CN115664906A - Unsupervised clustering method and unsupervised clustering device for TDMA signal protocol - Google Patents
Unsupervised clustering method and unsupervised clustering device for TDMA signal protocol Download PDFInfo
- Publication number
- CN115664906A CN115664906A CN202211276019.XA CN202211276019A CN115664906A CN 115664906 A CN115664906 A CN 115664906A CN 202211276019 A CN202211276019 A CN 202211276019A CN 115664906 A CN115664906 A CN 115664906A
- Authority
- CN
- China
- Prior art keywords
- signal
- training
- clustering
- tdma
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- 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
Landscapes
- Time-Division Multiplex Systems (AREA)
Abstract
The invention discloses a method and a device for unsupervised clustering of a TDMA signal protocol, wherein the method comprises the following steps: acquiring a receiving instruction; receiving a target frequency band signal by using the receiving instruction to obtain a TDMA signal data set; processing the TDMA signal data set to obtain IQ data; preprocessing the IQ data to obtain a TDMA signal clustering algorithm training set; utilizing the TDMA signal clustering algorithm training set to perform signal protocol unsupervised algorithm training to obtain a target signal clustering algorithm model; and clustering the received TDMA signals by using the target signal clustering algorithm model to obtain the protocol type of the TDMA signals, thereby realizing the unsupervised clustering of the TDMA signal protocol based on deep learning. The method of the invention realizes unsupervised clustering of the TDMA signal protocol by receiving, truncating and time-frequency transforming the signal and adopting a clustering technology based on deep learning.
Description
Technical Field
The invention relates to the field of unsupervised clustering of signal protocols, in particular to a method and a device for unsupervised clustering of Time Division Multiple Access (TDMA) signal protocols.
Background
The traditional protocol analysis needs to carry out parameter estimation layer by layer on a communication protocol, and can carry out bit stream analysis on protocol words on the premise of solving the problems of modulation pattern identification and demodulation, interleaving and scrambling code parameter estimation and de-interleaving and descrambling, and channel decoding parameter estimation and decoding. This analysis method has the following disadvantages: the analysis period is long, the expert dependence is strong, the algorithm complexity is high, and the requirement for real-time analysis of unknown and agile electromagnetic signals is difficult to solve.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a method and a device for unsupervised clustering of a TDMA signal protocol, which can acquire a receiving instruction; receiving a target frequency band signal by using the receiving instruction to obtain a TDMA signal data set; processing the TDMA signal data set to obtain IQ data; preprocessing the IQ data to obtain a TDMA signal clustering algorithm training set; utilizing the TDMA signal clustering algorithm training set to perform signal protocol unsupervised algorithm training to obtain a target signal clustering algorithm model; and clustering the received TDMA signals by using the target signal clustering algorithm model to obtain the protocol type of the TDMA signals, thereby realizing the unsupervised clustering of the TDMA signal protocol based on deep learning. The method for rapidly clustering the signal protocols through the deep learning method can be applied to electronic information systems, such as communication signal processing and the like, and lays a technical foundation for intelligent protocol analysis.
In order to solve the above technical problem, a first aspect of the embodiments of the present invention discloses a method for unsupervised clustering of TDMA signal protocols, where the method includes:
s1, acquiring a receiving instruction; the receiving instruction comprises a working frequency band and a line gain parameter;
s2, receiving a target frequency band signal by using the receiving instruction to obtain a TDMA signal data set; the TDMA signal data set comprises service burst TB data, reference burst AB data and synchronous burst RB data;
s3, processing the TDMA signal data set to obtain IQ data; the IQ data is in-phase orthogonal signal data, I is in-phase, and Q is orthogonal with the phase difference of 90 degrees between I and I;
s4, preprocessing the IQ data to obtain a TDMA signal clustering algorithm training set;
s5, training a TDMA signal clustering algorithm model by utilizing the TDMA signal clustering algorithm training set to obtain a target signal clustering algorithm model;
and S6, clustering the received TDMA signals by using the target signal clustering algorithm model to obtain the protocol type of the TDMA signals.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the preprocessing is performed on the IQ data to obtain a TDMA signal clustering algorithm training set, and the method includes:
s41, measuring the IQ data to obtain the length value of each TDMA signal data in the TDMA signal data set, and taking the maximum value as L max ；
S42, for any TDMA signal data, intercepting the TDMA signal data with the length of
The data header of (a) is the length of the TDMA signal data;
s43, judging whether l is equal to 128 or not to obtain a judgment result;
if the judgment result is negative, the data is supplemented from 0 to 128 to obtain second communication signal data with the shape of [2,128 ];
when the judgment result is yes, directly obtaining second communication signal data with the shape of [2,128 ];
s44, performing WVD time-frequency transformation on the second communication signal data to obtain a time-frequency spectrogram with the shape of [128,128 ]:
in the formula (I), the compound is shown in the specification,
is the instantaneous autocorrelation function R (t, tau) of the signal x (t), t is time shift, tau is an integral variable, f is frequency, S (t, f) is a time-frequency transformation function;
the WVD time-frequency transformation, namely Wigner-Ville distribution, is the Fourier transformation of a signal instantaneous correlation function and reflects the signal instantaneous time-frequency relation;
s45, constructing and obtaining training data with the shape of [128,128] by using the time-frequency spectrogram;
s46, marking the time-frequency spectrogram signal protocol type by using the signal protocol type of the time-frequency spectrogram to obtain a training data label with the shape of [128,128 ];
the signal protocol types include: a traffic burst TB, a reference burst AB, a synchronization burst RB;
s47, integrating the training data and the training data labels to obtain the signal clustering algorithm training data;
and S48, fusing all the signal clustering algorithm training data to obtain a TDMA signal clustering algorithm training set.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the TDMA signal clustering algorithm model is composed of a pre-training module and a clustering training module; training a TDMA signal clustering algorithm model by utilizing the TDMA signal clustering algorithm training set to obtain a target signal clustering algorithm model, wherein the method comprises the following steps of;
s51, pre-training the pre-training module by utilizing the TDMA signal training data set to obtain adjacent sample data and a characteristic extraction encoder of the TDMA signal training data set;
the characteristic extraction encoder is used for extracting the characteristics of the frequency domain information of the TDMA communication signal to obtain a characteristic expression vector;
s52, training the feature extraction encoder and the clustering training module by using the TDMA signal training data set and the adjacent sample data to obtain a target signal clustering algorithm model;
and the clustering training module trains the feature expression vectors to form a feature library.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the pre-training module is pre-trained by using the TDMA signal training data set to obtain neighboring sample data and feature extraction encoder of the TDMA signal training data set, where the method includes:
s511, inputting the TDMA signal training data set after the WVD preprocessing into a standard contrast learning BYOL network;
the contrast learning BYOL network consists of a target network with a parameter theta and an on-line network with a parameter epsilon; the online network is formed by an encoder f θ () Predictor g θ () And mapper q θ () Comprising a target network including an encoder f θ () Sum predictor g θ ()；
The predictor is used for the structural asymmetry of the target network and the online network;
s512, training the TDMA signal training data set by using the contrast learning BYOL network to obtain a feature expression vector;
the TDMA signal training data sets are respectively input into an online network and a target network of the contrast learning BYOL network to obtain q θ (Z θ ) And Z ε Using q θ (Z θ ) And Z ε Design loss function L B ：
In the formula, q θ (Z θ ) For the output of the online network, q θ () Predictor, Z θ Output vector, Z, for online network mapper ε Outputting for the target network;
by means of L B The gradient descent updates the on-line network parameter theta, and the updating method of the target network parameter epsilon comprises the following steps: :
ε←τε+(1-τ)θ
wherein epsilon is an online network parameter, theta is a target network parameter, and tau =0.5 is a hyperparameter;
s513, measuring adjacent sample data of the feature expression vector by using the similarity function to obtain position information of the adjacent sample in the TDMA signal training data set;
the cosine similarity function is expressed as:
wherein a and b are respectively a feature vector, and C (a and b) is the similarity between a and b;
s514, inputting the position information into an encoder for encoding and storing to obtain the adjacent sample data information of the TDMA signal training data set, and inputting the encoder f of the online network θ () And storing to obtain the characteristic extraction encoder.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the training the feature extraction encoder and the cluster training module by using the TDMA signal training data set and the adjacent sample data to obtain a target signal cluster algorithm model, includes:
s521, connecting the feature output extracted by the feature extraction encoder with the clustering module to obtain a clustering network phi (); the clustering module consists of full connection layers;
the cluster training module is used for carrying out classification processing on the extracted characteristic parameters by using a classification model to obtain a classification result;
s522, inputting the TDMA signal training data set and the adjacent sample data information into the clustering network phi ();
s523, for the clustering loss function L c And probability entropy L e Calculating to obtain a clustering loss function L c Value and probability entropy L e A value;
the cluster loss function L c The expression is as follows:
L c ＝λL n +L a
wherein:
in the formula, L n For an adjacency loss, L a To distribute losses, x i Is an element, n, in the second communication signal training data set X xi Is the calculated nearest neighbor information N of each training data set X of the second communication signal x Φ () is the clustering network, M is the number of all elements in the second communication signal training data set X, k =3, λ and γ are hyper-parameters, λ is used to balance L n And L a ，q i Training the second communication signal training data set X through the clustering network phi () to obtain a distribution probability matrix P N Elements of each column vector after being transformed into the column vector;
is to train the nearest neighbor information N through the clustering network phi () x Obtaining an assignment probability matrixP N Elements of each column vector after being transformed into the column vector;
the probability entropy L e The expression is as follows:
wherein, H is an entropy function, M is the number of all elements in the second communication signal training data set X, and k =3,q i Training the second communication signal training data set X through the clustering network phi () to obtain an assignment probability matrix P N Elements of each column vector after conversion into column vectors, j being q i Subscripts of elements in the vector;
s524, utilizing the clustering loss function L c And probability entropy L e Performing iterative training on the clustering network phi () to obtain training accuracy and standard mutual information of each time;
and S525, optimizing the loss function by using a batch gradient descent optimization algorithm based on the training accuracy and the standard mutual information, finishing the training of the TDMA signal clustering algorithm model, and obtaining the target signal clustering algorithm model.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the processing the TDMA signal data set to obtain IQ data includes:
carrying out low-noise power amplification, filtering, A/D conversion and digital down-conversion on the TDMA signal data set to obtain IQ data;
s31, amplifying the TDMA signal data set by using a low-noise power amplifier to obtain a first TDMA signal data set;
s32, processing the low-noise TDMA signal data set by using a band-pass filter to obtain a second TDMA signal data set;
s33, performing A/D conversion on the second TDMA signal data set by using an A/D converter to obtain a third TDMA signal data set;
and S34, carrying out digital down-conversion on the third TDMA signal data set by using a digital down-converter to obtain the IQ data.
The second aspect of the embodiment of the invention discloses a TDMA signal protocol unsupervised clustering device, which comprises:
the instruction receiving module is used for acquiring a receiving instruction; the receiving instruction comprises a working frequency band and a line gain parameter;
the first processing module is used for receiving the target frequency band signal by using the receiving instruction and acquiring a TDMA signal data set; the TDMA signal data set comprises service burst TB data, reference burst AB data and synchronous burst RB data;
the second processing module is used for processing the TDMA signal data set to obtain IQ data; the IQ data is in-phase orthogonal signal data, I is in-phase, and Q is orthogonal with the phase difference of 90 degrees between I and I;
the IQ data preprocessing module is used for preprocessing the IQ data to obtain a TDMA signal clustering algorithm training set;
the unsupervised clustering algorithm training module is used for training a TDMA signal clustering algorithm model by utilizing the TDMA signal clustering algorithm training set to obtain a target signal clustering algorithm model;
and the unsupervised clustering module is used for clustering the received TDMA signals by using the target signal clustering algorithm model to obtain the protocol type of the TDMA signals.
As an optional implementation manner, in the second aspect of the embodiment of the present invention, the IQ data preprocessing module preprocesses the IQ data to obtain a TDMA signal clustering algorithm training set:
measuring the IQ data to obtain the length value of each TDMA signal data in the TDMA signal data set, and taking the maximum value as L max ；
For any one of the TDMA signal data, truncating the length of the TDMA signal data from the starting position of the TDMATDMA signal data
The data header of (a) is the length of the TDMA signal data;
judging whether l is equal to 128 to obtain a judgment result;
if the judgment result is negative, the data is supplemented from 0 to 128 to obtain second communication signal data with the shape of [2,128 ];
when the judgment result is yes, directly obtaining second communication signal data with the shape of [2,128 ];
performing WVD time-frequency transformation on the second communication signal data to obtain a time-frequency spectrogram with the shape of [128,128 ]:
in the formula (I), the compound is shown in the specification,
is the instantaneous autocorrelation function R (t, tau) of the signal x (t), t is time shift, tau is an integral variable, f is frequency, S (t, f) is a time-frequency transformation function;
constructing and obtaining training data with the shape of [128,128] by using the time-frequency spectrogram;
marking the time-frequency spectrogram signal protocol type by using the signal protocol type of the time-frequency spectrogram to obtain a training data label with the shape of [128,128 ];
the signal protocol types include: a service burst TB, a reference burst AB, a synchronization burst RB;
integrating the training data and the training data labels to obtain the signal clustering algorithm training data;
and fusing all the signal clustering algorithm training data to obtain a TDMA signal clustering algorithm training set.
As an optional implementation manner, in the second aspect of the embodiment of the present invention, the unsupervised clustering algorithm training module trains a TDMA signal clustering algorithm model by using the TDMA signal clustering algorithm training set to obtain a target signal clustering algorithm model:
the TDMA signal clustering algorithm model consists of a pre-training module and a clustering training module;
pre-training the pre-training module by utilizing the TDMA signal training data set to obtain adjacent sample data and a characteristic extraction encoder of the TDMA signal training data set;
the characteristic extraction encoder is used for extracting the characteristics of the frequency domain information of the TDMA communication signal to obtain a characteristic expression vector;
training the feature extraction encoder and the clustering training module by using the TDMA signal training data set and the adjacent sample data to obtain a target signal clustering algorithm model;
and the clustering training module trains the feature expression vectors to form a feature library.
As an optional implementation manner, in the second aspect of the embodiment of the present invention, the unsupervised clustering algorithm training module performs pre-training on the pre-training module by using the TDMA signal training data set to obtain adjacent sample data and feature extraction encoder of the TDMA signal training data set, including:
inputting the TDMA signal training data set subjected to WVD preprocessing into a standard contrast learning BYOL network;
the contrast learning BYOL network consists of a target network with a parameter theta and an on-line network with a parameter epsilon; the online network is formed by an encoder f θ () Predictor g θ () And mapper q θ () Comprising a target network including an encoder f θ () Sum predictor g θ ()；
The predictor is used for the structural asymmetry of the target network and the online network;
training the TDMA signal training data set by using the contrast learning BYOL network to obtain a feature expression vector;
the TDMA signal training data sets are respectively input into an online network and a target network of the contrast learning BYOL network to obtain q θ (Z θ ) And Z ε By using q θ (Z θ ) And Z ε Design loss function L B ：
In the formula, q θ (Z θ ) For the output of the on-line network, q θ () Predictor, Z θ Output vector, Z, for online network mapper ε Outputting for the target network;
by means of L B The gradient descent updates the on-line network parameter theta, and the updating method of the target network parameter epsilon comprises the following steps:
ε←τε+(1-τ)θ
wherein epsilon is an online network parameter, theta is a target network parameter, and tau =0.5 is a hyperparameter;
measuring adjacent sample data of the feature expression vector by using the similarity function to obtain position information of the adjacent sample in the TDMA signal training data set;
the cosine similarity function is expressed as:
wherein a and b are feature vectors, respectively, and C (a and b) is the similarity between a and b;
inputting the position information into an encoder for encoding and storing to obtain the adjacent sample data information of the TDMA signal training data set, and inputting the encoder f of the online network θ () And storing to obtain the characteristic extraction encoder.
As an optional implementation manner, in a second aspect of the embodiment of the present invention, the unsupervised clustering algorithm training module trains the feature extraction encoder and the clustering training module by using the TDMA signal training data set and the adjacent sample data to obtain a target signal clustering algorithm model, where the method includes:
connecting the feature output extracted by the feature extraction encoder with the clustering module to obtain a clustering network phi (); the clustering module consists of full connection layers;
the cluster training module is used for carrying out classification processing on the extracted characteristic parameters by using a classification model to obtain a classification result;
inputting the TDMA signal training data set and the adjacent sample data information into the clustering network phi ();
for clustering loss function L c And probability entropy L e Calculating to obtain a clustering loss function L c Value and probability entropy L e A value;
the cluster loss function L c The expression is as follows:
L c ＝λL n +L a
wherein:
in the formula, L n For an adjacency loss, L a To distribute losses, x i Is an element, n, in the second communication signal training data set X xi Is the calculated nearest neighbor information N of each training data set X of the second communication signal x Φ () is the clustering network, M is the number of all elements in the second communication signal training data set X, k =3, λ and γ are hyper-parameters, λ is used to balance L n And L a ，q i Training the second communication signal training data set X through the clustering network phi (), and obtaining a distribution probability matrix P N Elements of each column vector after being transformed into the column vector;
training the nearest neighbor information N through the clustering network phi () x Obtaining an assignment probability matrix P N Elements of each column vector after being transformed into the column vector;
the describedProbability entropy L e The expression is as follows:
wherein, H is entropy function, M is the number of all elements in the training data set X of the second communication signal, k =3,q i Training the second communication signal training data set X through the clustering network phi () to obtain an assignment probability matrix P N Elements of each column vector after conversion into column vectors, j being q i Subscripts of elements in the vector;
using the clustering loss function L c And probability entropy L e Performing iterative training on the clustering network phi () to obtain training accuracy and standard mutual information of each time;
and optimizing a loss function by using a batch gradient descent optimization algorithm based on the training accuracy and the standard mutual information, finishing the training of the TDMA signal clustering algorithm model, and obtaining a target signal clustering algorithm model.
As an optional implementation manner, in the second aspect of the embodiment of the present invention, the second processing module processes the TDMA communication signal data set to obtain IQ data, and the method includes:
carrying out low-noise power amplification, filtering, A/D conversion and digital down-conversion on the TDMA communication signal data set to obtain IQ data;
carrying out low-noise power amplification on the TDMA signal data set by using a low-noise power amplifier to obtain a first TDMA signal data set;
processing the low-noise TDMA signal data set by using a band-pass filter to obtain a second TDMA signal data set;
performing A/D conversion on the second TDMA signal data set by using an A/D converter to obtain a third TDMA signal data set;
and carrying out digital down-conversion on the third TDMA signal data set by using a digital down-converter to obtain the IQ data.
The third aspect of the present invention discloses another TDMA signal protocol unsupervised clustering apparatus, which includes:
a memory storing executable program code;
a processor coupled with the memory;
the processor calls the executable program codes stored in the memory to execute part or all of the steps of the TDMA signal protocol unsupervised clustering method disclosed by the first aspect of the embodiment of the invention.
In a fourth aspect, the present invention discloses a computer storage medium, which stores computer instructions for performing some or all of the steps of the TDMA signal protocol unsupervised clustering method disclosed in the first aspect of the embodiments of the present invention when the computer instructions are called.
Compared with the prior art, the embodiment of the invention has the following beneficial effects:
in the embodiment of the invention, a receiving instruction can be obtained; receiving a target frequency band signal by using the receiving instruction to obtain a TDMA signal data set; processing the TDMA signal data set to obtain IQ data; preprocessing the IQ data to obtain a TDMA signal clustering algorithm training set; utilizing the TDMA signal clustering algorithm training set to perform signal protocol unsupervised algorithm training to obtain a target signal clustering algorithm model; and clustering the received TDMA signals by using the target signal clustering algorithm model to obtain the protocol type of the TDMA signals. The method for deep learning does not depend on experts to realize rapid clustering of the TDMA signal protocol, has short analysis period and simple and efficient algorithm, and meets the requirement of real-time analysis on unknown and agile electromagnetic signals.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a schematic flowchart of an unsupervised clustering method for TDMA signal protocol according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of an unsupervised clustering apparatus for TDMA signal protocol according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a TDMA signal clustering algorithm model adopted by the device for unsupervised clustering of TDMA signal protocols according to the embodiment of the present invention;
FIG. 4 is a schematic diagram of a band-pass filter employed by another device for unsupervised clustering of TDMA signal protocols according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of digital down-conversion employed by another device for TDMA signal protocol unsupervised clustering according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of another TDMA signal protocol unsupervised clustering apparatus according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The terms "first," "second," and the like in the description and claims of the present invention and in the above-described drawings are used for distinguishing between different objects and not for describing a particular order. Furthermore, the terms "include" and "have," as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, apparatus, product, or apparatus that comprises a list of steps or elements is not limited to those listed but may alternatively include other steps or elements not listed or inherent to such process, method, product, or apparatus.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the invention. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
The invention discloses a method and a device for unsupervised clustering of a Time Division Multiple Access (TDMA) signal protocol, which can acquire a receiving instruction; receiving a target frequency band signal by using the receiving instruction to obtain a TDMA signal data set; processing the TDMA signal data set to obtain IQ data; preprocessing the IQ data to obtain a TDMA signal clustering algorithm training set; performing TDMA signal protocol unsupervised algorithm training by utilizing the TDMA signal clustering algorithm training set to obtain a target signal clustering algorithm model; and clustering the received TDMA signals by using the target signal clustering algorithm model to obtain the protocol type of the TDMA signals. The TDMA signal protocols are quickly clustered on a physical layer through a deep learning method, and a technical foundation is laid for intelligent protocol analysis. The following are detailed below.
Example one
Referring to fig. 1, fig. 1 is a flowchart illustrating an unsupervised clustering method for TDMA signal protocols according to an embodiment of the present invention. The method for TDMA signal protocol unsupervised clustering described in fig. 1 is applied to an electronic information system, such as a local server or a cloud server for TDMA signal protocol unsupervised clustering management, and the embodiment of the present invention is not limited thereto. As shown in fig. 1, the method for unsupervised clustering of TDMA signal protocols may include the following operations:
s1, acquiring a receiving instruction; the receiving instruction comprises a working frequency band and a line gain parameter;
s2, receiving a target frequency band signal by using the receiving instruction to obtain a TDMA signal data set; the TDMA signal data set comprises service burst TB data, reference burst AB data and synchronous burst RB data;
s3, processing the TDMA signal data set to obtain IQ data; the IQ data is in-phase orthogonal signal data, I is in-phase, and Q is orthogonal with the phase difference of 90 degrees of I;
carrying out low-noise power amplification, filtering, A/D conversion and digital down-conversion on the TDMA communication signal data set to obtain IQ data;
s31, amplifying the TDMA signal data set by using a low-noise power amplifier to obtain a first TDMA signal data set;
s32, processing the low-noise TDMA signal data set by using a band-pass filter to obtain a second TDMA signal data set;
optionally, a band-pass filter is used to filter the received signal, fig. 4 is a basic schematic diagram of the band-pass filter, vi is an input voltage, vo is an output voltage, L is an inductor, and C is a capacitor;
optionally, filtering the received signal by using an extended kalman filtering algorithm with multidimensional variables, wherein the expression is shown in the specification;
wherein x is a variable, x k In the form of a state vector, the state vector,
as a Jacobian matrix, o n Representing the infinitesimal high order, n being the order, f (x) being a one-dimensional Taylor expansion function;
in the formula (I), the compound is shown in the specification,
as a Jacobian matrix, f (x) k ) Is an m-dimensional Taylor expansion function, f (x) is a function with derivatives of order n, x is a variable, x k Is a state vector, m isThe dimensions of the vector are such that,
s33, performing A/D conversion on the second TDMA signal data set by using an A/D converter to obtain a third TDMA signal data set;
s34, performing digital down-conversion on the third TDMA signal data set by using a digital down-converter to obtain the IQ data;
optionally, digital down-conversion is used to down-convert the received signal, as shown in fig. 5.
S4, preprocessing the IQ data to obtain a TDMA signal clustering algorithm training set, including;
s41, the IQ data are measured to obtain the length value of each TDMA signal data in the TDMA signal data set, and the maximum value is taken as L max ；
S42, for any TDMA signal data, intercepting the TDMA signal data with the length of
The data header of (a), l is the length of the TDMA signal data;
s43, judging whether l is equal to 128 or not to obtain a judgment result;
if the judgment result is negative, the data is supplemented from 0 to 128 to obtain second communication signal data with the shape of [2,128 ];
when the judgment result is yes, directly obtaining second communication signal data with the shape of [2,128 ];
s44, performing WVD time-frequency transformation on the second communication signal data to obtain a time-frequency spectrogram with the shape of [128,128 ]:
in the formula (I), the compound is shown in the specification,
is the instantaneous autocorrelation function R (t, tau) of the signal x (t), t is time shift, tau is an integral variable, f is frequency, S (t, f) is a time-frequency transformation function;
s45, constructing and obtaining training data with the shape of [128,128] by using the time-frequency spectrogram;
s46, marking the time-frequency spectrogram signal protocol type by using the signal protocol type of the time-frequency spectrogram to obtain a training data label with the shape of [128,128 ];
the signal protocol types include: a service burst TB, a reference burst AB, a synchronization burst RB;
s47, integrating the training data and the training data labels to obtain the signal clustering algorithm training data;
and S48, fusing all the signal clustering algorithm training data to obtain a TDMA signal clustering algorithm training set.
S5, training a TDMA signal clustering algorithm model by utilizing the TDMA signal clustering algorithm training set to obtain a target signal clustering algorithm model, wherein the method comprises the following steps:
the TDMA signal clustering algorithm model is shown in fig. 3, wherein Aug1 and Aug2 are two different enhancement modes, respectively, and Anchor and Neighbor are Anchor samples and their adjacent samples, respectively;
s51, pre-training the pre-training module by utilizing the TDMA signal training data set to obtain adjacent sample data and a characteristic extraction encoder of the TDMA signal training data set;
s511, inputting the TDMA signal training data set after the WVD preprocessing into a standard contrast learning BYOL network;
the contrast learning BYOL network consists of a target network with a parameter theta and an on-line network with a parameter epsilon; the online network is formed by an encoder f θ () Predictor g θ () And mapper q θ () Comprising a target network including an encoder f θ () Sum predictor g θ ()；
The predictor is used for the structural asymmetry of the target network and the online network;
s512, training the TDMA signal training data set by using the contrast learning BYOL network to obtain a feature expression vector;
the TDMA signal training data sets are respectively input into an online network and a target network of the contrast learning BYOL network to obtain q θ (Z θ ) And Z ε Using q θ (Z θ ) And Z ε Design loss function L B ：
In the formula, q θ (Z θ ) For the output of the on-line network, q θ () Predictor, Z θ Output vector, Z, for online network mapper ε Outputting for the target network;
by means of L B The gradient descent updates the on-line network parameter theta, and the updating method of the target network parameter epsilon comprises the following steps:
ε←τε+(1-τ)θ
in the formula, epsilon is an online network parameter, theta is a target network parameter, and tau =0.5 is a hyperparameter;
s513, measuring adjacent sample data of the feature expression vector by using the similarity function to obtain position information of the adjacent sample in the TDMA signal training data set;
the cosine similarity function is expressed as:
wherein a and b are feature vectors, respectively, and C (a and b) is the similarity between a and b;
s514, inputting the position information into an encoder for encoding and storing to obtain the adjacent sample data information of the TDMA signal training data set, and inputting the encoder f of the online network θ () Storing to obtainTo the feature extraction encoder.
S52, training the feature extraction encoder and the clustering training module by using the TDMA signal training data set and the adjacent sample data to obtain a target signal clustering algorithm model, which comprises the following steps:
s521, connecting the feature output extracted by the feature extraction encoder with the clustering module to obtain a clustering network phi (); the clustering module consists of full connection layers;
the cluster training module is used for carrying out classification processing on the extracted characteristic parameters by using a classification model to obtain a classification result;
s522, inputting the TDMA signal training data set and the adjacent sample data information into the clustering network phi ();
s523, for the clustering loss function L c And probability entropy L e Calculating to obtain a clustering loss function L c Value and probability entropy L e A value;
the cluster loss function L c The expression is as follows:
L c ＝λL n +L a
wherein:
in the formula, L n For an adjacency loss, L a To distribute losses, x i Is an element, n, in the second communication signal training data set X xi Is the calculated nearest neighbor information N of each training data set X of the second communication signal x Φ () is the clustering network, M is the number of all elements in the second communication signal training data set X, k =3, λ and γ are hyper-parameters, λ is used to balance L n And L a ，q i Is through the clustering networkPhi () trains the second communication signal training data set X to obtain an assignment probability matrix P N Elements of each column vector after being transformed into the column vector;
is to train the nearest neighbor information N through the clustering network phi () x Obtaining an assignment probability matrix P N Elements of each column vector after being transformed into the column vector;
the probability entropy L e The expression is as follows:
wherein, H is an entropy function, M is the number of all elements in the second communication signal training data set X, and k =3,q i Training the second communication signal training data set X through the clustering network phi () to obtain an assignment probability matrix P N Elements of each column vector after conversion into column vectors, j being q i Subscripts of elements in the vector;
s524, utilizing the clustering loss function L c And probability entropy L e Performing iterative training on the clustering network phi () to obtain training accuracy and standard mutual information of each time;
and S525, optimizing the loss function by using a batch gradient descent optimization algorithm based on the training accuracy and the standard mutual information, finishing the training of the TDMA signal clustering algorithm model, and obtaining the target signal clustering algorithm model.
S6, clustering the received TDMA signals by using the target signal clustering algorithm model to obtain the protocol type of the TDMA signals;
optionally, a batch gradient descent algorithm is used for iterative optimization, and a parameter updating method is as follows:
wherein Θ isAs a parameter of the clustering network, a =0.001 is a learning rate,
in order to reduce the direction of the fastest descending of the loss function, the theta parameter in the loss function is obtained by partial derivation.
Therefore, by implementing the TDMA signal protocol unsupervised clustering method, a receiving instruction can be obtained; receiving a target frequency band signal by using the receiving instruction to obtain a TDMA signal data set; processing the TDMA signal data set to obtain IQ data; preprocessing the IQ data to obtain a TDMA signal clustering algorithm training set; utilizing the TDMA signal clustering algorithm training set to perform signal protocol unsupervised algorithm training to obtain a target signal clustering algorithm model; and clustering the received TDMA signals by using the target signal clustering algorithm model to obtain the protocol type of the TDMA signals. The signal protocol is subjected to rapid clustering by a deep learning method without depending on experts, and a technical basis is laid for intelligent protocol analysis.
Example two
Referring to fig. 2, fig. 2 is a schematic structural diagram of an unsupervised clustering apparatus for TDMA signal protocol according to an embodiment of the present invention. The apparatus described in fig. 2 can be applied to a TDMA signal processing system, such as a local server or a cloud server for TDMA signal processing, and the embodiments of the present invention are not limited thereto. As shown in fig. 2, the apparatus may include:
an instruction receiving module 201, configured to obtain a receiving instruction; the receiving instruction comprises a working frequency band and a line gain parameter;
a first processing module 202, configured to receive a target frequency band signal by using the receiving instruction, and obtain a TDMA signal data set; the TDMA signal data set comprises service burst TB data, reference burst AB data and synchronous burst RB data;
a second processing module 203, configured to process the TDMA signal data set to obtain IQ data; the IQ data is in-phase orthogonal signal data, I is in-phase, and Q is orthogonal with the phase difference of 90 degrees between I and I;
an IQ data preprocessing module 204, configured to preprocess the IQ data to obtain a TDMA signal clustering algorithm training set;
the unsupervised clustering algorithm training module 205 is configured to train a TDMA signal clustering algorithm model by using the TDMA signal clustering algorithm training set to obtain a target signal clustering algorithm model;
and the unsupervised clustering module 206 is configured to perform clustering processing on the received TDMA signals by using the target signal clustering algorithm model to obtain a protocol type of the TDMA signals.
It can be seen that, implementing the apparatus for unsupervised clustering of TDMA signal protocol described in fig. 2, can obtain a TDMA signal data set by obtaining a receiving instruction, receiving a target frequency band signal; processing the TDMA signal data set to obtain IQ data; preprocessing the IQ data to obtain a TDMA signal clustering algorithm training set; utilizing the TDMA signal clustering algorithm training set to perform signal protocol unsupervised algorithm training to obtain a target signal clustering algorithm model; and clustering the received TDMA signals by using the target signal clustering algorithm model to obtain the protocol type of the TDMA signals. The TDMA signal protocol is subjected to rapid clustering independent of experts by a deep learning method, and a technical foundation is laid for intelligent protocol analysis.
In another alternative embodiment, as shown in fig. 2, the IQ data preprocessing module 204 preprocesses the IQ data to obtain a training set of the TDMA signal protocol unsupervised clustering algorithm, which includes:
measuring the IQ data to obtain the length value of each TDMA signal data in the TDMA signal data set, and taking the maximum value as L max ；
For any one of the TDMA signal data, truncating the length of the TDMA signal data from the starting position of the TDMATDMA signal data
The data header of (a), l is the length of the TDMA signal data;
judging whether l is equal to 128 to obtain a judgment result;
if the judgment result is negative, the data is supplemented from 0 to 128, and second communication signal data with the shape of [2,128] is obtained;
when the judgment result is yes, directly obtaining second communication signal data with the shape of [2,128 ];
performing WVD time-frequency transformation on the second communication signal data to obtain a time-frequency spectrogram with the shape of [128,128 ]:
in the formula (I), the compound is shown in the specification,
is the instantaneous autocorrelation function R (t, tau) of the signal x (t), t is time shift, tau is an integral variable, f is frequency, S (t, f) is a time-frequency transformation function;
constructing and obtaining training data with the shape of [128,128] by using the time-frequency spectrogram;
marking the time-frequency spectrogram signal protocol type by using the signal protocol type of the time-frequency spectrogram to obtain a training data label with the shape of [128,128 ];
the signal protocol types include: a service burst TB, a reference burst AB, a synchronization burst RB;
integrating the training data and the training data labels to obtain the signal clustering algorithm training data;
and fusing all the signal clustering algorithm training data to obtain a TDMA signal clustering algorithm training set.
Therefore, by implementing the reinforcement learning-based TDMA signal protocol unsupervised clustering device for the TDMA signal protocol unsupervised clustering described in the figure 2, IQ data can be preprocessed to obtain a TDMA signal protocol unsupervised clustering algorithm training set, and the TDMA signal protocol unsupervised algorithm training is performed, so that the TDMA signal protocol is favorably subjected to expert-independent fast clustering, and a technical basis is laid for intelligent protocol analysis.
In yet another alternative embodiment, as shown in fig. 2, the unsupervised clustering algorithm training module 205 trains a TDMA signal clustering algorithm model by using the TDMA signal clustering algorithm training set to obtain a target signal clustering algorithm model, and the method includes;
s51, pre-training the pre-training module by utilizing the TDMA signal training data set to obtain adjacent sample data and a characteristic extraction encoder of the TDMA signal training data set;
the characteristic extraction encoder is used for extracting the characteristics of the frequency domain information of the TDMA communication signal to obtain a characteristic expression vector;
s52, training the feature extraction encoder and the clustering training module by using the TDMA signal training data set and the adjacent sample data to obtain a target signal clustering algorithm model;
and the clustering training module trains the feature expression vectors to form a feature library.
Therefore, by implementing the TDMA signal protocol unsupervised clustering device for unsupervised clustering of the TDMA signal protocol described in fig. 2, the TDMA signal clustering algorithm model can be used for processing the received signals, thereby facilitating the rapid clustering of the TDMA signal protocol independent of experts and laying a technical foundation for intelligent protocol analysis.
In yet another alternative embodiment, as shown in fig. 2, the unsupervised clustering algorithm training module 205 performs pre-training on the pre-training module by using the TDMA signal training data set to obtain neighboring sample data and feature extraction encoder of the TDMA signal training data set, including;
inputting the TDMA signal training data set subjected to WVD preprocessing into a standard contrast learning BYOL network;
the contrast learning BYOL network consists of a target network with a parameter theta and an on-line network with a parameter epsilon; the online network is formed by an encoder f θ () Predictor g θ () And mapper q θ () Comprising a target network including an encoder f θ () Sum predictor g θ ()；
The predictor is used for the structural asymmetry of the target network and the online network;
training the TDMA signal training data set by using the contrast learning BYOL network to obtain a feature expression vector;
the TDMA signal training data sets are respectively input into an online network and a target network of the contrast learning BYOL network to obtain q θ (Z θ ) And Z ε Using q θ (Z θ ) And Z ε Design loss function L B ：
In the formula, q θ (Z θ ) For the output of the on-line network, q θ () Predictor, Z θ Output vector, Z, for online network mapper ε Outputting for the target network;
by means of L B Gradient descent updates the online network parameter theta, and the updating method of the target network parameter epsilon comprises the following steps:
ε←τε+(1-τ)θ
in the formula, epsilon is an online network parameter, theta is a target network parameter, and tau =0.5 is a hyperparameter;
measuring adjacent sample data of the feature expression vector by using the similarity function to obtain position information of the adjacent sample in the TDMA signal training data set;
the cosine similarity function is expressed as:
wherein a and b are feature vectors, respectively, and C (a and b) is the similarity between a and b;
inputting the position information into an encoder for encoding and storing to obtain the adjacent sample data information of the TDMA signal training data set, and encoding the encoder f of the online network θ () Saving, obtaining said feature extraction encoder。
It can be seen that, by implementing the TDMA signal protocol unsupervised clustering device for the TDMA signal protocol unsupervised clustering described in fig. 2, the TDMA signal training data set can be pre-trained to obtain the adjacent sample data and the feature extraction encoder of the TDMA signal training data set, which is beneficial to realizing the rapid clustering of the TDMA signal protocol independent of experts and lays a technical foundation for intelligent protocol analysis.
In yet another alternative embodiment, as shown in fig. 2, the unsupervised clustering algorithm training module 205 trains the feature extraction encoder and the clustering training module by using the TDMA signal training data set and the adjacent sample data to obtain a target signal clustering algorithm model, which includes:
connecting the feature output extracted by the feature extraction encoder with the clustering module to obtain a clustering network phi (); the clustering module consists of full connection layers;
the cluster training module is used for carrying out classification processing on the extracted characteristic parameters by using a classification model to obtain a classification result;
inputting the TDMA signal training data set and the adjacent sample data information into the clustering network phi ();
for clustering loss function L c And probability entropy L e Calculating to obtain a clustering loss function L c Value and probability entropy L e A value;
the cluster loss function L c The expression is as follows:
L c ＝λL n +L a
wherein:
in the formula, L n For adjacent loss, L a To distribute losses, x i Is an element, n, in the training data set X of the second communication signal xi Is the calculated nearest neighbor information N of each training data set X of the second communication signal x Φ () is the clustering network, M is the number of all elements in the second communication signal training data set X, k =3, λ and γ are hyper-parameters, λ is used to balance L n And L a ，q i Training the second communication signal training data set X through the clustering network phi () to obtain a distribution probability matrix P N Elements of each column vector after being transformed into the column vector;
training the nearest neighbor information N through the clustering network phi () x Obtaining an assignment probability matrix P N Elements of each column vector after being transformed into the column vector;
the probability entropy L e The expression is as follows:
wherein, H is an entropy function, M is the number of all elements in the second communication signal training data set X, and k =3,q i Training the second communication signal training data set X through the clustering network phi () to obtain an assignment probability matrix P N Elements of each column vector after conversion into column vectors, j being q i Subscripts of elements in the vector;
using the clustering loss function L c And probability entropy L e Performing iterative training on the clustering network phi () to obtain training accuracy and standard mutual information of each time;
and optimizing a loss function by using a batch gradient descent optimization algorithm based on the training accuracy and the standard mutual information, finishing the training of the TDMA signal clustering algorithm model, and obtaining a target signal clustering algorithm model.
Therefore, the implementation of the TDMA signal protocol unsupervised clustering device for the TDMA signal protocol unsupervised clustering described in fig. 2 can process the TDMA signal training data set and the adjacent sample data to obtain the signal protocol unsupervised clustering model, which is beneficial to realizing the fast clustering of the TDMA signal protocol independent of experts and lays a technical foundation for intelligent protocol analysis.
EXAMPLE III
Referring to fig. 6, fig. 6 is a schematic structural diagram of another TDMA signal protocol unsupervised clustering apparatus according to an embodiment of the present invention. The apparatus described in fig. 6 can be applied to an electronic information system, such as a local server or a cloud server for TDMA signal protocol unsupervised cluster management, and the embodiment of the present invention is not limited thereto. As shown in fig. 6, the apparatus may include:
a memory 301 storing executable program code;
a processor 302 coupled to the memory 301;
the processor 302 invokes executable program code stored in the memory 301 to perform the steps of the TDMA signal protocol unsupervised clustering method described in the first embodiment.
Example four
The embodiment of the invention discloses a computer-readable storage medium which stores a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps in the TDMA signal protocol unsupervised clustering method described in the first embodiment.
EXAMPLE five
An embodiment of the invention discloses a computer program product, which comprises a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute the steps in the TDMA signal protocol unsupervised clustering method described in the first embodiment.
The above-described embodiments of the apparatus are merely illustrative, and the modules described as separate parts may or may not be physically separate, and the parts displayed as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above detailed description of the embodiments, those skilled in the art will clearly understand that the embodiments may be implemented by software plus a necessary general hardware platform, and may also be implemented by hardware. Based on such understanding, the above technical solutions may be embodied in the form of a software product, which may be stored in a computer-readable storage medium, wherein the storage medium includes a Read-Only Memory (ROM), a Random Access Memory (RAM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), a One-time Programmable Read-Only Memory (OTPROM), an Electrically Erasable Programmable Read-Only Memory (EEPROM), a Compact Disc-Read-Only Memory (CD-ROM) or other Memory capable of storing data, a magnetic tape, or any other computer-readable medium capable of storing data.
Finally, it should be noted that: the embodiment of the invention discloses an unsupervised clustering method and device for TDMA signal protocol based on reinforcement learning, which is only a preferred embodiment of the invention, and is only used for explaining the technical scheme of the invention, but not limiting the technical scheme; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those skilled in the art; the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims (9)
1. A method for unsupervised clustering of TDMA signal protocols, the method comprising:
s1, acquiring a receiving instruction; the receiving instruction comprises a working frequency band and a line gain parameter;
s2, receiving a target frequency range signal by using the receiving instruction to obtain a TDMA signal data set; the TDMA signal data set comprises service burst TB data, reference burst AB data and synchronous burst RB data;
s3, processing the TDMA signal data set to obtain IQ data; the IQ data is in-phase orthogonal signal data, I is in-phase, and Q is orthogonal with the phase difference of 90 degrees between I and I;
s4, preprocessing the IQ data to obtain a TDMA signal clustering algorithm training set;
s5, training a TDMA signal clustering algorithm model by utilizing the TDMA signal clustering algorithm training set to obtain a target signal clustering algorithm model;
and S6, clustering the received TDMA signals by using the target signal clustering algorithm model to obtain the protocol type of the TDMA signals.
2. The method for unsupervised clustering of TDMA signal protocol according to claim 1, wherein said preprocessing said IQ data to obtain a training set of TDMA signal clustering algorithm, the method comprising:
s41, measuring the IQ data to obtain the length value of each TDMA signal data in the TDMA signal data set, and taking the maximum value as L max ；
S42, for any TDMA signal data, intercepting the TDMA signal data with the length of
The data header of (a), l is the length of the TDMA signal data;
s43, judging whether l is equal to 128 or not to obtain a judgment result;
if the judgment result is negative, the data is supplemented from 0 to 128 to obtain second communication signal data with the shape of [2,128 ];
when the judgment result is yes, directly obtaining second communication signal data with the shape of [2,128 ];
s44, performing WVD time-frequency transformation on the second communication signal data to obtain a time-frequency spectrogram with the shape of [128,128 ]:
in the formula (I), the compound is shown in the specification,
is the instantaneous autocorrelation function R (t, tau) of the signal x (t), t is time shift, tau is an integral variable, f is frequency, S (t, f) is a time-frequency transformation function;
s45, constructing and obtaining training data with the shape of [128,128] by using the time-frequency spectrogram;
s46, marking the time-frequency spectrogram signal protocol type by using the signal protocol type of the time-frequency spectrogram to obtain a training data label with the shape of [128,128 ];
the signal protocol types include: a traffic burst TB, a reference burst AB, a synchronization burst RB;
s47, integrating the training data and the training data labels to obtain the signal clustering algorithm training data;
and S48, fusing all the signal clustering algorithm training data to obtain a TDMA signal clustering algorithm training set.
3. The method for unsupervised clustering of TDMA signal protocol according to claim 1 wherein said TDMA signal clustering algorithm model is comprised of a pre-training module and a cluster training module;
the TDMA signal clustering algorithm training set is utilized to train a TDMA signal clustering algorithm model to obtain a target signal clustering algorithm model, and the method comprises the following steps of;
s51, pre-training the pre-training module by utilizing the TDMA signal training data set to obtain adjacent sample data and a characteristic extraction encoder of the TDMA signal training data set;
the characteristic extraction coder is used for extracting the characteristics of the frequency domain information of the TDMA communication signal to obtain a characteristic expression vector;
s52, training the feature extraction encoder and the clustering training module by using the TDMA signal training data set and the adjacent sample data to obtain a target signal clustering algorithm model;
and the clustering training module trains the feature expression vectors to form a feature library.
4. The method for unsupervised clustering of TDMA signal protocol according to claim 3, wherein said pre-training module using said TDMA signal training dataset, resulting in neighboring sample data and feature extraction encoders of said TDMA signal training dataset, the method comprising:
s511, inputting the TDMA signal training data set after the WVD preprocessing into a standard contrast learning BYOL network;
the contrast learning BYOL network consists of a target network with a parameter theta and an on-line network with a parameter epsilon; said online network is formed by an encoder f θ () Predictor g θ () And mapper q θ () Comprising a target network including an encoder f θ () Sum predictor g θ ()；
The predictor is used for the structural asymmetry of the target network and the online network;
s512, training the TDMA signal training data set by using the contrast learning BYOL network to obtain a feature expression vector;
the TDMA signal training data sets are respectively input into an online network and a target network of the contrast learning BYOL network to obtain q θ (Z θ ) And Z ε By using q θ (Z θ ) And Z ε Design loss function L B ：
In the formula, q θ (Z θ ) For the output of the on-line network, q θ () Predictor, Z θ Output vector, Z, for online network mapper ε Outputting for the target network;
by means of L B The gradient descent updates the on-line network parameter theta, and the updating method of the target network parameter epsilon comprises the following steps:
ε←τε+(1-τ)θ
in the formula, epsilon is an online network parameter, theta is a target network parameter, and tau =0.5 is a hyperparameter;
s513, measuring adjacent sample data of the feature expression vector by using the similarity function to obtain position information of the adjacent sample in the TDMA signal training data set;
the cosine similarity function is expressed as:
wherein a and b are respectively a feature vector, and C (a and b) is the similarity between a and b;
s514, inputting the position information into an encoder for encoding and storing to obtain the adjacent sample data information of the TDMA signal training data set, and inputting the encoder f of the online network θ () And storing to obtain the characteristic extraction encoder.
5. The method for unsupervised clustering of TDMA signal protocol according to claim 3, wherein said training said feature extraction encoder and said cluster training module with said TDMA signal training data set and said neighboring sample data to obtain a target signal clustering algorithm model, comprises:
s521, connecting the feature output extracted by the feature extraction encoder with the clustering module to obtain a clustering network phi (); the clustering module consists of full connection layers;
the cluster training module is used for carrying out classification processing on the extracted characteristic parameters by using a classification model to obtain a classification result;
s522, inputting the TDMA signal training data set and the adjacent sample data information into the clustering network Φ ();
s523, for the clustering loss function L c And probability entropy L e Calculating to obtain a clustering loss function L c Value and probability entropy L e A value;
the cluster loss function L c The expression is as follows:
L c ＝λL n +L a
wherein:
in the formula, L n For an adjacency loss, L a To distribute losses, x i Is an element, n, in the second communication signal training data set X xi Is the calculated nearest neighbor information N of each training data set X of the second communication signal x Wherein Φ () is said clustering network, M is the number of all elements in said second communication signal training data set X, k =3, λ and γ are hyper-parameters, λ is used to balance L n And L a ,q i Training the second communication signal training data set X through the clustering network phi () to obtain an allocation probability matrix P N Elements of each column vector after being transformed into the column vector;
the nearest neighbor information N is trained by the clustering network phi () x Obtaining an assignment probability matrix P N Elements of each column vector after being transformed into the column vector;
the probability entropy L e The expression is as follows:
wherein, H is an entropy function, M is the number of all elements in the second communication signal training data set X, and k =3,q i Training the second communication signal training data set X through the clustering network phi () to obtain an allocation probability matrix P N Elements of each column vector after conversion into column vectors, j being q i Subscripts of elements in the vector;
s524, utilizing the clustering loss function L c And probability entropy L e Performing iterative training on the clustering network phi () to obtain training accuracy and standard mutual information each time;
and S525, optimizing the loss function by using a batch gradient descent optimization algorithm based on the training accuracy and the standard mutual information, finishing the training of the TDMA signal clustering algorithm model, and obtaining the target signal clustering algorithm model.
6. The method for unsupervised clustering of TDMA signal protocols according to claim 1, wherein said processing of said TDMA signal data sets resulting in IQ data, comprises:
carrying out low-noise power amplification, filtering, A/D conversion and digital down-conversion on the TDMA signal data set to obtain IQ data;
s31, amplifying the TDMA signal data set by using a low-noise power amplifier to obtain a first TDMA signal data set;
s32, processing the low-noise TDMA signal data set by using a band-pass filter to obtain a second TDMA signal data set;
s33, performing A/D conversion on the second TDMA signal data set by using an A/D converter to obtain a third TDMA signal data set;
and S34, carrying out digital down-conversion on the third TDMA signal data set by using a digital down-converter to obtain the IQ data.
7. An apparatus for unsupervised clustering of TDMA signal protocols, the apparatus comprising:
the instruction receiving module is used for acquiring a receiving instruction; the receiving instruction comprises a working frequency band and a line gain parameter;
the first processing module is used for receiving a target frequency band signal by using the receiving instruction and acquiring a TDMA signal data set; the TDMA signal data set comprises traffic burst TB data, reference burst AB data, and synchronization burst RB data;
the second processing module is used for processing the TDMA signal data set to obtain IQ data; the IQ data is in-phase orthogonal signal data, I is in-phase, and Q is orthogonal with the phase difference of 90 degrees of I;
the IQ data preprocessing module is used for preprocessing the IQ data to obtain a TDMA signal clustering algorithm training set;
the unsupervised clustering algorithm training module is used for training a TDMA signal clustering algorithm model by utilizing the TDMA signal clustering algorithm training set to obtain a target signal clustering algorithm model;
and the unsupervised clustering module is used for clustering the received TDMA signals by using the target signal clustering algorithm model to obtain the protocol type of the TDMA signals.
8. An apparatus for unsupervised clustering of TDMA signal protocols, the apparatus comprising:
a memory storing executable program code;
a processor coupled with the memory;
the processor invokes the executable program code stored in the memory to perform the TDMA signal protocol unsupervised clustering method of any of claims 1-6.
9. A computer-readable storage medium storing computer instructions which, when invoked, perform the TDMA signal protocol unsupervised clustering method of any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211276019.XA CN115664906B (en) | 2022-10-18 | 2022-10-18 | Method and device for unsupervised clustering of TDMA signal protocol |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211276019.XA CN115664906B (en) | 2022-10-18 | 2022-10-18 | Method and device for unsupervised clustering of TDMA signal protocol |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115664906A true CN115664906A (en) | 2023-01-31 |
| CN115664906B CN115664906B (en) | 2023-05-02 |
Family
ID=84989300
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211276019.XA Active CN115664906B (en) | 2022-10-18 | 2022-10-18 | Method and device for unsupervised clustering of TDMA signal protocol |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115664906B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116578874A (en) * | 2023-07-10 | 2023-08-11 | 武汉能钠智能装备技术股份有限公司四川省成都市分公司 | Satellite signal attribute appraising method and device based on network protocol |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107682216A (en) * | 2017-09-01 | 2018-02-09 | 南京南瑞集团公司 | A kind of network traffics protocol recognition method based on deep learning |
| CN108256488A (en) * | 2018-01-19 | 2018-07-06 | 中国人民解放军陆军装甲兵学院 | A kind of radar target identification method based on micro-Doppler feature extraction and deep learning |
| CN110071885A (en) * | 2019-04-17 | 2019-07-30 | 成都华日通讯技术有限公司 | A kind of deep learning method of discrimination of PSK digital signal subclass Modulation Identification |
| CN113971440A (en) * | 2021-10-21 | 2022-01-25 | 中国人民解放军空军航空大学 | Unsupervised radar signal sorting method based on deep clustering |
| CN114666273A (en) * | 2022-04-30 | 2022-06-24 | 东南大学 | Application layer unknown network protocol oriented traffic classification method |
| US20220217619A1 (en) * | 2021-01-06 | 2022-07-07 | Bae Systems Information And Electronic Systems Integration Inc. | Artificial intelligence radio classifier and identifier |
| CN115037805A (en) * | 2022-06-08 | 2022-09-09 | 中国人民解放军陆军工程大学 | Unknown network protocol identification method, system, device and storage medium based on deep clustering |
| CN115130581A (en) * | 2022-04-02 | 2022-09-30 | 北京百度网讯科技有限公司 | Sample generation method, training method, data processing method and electronic device |
-
2022
- 2022-10-18 CN CN202211276019.XA patent/CN115664906B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107682216A (en) * | 2017-09-01 | 2018-02-09 | 南京南瑞集团公司 | A kind of network traffics protocol recognition method based on deep learning |
| CN108256488A (en) * | 2018-01-19 | 2018-07-06 | 中国人民解放军陆军装甲兵学院 | A kind of radar target identification method based on micro-Doppler feature extraction and deep learning |
| CN110071885A (en) * | 2019-04-17 | 2019-07-30 | 成都华日通讯技术有限公司 | A kind of deep learning method of discrimination of PSK digital signal subclass Modulation Identification |
| US20220217619A1 (en) * | 2021-01-06 | 2022-07-07 | Bae Systems Information And Electronic Systems Integration Inc. | Artificial intelligence radio classifier and identifier |
| CN113971440A (en) * | 2021-10-21 | 2022-01-25 | 中国人民解放军空军航空大学 | Unsupervised radar signal sorting method based on deep clustering |
| CN115130581A (en) * | 2022-04-02 | 2022-09-30 | 北京百度网讯科技有限公司 | Sample generation method, training method, data processing method and electronic device |
| CN114666273A (en) * | 2022-04-30 | 2022-06-24 | 东南大学 | Application layer unknown network protocol oriented traffic classification method |
| CN115037805A (en) * | 2022-06-08 | 2022-09-09 | 中国人民解放军陆军工程大学 | Unknown network protocol identification method, system, device and storage medium based on deep clustering |
Non-Patent Citations (2)
| Title |
|---|
| 林荣强;李鸥;李青;刘琰;: "基于半监督聚类集成的未知网络协议识别方法" * |
| 林荣强;李鸥;李青;刘琰;: "基于半监督聚类集成的未知网络协议识别方法", 小型微型计算机*** * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116578874A (en) * | 2023-07-10 | 2023-08-11 | 武汉能钠智能装备技术股份有限公司四川省成都市分公司 | Satellite signal attribute appraising method and device based on network protocol |
| CN116578874B (en) * | 2023-07-10 | 2023-11-21 | 武汉能钠智能装备技术股份有限公司四川省成都市分公司 | Satellite signal attribute appraising method and device based on network protocol |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115664906B (en) | 2023-05-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107770263B (en) | safe access method and system for Internet of things terminal based on edge calculation | |
| CN110197127B (en) | Wireless signal detection and electromagnetic interference classification system and method based on deep learning | |
| CN115664906A (en) | Unsupervised clustering method and unsupervised clustering device for TDMA signal protocol | |
| CN110838133B (en) | Multi-target tracking method and related equipment | |
| CN114154545B (en) | Intelligent unmanned aerial vehicle measurement and control signal identification method under strong mutual interference condition | |
| CN111968046B (en) | Target association fusion method for radar photoelectric sensor based on topological structure | |
| CN111050315B (en) | Wireless transmitter identification method based on multi-core two-way network | |
| CN111915015B (en) | Abnormal value detection method and device, terminal equipment and storage medium | |
| CN112364916B (en) | Image classification method based on transfer learning, related equipment and storage medium | |
| CN110826403A (en) | Tracking target determination method and related equipment | |
| CN113075462B (en) | Electromagnetic field distribution positioning method based on neural network | |
| CN115577305B (en) | Unmanned aerial vehicle signal intelligent recognition method and device | |
| CN110188753A (en) | One kind being based on dense connection convolutional neural networks target tracking algorism | |
| CN114218984A (en) | Radio frequency fingerprint identification method based on sample multi-view learning | |
| CN103345552A (en) | Method and device for assessing reliability of power ICT communication network | |
| CN108833145B (en) | Segmented two-dimensional predistortion system and method using self-organizing mapping network | |
| CN112055419B (en) | Communication signal correlation method based on statistics | |
| CN115600101A (en) | Unmanned aerial vehicle signal intelligent detection method and device based on priori knowledge | |
| CN110471030A (en) | Based on the radio frequency tomography passive type Position-Solving method for improving conjugate gradient | |
| CN113052219B (en) | Abnormal track detection method and device and electronic equipment | |
| WO2022018847A1 (en) | Signal specification identifying device, control circuit, and program storage medium | |
| CN114333840A (en) | Voice identification method and related device, electronic equipment and storage medium | |
| Buhler et al. | The CMW extraction algorithm for MTP/MPEF | |
| CN115333903B (en) | Synchronization head detection method, synchronization device, receiver and communication system | |
| CN116431991B (en) | Satellite communication radiation source individual identification method based on phase modulation signal feature fusion |
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 |