B (en) | Radar radiation source individual identification method under differential signal-to-noise ratio condition | |
| CN113111786B (en) | Underwater target identification method based on small sample training diagram convolutional network | |
| CN114595732A (en) | Radar radiation source sorting method based on depth clustering | |
| CN114897023A (en) | Underwater sound target identification method based on underwater sound target sensitivity difference feature extraction | |
| CN112347910A (en) | Signal fingerprint identification method based on multi-mode deep learning | |
| CN116047427A (en) | Small sample radar active interference identification method | |
| CN115982613A (en) | Signal modulation identification system and method based on improved convolutional neural network | |
| CN113780521B (en) | Radiation source individual identification method based on deep learning | |
| CN114757224A (en) | Specific radiation source identification method based on continuous learning and combined feature extraction | |
| Limin et al. | Low probability of intercept radar signal recognition based on the improved AlexNet model | |
| CN116383719A (en) | MGF radio frequency fingerprint identification method for LFM radar | |
| CN115563480A (en) | Gear fault identification method for screening octave geometric modal decomposition based on kurtosis ratio coefficient | |
| CN110555483B (en) | Polarized SAR classification method based on FW-DCGAN feature generation | |
| CN114520758A (en) | Signal modulation identification method based on instantaneous characteristics |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200414 |