CN112380931B - Modulation signal classification method and system based on sub-graph network - Google Patents

Modulation signal classification method and system based on sub-graph network Download PDF

Info

Publication number
CN112380931B
CN112380931B CN202011196212.3A CN202011196212A CN112380931B CN 112380931 B CN112380931 B CN 112380931B CN 202011196212 A CN202011196212 A CN 202011196212A CN 112380931 B CN112380931 B CN 112380931B
Authority
CN
China
Prior art keywords
network
modulation signal
signal
weighted directed
visual
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.)
Active
Application number
CN202011196212.3A
Other languages
Chinese (zh)
Other versions
CN112380931A (en
Inventor
宣琦
周锦超
裘坤锋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University of Technology ZJUT
Original Assignee
Zhejiang University of Technology ZJUT
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhejiang University of Technology ZJUT filed Critical Zhejiang University of Technology ZJUT
Priority to CN202011196212.3A priority Critical patent/CN112380931B/en
Publication of CN112380931A publication Critical patent/CN112380931A/en
Application granted granted Critical
Publication of CN112380931B publication Critical patent/CN112380931B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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
    • G06F2218/10—Feature extraction by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
    • 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
    • G06F2218/16—Classification; Matching by matching signal segments
    • G06F2218/18—Classification; Matching by matching signal segments by plotting the signal segments against each other, e.g. analysing scattergrams
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Abstract

A modulated signal classification method based on a sub-graph network, comprising: s1, converting a dual-channel I/Q modulation signal into a four-channel signal, and carrying out peak detection processing; s2, respectively converting the four-channel signals into weighted directed network graphs, and respectively mapping the four weighted directed network graphs into first-order sub-networks; s3, respectively carrying out feature extraction on the four weighted directed network graphs and the four first-order sub-networks to obtain eight feature vectors, and carrying out space expansion on the feature vectors to obtain a fusion feature vector of each I/Q modulation signal; s4, training the modulation signal classification model, wherein the classification precision is smaller than a preset threshold value, adjusting the super-parameters in the circular system limited crossing visual view networking method, and repeating the steps S2-S3 until the classification precision is larger than or equal to the preset threshold value, and completing the classification of the I/Q modulation signals through the trained classification model. The invention also provides a modulation signal classification system based on the sub-graph network.

Description

Modulation signal classification method and system based on sub-graph network
Technical Field
The invention relates to the technical field of I/Q modulation signal classification, in particular to a modulation signal classification method and system based on a sub-graph network.
Background
Aiming at the problem of modulation signal classification, in the prior art, an I/Q modulation signal with lower signal-to-noise ratio cannot be identified, for example, a signal modulation mode below-10 dB. In order to solve the technical problem, researchers couple the time sequence with the complex network field, and the structural characteristics of the time sequence are mined through researching the topological structure of the network, so that the purpose of signal classification is achieved, but how to map the time sequence into the complex network is a problem which needs to be solved in a key way in the time sequence research of the direction. There are many methods in the current academy of analyzing time series by visual algorithms, such as visual VG, horizontal visual HVG, and limited crossing visual LPVG, and experiments prove that these methods of creating a network map from time series can extract and retain some of the basic features of the time series. However, the network map obtained by these networking algorithms is relatively single, and cannot be built from a time sequence according to different tasks or user requirements, so that the network map containing more effective information can not be built. Meanwhile, none of the prior art applies a visual algorithm to the classification of radio modulated signals.
Therefore, there is a need for a method and system for classifying I/Q modulated signals quickly and accurately.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, and provides a modulation signal classification method and a modulation signal classification system based on a sub-graph network, which are used for solving the technical problems in the prior art and can be used for rapidly and accurately classifying modulation signals.
In order to achieve the above object, the present invention provides the following solutions: a modulation signal classification method based on a sub-graph network comprises the following steps:
s1, acquiring an I/Q modulation signal, converting the two-channel I/Q modulation signal into a four-channel signal, and carrying out peak detection processing on the four-channel signal;
s2, converting the four-channel signals into weighted directed network diagrams respectively by adopting a circular finite-crossing visual network building method, and mapping the four weighted directed network diagrams into first-order sub-networks respectively;
s3, respectively carrying out feature extraction on the four weighted directed network graphs and the four first-order sub-networks to obtain eight feature vectors, and carrying out space expansion on the feature vectors to obtain a fusion feature vector of each I/Q modulation signal;
s4, constructing a modulation signal classification model based on a limit random tree, training the modulation signal classification model by adopting the fusion feature vector, wherein the classification precision is smaller than a preset threshold value, adjusting super parameters in a circular system limited crossing visual view networking method, and repeating the steps S2-S3 until the classification precision is greater than or equal to the preset threshold value to obtain a trained modulation signal classification model, and completing classification of I/Q modulation signals through the trained classification model.
Preferably, the specific method for performing peak detection processing on the four-channel signal in step S1 includes:
setting a super parameter v, which is used for controlling the size of a sliding window, adding v 0 values at the beginning and the end of a signal S, and expanding the signal into: s is S 1 ＝{0，0，...，0，y 1 ，y 2 ，...，y N ，0，0，...，0}；
Will y k (k∈[1，N]) Signal segment { y with length v+1 on left k-v ,…,y k-1 ,y k The maximum value in } is defined as y left-max Will y k (k∈[1，N]) Signal segment { y with length v+1 on right k ,y k+1 ,…,y k+v The maximum value in } is defined as y right-max If y k Satisfy (y) left-max +y right-max )/2≤y k The point is kept as the local maximum peak, otherwise the point is deleted.
Preferably, step S2 adopts a circular finite-crossing visual network building method to convert the four-way signals into weighted directed network graphs, which specifically includes: the time sequence y= { Y i } i＝1,2,...,N Selecting points separated by l data points as nodes of the weighted directed network graph, wherein N represents the number of the data points in the time sequence Y; two nodes (t) m ,y m ) And (t) n ,y n ) The limited crossing sight distance between the two points is L, the actual crossing sight distance is M, wherein L represents the maximum cut-off number allowed between the two points, M represents the actual cut-off number between the two points, t m Representing data point y m Corresponding time point, t n Representing data point y n A corresponding point in time; if the two nodes meet the visibility criterion, the connection line of the two nodes forms the connection edge of the weighted directed network graph, and the weight of the connection edge is the actual crossing line of sight M+1.
Wherein the visibility criterion is described as:
with nodes (t) m ,y m ) And (t) n ,y n ) Wire section of (1) is a chordThe circular arc is taken as a visible line, as shown in formula (3):
f(t,y')＝(t-t m )(t-t n )+(y'-y m )(y'-y n )+a[(t-t m )(y n -y m )-(y'-y m )(t n -t m )]＝0…(3)
wherein t represents time, y' represents the value of a data point on the visual line, alpha represents a super parameter, the size of alpha is adjustable, when alpha is more than 0, the visual line is concave, and when alpha is less than 0, the visual line is convex.
By adjusting the alpha, arc visual lines with different radians can be obtained; calculating the time point t according to formula (3) k Value y 'on the corresponding visual line of (m < k < n)' k The method comprises the steps of carrying out a first treatment on the surface of the Due to y' k At time t k There are two solutions, to ensure visual meaning, the solution on the arc less than 180 ° is chosen to be y' k ，
If node (t) m ,y m ) And (t) n ,y n ) Between which there are K data points (t k ,y k ) K is more than or equal to 0 and less than or equal to L, and y is satisfied k >y' k The remaining l-K data points (t k ,y k ) Satisfy y k <y' k The visibility criterion is satisfied.
Preferably, the method of mapping the four weighted directed network graphs into first-order sub-networks in step S2 includes:
and mapping all the connected edges in the weighted directed network graph into different nodes in the first-order sub-graph network, and if two edges in the weighted directed network graph share the same node, connecting the edges between the two nodes in the weighted directed network graph to form the first-order sub-graph network.
Preferably, in step S3, a Graph2vec automatic feature extraction method is adopted to perform feature extraction on the four weighted directed network graphs and the four first-order sub-networks.
The system for implementing the modulation signal classification method based on the sub-graph network comprises a data acquisition and processing unit, a network graph construction unit, a characteristic extraction unit and a modulation signal classification unit which are sequentially connected and unidirectionally feed data, wherein the modulation signal classification unit feeds a super-parameter adjustment control signal to the network graph construction unit, and the network graph construction unit, the characteristic extraction unit and the modulation signal classification unit form a cycle;
the data acquisition and processing unit is used for acquiring I/Q modulation signals, converting the two-channel I/Q modulation signals into four-channel signals, carrying out peak detection processing on the four-channel signals and then transmitting the four-channel signals to the network diagram construction unit;
the network diagram construction unit adopts a circular finite-crossing visual diagram construction method to respectively convert the four-channel signals after peak detection processing into weighted directed network diagrams, respectively map the four weighted directed network diagrams into first-order sub-networks, and transmit the four weighted directed network diagrams and the corresponding four first-order sub-networks to the feature extraction unit; the network diagram construction unit can also adjust super parameters of the circular finite-crossing visual network construction method according to control signals;
the feature extraction unit is used for extracting features of the four weighted directed network graphs and the four first-order sub-networks to obtain eight feature vectors, performing spatial expansion on the feature vectors to obtain a fusion feature vector of each I/Q modulation signal, and transmitting the fusion feature vector of each I/Q modulation signal to the modulation signal classification unit;
the modulating signal classifying unit builds a modulating signal classifying model based on a limit random tree, trains the modulating signal classifying model by adopting the fusion feature vector, transmits a super-parameter adjustment control signal to the network diagram building unit until the classifying precision is larger than or equal to a preset threshold value, obtains a trained modulating signal classifying model, and classifies the I/Q modulating signals through the trained modulating signal classifying model.
The invention discloses the following technical effects:
(1) According to the invention, a round-system limited-crossing visual network construction method is adopted to convert an I/Q modulation signal time sequence classification task into a network diagram classification task, and in the training process of a classification model, a network diagram containing more effective information is obtained from the I/Q modulation signal time sequence by adjusting super parameter setting in the round-system limited-crossing visual network construction method, so that potential structure information of the network diagram can be fully utilized, the classification precision of the I/Q modulation signal is improved, and the modulation type of the signal can be accurately classified from the I/Q modulation signal, thereby providing a technical foundation for subsequent processing of the signal and obtaining information in the signal; meanwhile, by adjusting the setting of the super parameters, the method can be suitable for I/Q modulation signals of different types, and the flexibility is greatly improved;
(2) The invention uses the sub-graph network SGN to expand the network feature space, acquires the hidden information in the original network graph from the basic structure of the sub-graph, fully utilizes the potential structural information of the network graph, and further improves the classification precision;
(3) The invention applies the improved visual algorithm to the task of classifying the modulation signal sequence, reduces the signal scale through the peak detection algorithm in the preprocessing stage, and effectively shortens the operation time on the premise of ensuring small change of precision.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic diagram of a modulated signal classification system based on a sub-graph network according to the present invention;
FIG. 2 is a flow chart of a method for classifying modulated signals based on a sub-graph network according to the present invention;
FIG. 3 is a schematic diagram of a circular finite-crossing visual networking in accordance with an embodiment of the present invention;
FIG. 4 is a schematic diagram of an arc with two points connected by a chord according to an embodiment of the present invention;
fig. 5 is a schematic diagram of a first-order sub-graph network constructed in an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
Referring to fig. 1, the present embodiment provides a modulated signal classification system based on a sub-graph network, for classifying I/Q modulated signals in a communication process, including: the system comprises a data acquisition and processing unit, a network diagram construction unit, a characteristic extraction unit and a modulation signal classification unit; the data acquisition and processing unit, the network diagram construction unit, the characteristic extraction unit and the modulation signal classification unit are sequentially connected and feed data unidirectionally, and the modulation signal classification unit feeds a super-parameter adjustment control signal with the network diagram construction unit, so that the network diagram construction unit, the characteristic extraction unit and the modulation signal classification unit form a cycle;
the data acquisition and processing unit is used for acquiring I/Q modulation signals, converting the acquired I/Q modulation signals into four-channel signals by processing the acquired I/Q modulation signals, carrying out peak detection processing on the four-channel signals, and transmitting the four-channel signals to the network diagram construction unit;
in this embodiment, the I/Q modulated signal in the training sample is obtained from the Radio ml 2016.10a data set, where the Radio ml 2016.10a data set is a synthetic data set generated by GNU Radio, and the data set is first released on the sixth GNU annual Radio conference; in the actual classification process, the data acquisition and processing unit receives an aerial radio signal through a receiving antenna of a radio receiver, and then carries out filtering and IQ demodulation to obtain an I/Q modulation signal.
The I/Q modulation signal is a double-channel time sequence comprising an I channel and a Q channel; the specific method for converting the I/Q modulation signal into the four-channel signal comprises the following steps:
processing the time sequence of each I/Q modulation signal I channel and Q channel to obtain amplitude data A and phase data W of each I/Q modulation signal, thereby obtaining the time sequence of four channels of each I/Q modulation signal, wherein the four channels are respectively an I channel, a Q channel, an A channel and a W channel.
The calculation of the amplitude data A is shown in the formula (1):
wherein A is i Amplitude data representing the ith time point in channel A, I i Signal data representing the I-th time point in the I-channel, Q i Signal data representing an i-th point in time in the Q channel;
the phase data W is calculated by taking an I channel as an abscissa and a Q channel as an ordinate, and is shown in the formula (2):
in which W is i Phase data representing the ith time point in the W channel, I i Signal data representing the I-th time point in the I-channel, Q i Signal data representing the i-th point in time in the Q channel.
The specific method for carrying out peak detection processing on the four-channel signal comprises the following steps:
setting a super parameter v, which is used for controlling the size of a sliding window, adding v 0 values at the beginning and the end of a signal S, and expanding the signal into: s is S 1 ＝{0，0，...，0，y 1 ，y 2 ，...，y N ，0，0，...，0}；
Will y k (k∈[1，N]) Signal segment { y with length v+1 on left k-v ,…,y k-1 ,y k In }The maximum value is defined as y left-max Will y k (k∈[1，N]) Signal segment { y with length v+1 on right k ,y k+1 ,…,y k+v The maximum value in } is defined as y right-max If y k Satisfy (y) left-max +y right-max )/2≤y k And if not, deleting the point to reduce the signal scale of each channel, thereby effectively shortening the operation time.
The network diagram construction unit adopts a circular finite-crossing visual diagram construction method to respectively convert the four-channel signals after peak detection processing into weighted directed network diagrams, respectively map the four weighted directed network diagrams into first-order sub-networks, and transmit the four weighted directed network diagrams and the corresponding four first-order sub-networks to the feature extraction unit; the network diagram construction unit can also adjust super parameters of the circular finite-crossing visual network construction method according to control signals;
in this embodiment, the network map construction unit converts the time sequences of the I channel, the Q channel, the a channel, and the W channel into weighted directed network maps respectively by using a circular finite-crossing visual network construction methodAndthe nodes of the weighted directed network graph are mapped by time points of the time sequence, the directed connecting edge of the weighted directed network graph is formed by adding one to the round system visible line cut-off times, wherein the node with a larger signal value points to the node with a smaller signal value, and the connecting edge weight of the weighted directed network graph is formed by adding one to the round system visible line cut-off times, and the situation that the weight is 0 can be effectively avoided. The specific method for converting the four-channel signals into weighted directed network diagrams respectively comprises the following steps:
the time sequence y= { Y i } i＝1,2,...,N Selecting points separated by one data point as the weighted directed networkNodes of the graph, wherein N represents the number of data points in the time sequence Y; two nodes (t) m ,y m ) And (t) n ,y n ) The limited crossing sight distance between the two points is L, the actual crossing sight distance is M, wherein L represents the maximum cut-off number allowed between the two points, M represents the actual cut-off number between the two points, t m Representing data point y m Corresponding time point, t n Representing data point y n A corresponding point in time; if the two nodes meet the visibility criterion, the connection line of the two nodes forms the connection edge of the weighted directed network graph, and the weight of the connection edge is the actual crossing line of sight m+1, as shown in fig. 3.
Wherein the visibility criterion is described as:
as shown in fig. 4, a node (t m ,y m ) And (t) n ,y n ) The connecting line segment of (2) is a chord and is provided with an arc as a visual line, as shown in the formula (3):
f(t,y')＝(t-t m )(t-t n )+(y'-y m )(y'-y n )+a[(t-t m )(y n -y m )-(y'-y m )(t n -t m )]＝0…(3)
wherein t represents time, y' represents the value of a data point on the visual line, alpha represents a super parameter, the size of alpha is adjustable, when alpha is more than 0, the visual line is concave, and when alpha is less than 0, the visual line is convex. In fig. 4, the dotted line indicates the more edges than the visual view VG, and the thick dotted line indicates the more edges than the limited-crossing visual view LPVG.
By adjusting the alpha, arc visual lines with different radians can be obtained; calculating the time point t according to formula (3) k Value y 'on the corresponding visual line of (m < k < n)' k The method comprises the steps of carrying out a first treatment on the surface of the Due to y' k At time t k There are two solutions, to ensure visual meaning, the solution on the arc less than 180 ° is chosen to be y' k ；
If node (t) m ,y m ) And (t) n ,y n ) Between which there are K data points (t k ,y k ) K is more than or equal to 0 and less than or equal to L, and y is satisfied k >y' k The remaining l-K data points (t k ,y k ) Satisfy y k <y' k I.e. at most not more than L time points are truncated, none of the remaining L-K time points are allowed to be truncated, the visibility criterion is fulfilled.
The method for mapping the weighted directed network graph to the first-order sub-network respectively is shown in fig. 5, and includes:
mapping all the connected edges in the weighted directed network graph to different nodes in the first-order sub-graph, and if two edges in the weighted directed network graph share the same node, connecting edges between two nodes in the weighted directed network graph to form the first-order sub-graph network, thereby connecting the four weighted directed network graphsAnd->Mapping into four of said first-order sub-networks +.>And->
The feature extraction unit is used for extracting features of the four weighted directed network graphs and the four first-order sub-networks to obtain eight feature vectors, performing spatial expansion on the feature vectors to obtain a fusion feature vector of each I/Q modulation signal, and transmitting the fusion feature vector of each I/Q modulation signal to the modulation signal classification unit;
the feature extraction unit of this embodiment adopts Graph2vec automatic feature extraction method to respectively weight four directed network graphs of each I/Q modulation signal And->Four first-order sub-networksAnd->Extracting features to obtain eight K-dimensional feature vectors And->Wherein R is K A matrix representing the K dimension; performing feature space expansion on eight K-dimensional feature vectors obtained by each I/Q modulation signal in a transverse splicing mode to obtain a fusion feature vector of each I/Q modulation signalWhere merge () represents merging.
The modulating signal classifying unit builds a modulating signal classifying model based on a limit random tree, trains the modulating signal classifying model by adopting the fusion feature vector, transmits a super-parameter adjustment control signal to the network diagram building unit until the classifying precision is larger than or equal to a preset threshold value, obtains a trained modulating signal classifying model, and classifies the I/Q modulating signals through the trained modulating signal classifying model.
Referring to fig. 2, the embodiment further provides a modulation signal classification method based on a sub-graph network, which specifically includes the following steps:
s1, acquiring an I/Q modulation signal, converting the acquired I/Q modulation signal into a four-channel signal by processing the acquired I/Q modulation signal, and carrying out peak detection processing on the four-channel signal;
the I/Q modulation signal is a double-channel time sequence comprising an I channel and a Q channel; the specific method for converting the dual-channel I/Q modulation signal into the four-channel signal comprises the following steps:
processing the time sequence of each I/Q modulation signal I channel and Q channel to obtain amplitude data A and phase data W of each I/Q modulation signal, thereby obtaining the time sequence of four channels of each I/Q modulation signal, wherein the four channels are respectively an I channel, a Q channel, an A channel and a W channel.
The calculation of the amplitude data A is shown in the formula (1):
wherein A is i Amplitude data representing the ith time point in channel A, I i Signal data representing the I-th time point in the I-channel, Q i Signal data representing an i-th point in time in the Q channel; the method comprises the steps of carrying out a first treatment on the surface of the
The phase data W is calculated by taking an I channel as an abscissa and a Q channel as an ordinate, and is shown in the formula (2):
in which W is i Phase data representing the ith time point in the W channel, I i Signal data representing the I-th time point in the I-channel, Q i Signal data representing the i-th point in time in the Q channel.
The specific method for carrying out peak detection processing on the four-channel signal comprises the following steps:
setting a super parameter v, which is used for controlling the size of a sliding window, adding v 0 values at the beginning and the end of a signal S, and expanding the signal into: s is S 1 ＝{0，0，...，0，y 1 ，y 2 ，...，y N ，0，0，...，0}；
Will y k (k∈[1，N]) Signal segment { y with length v+1 on left k-v ,…,y k-1 ,y k The maximum value in } is defined as y left-max Will y k (k∈[1，N]) Signal segment { y with length v+1 on right k ,y k+1 ,…,y k+v The maximum value in } is defined as y right-max If y k Satisfy (y) left-max +y right-max )/2≤y k And if not, deleting the point to reduce the signal scale of each channel, thereby effectively shortening the operation time.
S2, converting the four-channel signals into weighted directed network diagrams respectively by adopting a circular finite-crossing visual network building method, and mapping the four weighted directed network diagrams into first-order sub-networks respectively;
the nodes of the weighted directed network graph are mapped by time points of the time sequence, the directed connecting edge of the weighted directed network graph is formed by adding one to the round system visible line cut-off times formed by two nodes, wherein the node with a larger signal value points to the node with a smaller signal value, and the connecting edge weight of the weighted directed network graph is obtained by adding one to effectively avoid the condition that the weight is 0.
The specific method for converting the four-channel signals into the weighted directed network diagram comprises the following steps of:
the time sequence y= { Y i } i＝1,2,...,N Selecting points separated by l data points as nodes of the weighted directed network graph, wherein N represents the number of the data points in the time sequence Y; two nodes (t) m ,y m ) And (t) n ,y n ) The limited crossing sight distance between the two points is L, the actual crossing sight distance is M, wherein L represents the maximum cut-off number allowed between the two points, M represents the actual cut-off number between the two points, t m Representing data point y m Corresponding time point, t n Representing data point y n A corresponding point in time; if the two nodes meet the visibility criterion, the connection line of the two nodes forms the connection edge of the weighted directed network graph, and the weight of the connection edge is the actual crossing line of sight M+1.
Wherein the visibility criterion is described as:
with nodes (t) m ,y m ) And (t) n ,y n ) The connecting line segment of (2) is a chord and is provided with an arc as a visual line, as shown in the formula (3):
f(t,y')＝(t-t m )(t-t n )+(y'-y m )(y'-y n )+a[(t-t m )(y n -y m )-(y'-y m )(t n -t m )]＝0…(3)
wherein t represents time, y' represents the value of a data point on the visual line, alpha represents a super parameter, the size of alpha is adjustable, when alpha is more than 0, the visual line is concave, and when alpha is less than 0, the visual line is convex.
By adjusting the alpha, arc visual lines with different radians can be obtained; calculating the time point t according to formula (3) k Value y 'on the corresponding visual line of (m < k < n)' k The method comprises the steps of carrying out a first treatment on the surface of the Due to y' k At time t k There are two solutions, to ensure visual meaning, the solution on the arc less than 180 ° is chosen to be y' k ，
If node (t) m ,y m ) And (t) n ,y n ) Between which there are K data points (t k ,y k ) K is more than or equal to 0 and less than or equal to L, and y is satisfied k >y' k The remaining l-K data points (t k ,y k ) Satisfy y k <y' k The visibility criterion is satisfied.
The method for mapping the weighted directed network graph into the first-order sub-network respectively comprises the following steps:
mapping all the connected edges in the weighted directed network graph to different nodes in the first-order sub-graph, and if two edges in the weighted directed network graph share the same node, connecting edges between two nodes in the weighted directed network graph to form the first-order sub-graph network, thereby connecting the four weighted directed network graphsAnd->Mapping into four of said first-order sub-networks +.>And->
S3, respectively carrying out feature extraction on the four weighted directed network graphs and the four first-order sub-networks to obtain eight feature vectors, and carrying out space expansion on the feature vectors to obtain a fusion feature vector of each I/Q modulation signal;
in this embodiment, a Graph2vec automatic feature extraction method is adopted to respectively implement four weighted directed network graphs for each of the I/Q modulated signalsAnd->Four first-order sub-networksAnd->Extracting features to obtain eight K-dimensional feature vectors And->Wherein R is K A matrix representing the K dimension; performing feature space expansion on eight K-dimensional feature vectors obtained by each I/Q modulation signal in a transverse splicing mode to obtain a fusion feature vector of each I/Q modulation signalWhere merge () represents merging.
S4, constructing a modulation signal classification model based on a limit random tree, training the modulation signal classification model by adopting the fusion feature vector, wherein the classification precision is smaller than a preset threshold value, adjusting super parameters in a circular system limited crossing visual view networking method, and repeating the steps S2-S3 until the classification precision is greater than or equal to the preset threshold value to obtain a trained modulation signal classification model, and completing classification of I/Q modulation signals through the trained classification model.
The above embodiments are only illustrative of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention, and various modifications and improvements made by those skilled in the art to the technical solutions of the present invention should fall within the protection scope defined by the claims of the present invention without departing from the design spirit of the present invention.

Claims (5)

1. A modulation signal classification method based on a sub-graph network is characterized in that: the method comprises the following steps:
s1, acquiring an I/Q modulation signal, converting the two-channel I/Q modulation signal into a four-channel signal, and carrying out peak detection processing on the four-channel signal;
s2, converting the four-channel signals into weighted directed network diagrams respectively by adopting a circular finite-crossing visual network building method, and mapping the four weighted directed network diagrams into first-order sub-networks respectively; the method specifically comprises the following steps: the time sequence y= { Y i } i＝1,2,...,N Selecting points separated by l data points as nodes of the weighted directed network graph, wherein N represents the number of the data points in the time sequence Y; two nodes (t) m ,y m ) And (t) n ,y n ) The limited crossing sight distance between the two points is L, the actual crossing sight distance is M, wherein L represents the maximum cut-off number allowed between the two points, M represents the actual cut-off number between the two points, t m Representing data point y m Corresponding time point, t n Representing data point y n A corresponding point in time; if the two nodes meet the visibility criterion, connecting lines of the two nodes form connecting edges of the weighted directed network graph, and the weight of the connecting edges is the actual crossing visual distance M+1;
wherein the visibility criterion is described as:
with nodes (t) m ,y m ) And (t) n ,y n ) The connecting line segment of (2) is a chord and is provided with an arc as a visual line, as shown in the formula (3):
f(t,y')＝(t-t m )(t-t n )+(y'-y m )(y'-y n )+a[(t-t m )(y n -y m )-(y'-y m )(t n -t m )]＝0…(3)
wherein t represents time, y' represents the value of a data point on the visual line, alpha represents a super parameter, the size of alpha is adjustable, when alpha is more than 0, the visual line is concave, and when alpha is less than 0, the visual line is convex;
by adjusting the alpha, arc visual lines with different radians can be obtained; calculating the time point t according to formula (3) k The value y 'on the corresponding visual line' k Wherein m is<k<n; due to y' k At time t k There are two solutions, to ensure visual meaning, the solution on the arc less than 180 ° is chosen to be y' k If node (t) m ，y m ) And (t) n ，y k ) Between which there are K data points (t k ，y k ) K is more than or equal to 0 and less than or equal to L, and y is satisfied k >y′ k The remaining l-K data points (t k ，y k ) Satisfy y k <y′ k Then the visibility criterion is satisfied;
s3, respectively carrying out feature extraction on the four weighted directed network graphs and the four first-order sub-networks to obtain eight feature vectors, and carrying out space expansion on the feature vectors to obtain a fusion feature vector of each I/Q modulation signal;
s4, constructing a modulation signal classification model based on a limit random tree, training the modulation signal classification model by adopting the fusion feature vector, wherein the classification precision is smaller than a preset threshold value, adjusting super parameters in a circular system limited crossing visual view networking method, and repeating the steps S2-S3 until the classification precision is greater than or equal to the preset threshold value to obtain a trained modulation signal classification model, and completing classification of I/Q modulation signals through the trained classification model.
2. The sub-picture network based modulation signal classification method according to claim 1, wherein: the specific method for carrying out peak detection processing on the four-channel signal in the step s1 comprises the following steps:
setting a super parameter v, which is used for controlling the size of a sliding window, adding v 0 values at the beginning and the end of a signal s, and expanding the signal into: s is(s) 1 ＝{0，0，...，0，y 1 ，y 2 ，...，y N ，0，0，...，0}；
Will y k ，k∈[1，N]Signal segment { y with length v+1 on left k-v ，…，y k-1 ，y k The maximum value in } is defined as y left-max Will y k ，k∈[1，N]Signal segment { y with length v+1 on right k ，y k+1 ，…，y k+v The maximum value in } is defined as y right-max If y k Satisfy (y) left-max +y right-max )/2≤y k The point is kept as the local maximum peak, otherwise the point is deleted.
3. The sub-picture network based modulation signal classification method according to claim 1, wherein: the step S2 of mapping the four weighted directed network graphs into first-order sub-networks respectively comprises the following steps:
and mapping all the connected edges in the weighted directed network graph into different nodes in the first-order sub-graph network, and if two edges in the weighted directed network graph share the same node, connecting the edges between the two nodes in the weighted directed network graph to form the first-order sub-graph network.
4. The sub-picture network based modulation signal classification method according to claim 1, wherein: and step S3, the feature extraction unit adopts a Graph2vec automatic feature extraction method to extract features of the four weighted directed network graphs and the four first-order sub-networks.
5. A modulating signal classifying system based on a sub-graph network is characterized in that a data acquisition and processing unit, a network graph constructing unit, a characteristic extracting unit and a modulating signal classifying unit are sequentially connected and feed data unidirectionally, the modulating signal classifying unit feeds a super-parameter adjusting control signal to the network graph constructing unit, and the network graph constructing unit, the characteristic extracting unit and the modulating signal classifying unit form a cycle;
the data acquisition and processing unit is used for acquiring I/Q modulation signals, converting the two-channel I/Q modulation signals into four-channel signals, carrying out peak detection processing on the four-channel signals and then transmitting the four-channel signals to the network diagram construction unit;
the network diagram construction unit adopts a circular finite-crossing visual diagram construction method to respectively convert the four-channel signals after peak detection processing into weighted directed network diagrams, respectively map the four weighted directed network diagrams into first-order sub-networks, and transmit the four weighted directed network diagrams and the corresponding four first-order sub-networks to the feature extraction unit; the network diagram construction unit can also adjust super parameters of the circular finite-crossing visual network construction method according to control signals; the method specifically comprises the following steps: the time sequence y= { Y i } i＝1,2,...,N Selecting points separated by l data points as nodes of the weighted directed network graph, wherein N represents the number of the data points in the time sequence Y; two nodes (t) m ,y m ) And (t) n ,y n ) The limited crossing sight distance between the two points is L, the actual crossing sight distance is M, wherein L represents the maximum cut-off number allowed between the two points, M represents the actual cut-off number between the two points, t m Representing data point y m Corresponding time point, t n Representing data point y n A corresponding point in time; if the two nodes meet the visibility criterion, connecting lines of the two nodes form a connecting edge of the weighted directed network graph, and the weight of the connecting edge is realThe inter-crossing visual distance M+1;
wherein the visibility criterion is described as:
with nodes (t) m ,y m ) And (t) n ,y n ) The connecting line segment of (2) is a chord and is provided with an arc as a visual line, as shown in the formula (3):
f(t,y')＝(t-t m )(t-t n )+(y'-y m )(y'-y n )+a[(t-t m )(y n -y m )-(y'-y m )(t n -t m )]＝0…(3)
wherein t represents time, y' represents the value of a data point on the visual line, alpha represents a super parameter, the size of alpha is adjustable, when alpha is more than 0, the visual line is concave, and when alpha is less than 0, the visual line is convex;
by adjusting the alpha, arc visual lines with different radians can be obtained; calculating the time point t according to formula (3) k The value y 'on the corresponding visual line' k Wherein m < k < n; due to y' k At time t k There are two solutions, to ensure visual meaning, the solution on the arc less than 180 ° is chosen to be y' k If node (t) m ,y m ) And (t) n ,y n ) Between which there are K data points (t k ,y k ) K is more than or equal to 0 and less than or equal to L, and y is satisfied k >y' k The remaining l-K data points (t k ,y k ) Satisfy y k ＜y' k Then the visibility criterion is satisfied;
the feature extraction unit is used for extracting features of the four weighted directed network graphs and the four first-order sub-networks to obtain eight feature vectors, performing spatial expansion on the feature vectors to obtain a fusion feature vector of each I/Q modulation signal, and transmitting the fusion feature vector of each I/Q modulation signal to the modulation signal classification unit;
the modulating signal classifying unit builds a modulating signal classifying model based on a limit random tree, trains the modulating signal classifying model by adopting the fusion feature vector, transmits a super-parameter adjustment control signal to the network diagram building unit until the classifying precision is larger than or equal to a preset threshold value, obtains a trained modulating signal classifying model, and classifies the I/Q modulating signals through the trained modulating signal classifying model.
CN202011196212.3A 2020-10-30 2020-10-30 Modulation signal classification method and system based on sub-graph network Active CN112380931B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011196212.3A CN112380931B (en) 2020-10-30 2020-10-30 Modulation signal classification method and system based on sub-graph network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011196212.3A CN112380931B (en) 2020-10-30 2020-10-30 Modulation signal classification method and system based on sub-graph network

Publications (2)

Publication Number Publication Date
CN112380931A CN112380931A (en) 2021-02-19
CN112380931B true CN112380931B (en) 2024-02-20

Family

ID=74576517

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011196212.3A Active CN112380931B (en) 2020-10-30 2020-10-30 Modulation signal classification method and system based on sub-graph network

Country Status (1)

Country Link
CN (1) CN112380931B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113111914A (en) * 2021-03-12 2021-07-13 浙江工业大学 Graph width learning classification method and system based on global sampling subgraph
CN113392731B (en) * 2021-05-31 2023-06-23 浙江工业大学 Modulation signal classification method and system based on graph neural network

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108764077A (en) * 2018-05-15 2018-11-06 重庆邮电大学 A kind of digital signal modulated sorting technique based on convolutional neural networks
CN110367933A (en) * 2019-07-15 2019-10-25 天津大学 Sleep stage classification method and application based on complex network and deep learning
CN110956199A (en) * 2019-11-05 2020-04-03 浙江工业大学 Node classification method based on sampling subgraph network
CN111599170A (en) * 2020-04-13 2020-08-28 浙江工业大学 Traffic running state classification method based on time sequence traffic network diagram

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108764077A (en) * 2018-05-15 2018-11-06 重庆邮电大学 A kind of digital signal modulated sorting technique based on convolutional neural networks
CN110367933A (en) * 2019-07-15 2019-10-25 天津大学 Sleep stage classification method and application based on complex network and deep learning
CN110956199A (en) * 2019-11-05 2020-04-03 浙江工业大学 Node classification method based on sampling subgraph network
CN111599170A (en) * 2020-04-13 2020-08-28 浙江工业大学 Traffic running state classification method based on time sequence traffic network diagram

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
"基于有限穿越可视图的时间序列网络模型";周婷婷等;《物理学报》;第1-6页 *

Also Published As

Publication number Publication date
CN112380931A (en) 2021-02-19

Similar Documents

Publication Publication Date Title
CN112380931B (en) Modulation signal classification method and system based on sub-graph network
WO2018176889A1 (en) Method for automatically identifying modulation mode for digital communication signal
CN106023177A (en) Thunderstorm cloud cluster identification method and system for meteorological satellite cloud picture
CN108173599B (en) Stokes spatial coherent light modulation format identification method based on DENC L UE clustering
CN112380928B (en) Modulation signal classification method and system based on circular finite-crossing visual networking
CN113392731B (en) Modulation signal classification method and system based on graph neural network
CN103824070A (en) Rapid pedestrian detection method based on computer vision
CN103020992A (en) Video image significance detection method based on dynamic color association
CN110298914A (en) A kind of method of fruit tree canopy characteristic map in orchard establishing
CN108197581B (en) Unmanned aerial vehicle signal identification detection method based on improved AC-WGANs
Zhang et al. A dynamic hand gesture recognition algorithm based on CSI and YOLOv3
CN111612130B (en) Frequency shift keying communication signal modulation mode identification method
CN110062378A (en) Identity identifying method based on channel state information under a kind of gesture scene
CN106230544A (en) The monitoring identification of a kind of automobile remote-control interference signal and localization method
CN110378892A (en) A kind of method of quick detection electric transmission line channel hidden danger
CN113252584B (en) Crop growth detection method and system based on 5G transmission
CN108508425B (en) Method for detecting foreground target based on neighborhood characteristics under radar near-earth background noise
CN106683105A (en) Image segmentation method and image segmentation device
CN109063940B (en) Intelligent vehicle threat estimation system and method based on variable structure Bayesian network
Liang et al. Fall detection system based on millimeter wave radar and machine learning
Ma et al. Deep learning based cognitive radio modulation parameter estimation
CN109995690A (en) The neural network self-organization method of MFSK digital signal subclass Modulation Identification
CN112596024B (en) Motion identification method based on environment background wireless radio frequency signal
Cao et al. Feature extraction of remote sensing images based on bat algorithm and normalized chromatic aberration
Zhao et al. Weather classification based on convolutional neural networks

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