CN119052040B

CN119052040B - MPSK signal modulation mode identification method, device, medium and product

Info

Publication number: CN119052040B
Application number: CN202411262720.5A
Authority: CN
Inventors: 臧淼; 邢志强; 李响
Original assignee: North China University of Technology
Current assignee: North China University of Technology
Priority date: 2024-09-10
Filing date: 2024-09-10
Publication date: 2025-09-02
Anticipated expiration: 2044-09-10
Also published as: CN119052040A

Abstract

The present invention discloses a method, device, medium and product for identifying MPSK signal modulation modes, relating to the field of signal recognition. The method comprises obtaining a modulation signal sample and extracting a modulation signal constellation diagram based on the modulation signal sample; denoising and restoring the modulation signal constellation diagram using a trained constellation diagram pre-noising and denoising model; the trained constellation diagram pre-noising and denoising model is a one-dimensional convolutional neural network and includes an embedding layer; and performing MSE-based feature matching on the denoised modulation signal constellation diagram and the ideal constellation diagrams of different MPSK modulation modes to obtain a modulation mode recognition result. The present invention can improve the accuracy and efficiency of signal recognition and can be applied to blind demodulation scenarios in non-cooperative communications.

Description

MPSK signal modulation mode identification method, equipment, medium and product

Technical Field

The present invention relates to the field of signal identification, and in particular, to a method, an apparatus, a medium, and a product for identifying a MPSK signal modulation mode.

Background

In recent years, with rapid development of digital communication technology, wireless signal transmission technology is widely used in the fields of mobile communication, satellite communication, radar, etc., and is one of the important means for transmitting information. However, in the transmission process, due to the influence of environmental noise interference, channel influence, multipath effect and other factors, the signal is easy to be interfered, so that signal information is lost and distorted, and the signal receiving quality and the demodulation performance of the system are further influenced. Therefore, in order to improve the performance of the wireless communication system, the transmission quality of the signal is guaranteed, and the enhancement and identification of the signal are particularly critical for subsequent demodulation.

Conventional signal processing generally occurs in the case of cooperative communication, that is, the receiving end knows information such as parameters, modulation modes, and the like of a received signal. However, in practical applications, the receiving end may have a problem that the signal parameters of the transmitting end are not known, and it is necessary to estimate the parameters and then perform demodulation processing. This communication scheme requiring a preliminary estimation of the received signal parameters is called uncooperative communication.

Uncooperative communication is of great strategic importance in the military field. For example, in the field of radio detection, when an enemy target appears in a space, under the condition of not affecting a transmitting end, a radio signal transmitted by an enemy is intercepted by a radio technology, so that the reception and the cracking of enemy signals are completed, and powerful support is provided for enemy analysis and battlefield decision. Therefore, the uncooperative communication technology plays an important role in military investigation and has a vital significance for maintaining national security.

In uncooperative communication, demodulation of a completed signal is referred to as blind demodulation. The wireless signal modulation mode identification is between signal receiving and demodulation, plays a crucial role in blind demodulation, and is a precondition for realizing correct signal demodulation. Along with the wide application of the deep learning technology, the intelligent recognition technology of the signal modulation mode is widely applied by virtue of the advantages of being capable of automatically learning signal characteristics, better processing complex data, stronger in flexibility and the like.

The types of modulation signals in the wireless communication system are various, wherein the multi-system digital phase modulation (Multiple PHASE SHIFT KEYING, MPSK) is one of the common modulation types, has stronger anti-interference performance and good safety, is widely applied to the civil and military communication fields, and has strong practical significance as a research object.

In a non-cooperative communication environment, a receiver cannot acquire relevant parameter information of a received signal, and noise interference is easily caused in a transmission process, so that a modulation mode of the signal is difficult to accurately identify. Therefore, how to improve the recognition effect of the modulated signal has been a challenging task.

The modulation type recognition method is mainly divided into a likelihood ratio based decision theory recognition method and a statistical pattern based recognition method.

Based on likelihood ratio decision theory identification method, hypothesis is established according to statistical characteristics of signals, after a cost function is determined, a proper threshold is selected for comparison through calculating data likelihood ratio, and finally modulation type identification is achieved. However, such methods have certain limitations, such as high complexity of likelihood function calculation, and need to obtain accurate signal prior probability distribution information in advance. In contrast, by extracting relevant features from the signal to be identified based on the statistical pattern recognition method, and then training the classification model by utilizing the most distinguishable features, the trained classification model can realize signal modulation type recognition. The method has stronger adaptability, relatively lower complexity, stronger robustness and capability of adapting to noise change to a certain extent. Therefore, in the field of non-cooperative communication, the statistical pattern recognition method is widely used.

The modulation signal has various characteristics, besides the traditional signal spectrum, high-order cumulant and instantaneous characteristics, the constellation diagram is widely applied to the field of digital signal modulation type identification because of the advantages of intuitiveness, simplicity, easiness in implementation, strong anti-interference performance and the like. In 2020, dean et al propose a modulation type recognition method that combines deep learning with constellation, which can achieve a classification accuracy of about 87% at low signal-to-noise ratio. The constellation position is shifted due to the fact that the signal is susceptible to noise interference during transmission. In order to eliminate redundancy in the original constellation, key information is extracted from the original constellation to synthesize an enhanced constellation, song et al propose an enhanced constellation modulation recognition method based on a convolutional neural network, and the constellation is improved by capturing correlation information between adjacent points of a signal constellation and combining amplitude information into a pixel value. However, the methods treat the constellation diagram as an image, and directly introduce image classification and identification methods in deep learning, so that the processing speed is low and the computational complexity is high.

Therefore, to improve the accuracy and efficiency of signal recognition, it is desirable to provide a new signal recognition method or system.

Disclosure of Invention

The invention aims to provide a MPSK signal modulation mode identification method, equipment, medium and product, which can improve the accuracy and efficiency of signal identification and can be applied to a blind demodulation scene of non-cooperative communication.

In order to achieve the above object, the present invention provides the following solutions:

a method for identifying a modulation mode of an MPSK signal, the method comprising:

obtaining a certain number of modulation signal samples, and extracting a modulation signal constellation diagram according to the modulation signal;

Denoising and recovering the modulated signal constellation diagram by using a trained constellation diagram pre-denoising model (Add-Denoising Models for Constellation, ADMC), wherein the trained ADMC is a one-dimensional convolutional neural network and comprises embedding layers;

and performing MSE-based feature matching on the denoised modulation signal constellation and ideal constellation of different MPSK modulation modes to obtain a recognition result of the modulation mode.

Optionally, the obtaining a modulation signal and extracting a modulation signal constellation according to the modulation signal specifically includes:

Carrying out carrier frequency estimation on the modulated signal based on a Welch power spectrum method;

estimating the symbol rate based on a complex envelope detection method;

the symbol synchronization is realized by a self-synchronization method based on the Gardner algorithm;

carrier synchronization is achieved based on Costas in-phase quadrature loops, resulting in a modulated signal constellation.

Optionally, the process of determining the trained denoising diffusion probability model is as follows:

constructing a training data set according to ideal BPSK modulation signal constellation diagram samples;

the input layer of the ADMC convolutional neural network is used for receiving a modulation signal constellation diagram sample, and then the modulation signal constellation diagram sample passes through a convolutional layer, a BN layer, a ReLU activation function layer and a full connection layer respectively, and finally predicted noise data is output;

training a data set sample for noise addition based on ADMC pre-noise addition processes;

Training ADMC a convolutional neural network by using the denoised training data set samples, and learning ADMC a denoising process to obtain a trained ADMC applied to the BPSK modulation signal;

A trained ADMC applied to the QPSK modulated signal and the 8PSK modulated signal is determined.

Optionally, performing MSE-based feature matching on the denoised modulation signal constellation and ideal constellations of different MPSK modulation modes to obtain a recognition result of the modulation mode, which specifically includes:

ideal constellation pattern points of three modulation signals of BPSK, QPSK and 8PSK are used as matching templates;

matching and comparing the denoised modulation signal constellation with a matching template based on MSE;

judging the similarity degree of the denoised modulation signal constellation diagram and the matching template according to the comparison result;

And selecting the best matching modulation mode as the modulation mode of the modulation signal according to the MSE minimum rule.

A computer device includes a memory, a processor to store a computer program on the memory and executable on the processor, the processor executing the computer program to implement the one MPSK signal modulation mode identification method.

A computer readable storage medium having stored thereon a computer program which when executed by a processor implements the one MPSK signal modulation mode recognition method.

A computer program product comprising a computer program which when executed by a processor implements the one MPSK signal modulation mode identification method.

According to the specific embodiment provided by the invention, the invention discloses the following technical effects:

According to the method, the device, the medium and the product for identifying the MPSK signal modulation mode, which are provided by the invention, the sample point coordinates of the modulated signal constellation diagram are used as input data, the trained ADMC is utilized to conduct denoising recovery on the modulated signal constellation diagram, and then the denoised modulated signal constellation diagram is subjected to MSE-based feature matching with ideal constellation diagrams of different MPSK modulation modes, so that intelligent identification of the modulation mode is realized. The method and the device can improve the accuracy and the efficiency of signal identification, and can be applied to a blind demodulation scene of non-cooperative communication.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a flow chart of a MPSK signal modulation method recognition method provided in embodiment 1 of the present invention;

fig. 2 is a schematic diagram of a ADMC convolutional neural network according to embodiment 1 of the present invention;

Fig. 3 is a diagram of a signal modulation recognition framework provided in embodiment 1 of the present invention;

FIG. 4 is a graph showing recognition results of three modulation signals under different signal-to-noise ratios;

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.

Example 1

As shown in fig. 1, the method for identifying a modulation mode of a MPSK signal provided by the present invention includes:

s101, obtaining a modulation signal, and extracting a modulation signal constellation diagram according to the modulation signal;

The method specifically comprises the following steps:

estimating the symbol rate based on a complex envelope detection method;

S102, denoising and recovering a modulation signal constellation diagram by using a trained ADMC, wherein the trained ADMC is a one-dimensional convolutional neural network and comprises embedding layers;

the trained ADMC determination process is as follows:

And constructing a training data set according to ideal BPSK modulation signal constellation diagram samples, wherein each sample represents a sample point in the constellation diagram by two-dimensional coordinates. The modulation signal type is the same for all samples.

The method comprises the steps of constructing ADMC a convolutional neural network, wherein an input layer of the ADMC convolutional neural network is used for receiving a modulation signal constellation diagram sample, then the modulation signal constellation diagram sample passes through a convolutional layer, a BN (BatchNormalization) layer, a ReLU (Rectified LinearUnit) activation function layer and a full connection layer respectively, and finally predicted noise data is output, and the ADMC convolutional neural network comprises three convolutional layers in total, wherein the first layer of convolutional kernel size is set to 2, and the second layer and the third layer of convolutional kernel size are set to 1. And a BN layer and a ReLU activation function layer are respectively used behind each convolution layer, wherein the BN layer has the function of normalizing input data, improving training speed, preventing gradient explosion and effectively improving the stability of the model. The function of the ReLU layer is to enhance the nonlinear relation among the network layers of the model by introducing an activation function, so as to improve the sparsity of the model. Since ADMC contains time parameters in the whole training process, embedding layers are added into the model framework, so that the model considers the influence of time steps in the generating and predicting processes, and the generated data is processed better.

Training data set samples are subjected to noise adding based on ADMC constellation diagram pre-noise adding process until training constellation diagram sample points are scattered in three MPSK constellation point circumscribed circles in a Gaussian noise mode, wherein the specific noise adding principle is as follows:

For MPSK modulated signals, this can be expressed as:

S_N＝A_mm(t)cos(2πf_ct+φ_c)+n(t) (1)

Wherein A _m represents the signal amplitude, m (t) represents the low-pass pulse signal, cos (2πf _ct+φ_c) represents the high-frequency carrier signal, f _c represents the carrier frequency, φ _c represents the initial phase of the carrier signal, and n (t) represents the receiving-side noise. The method comprises the following steps of:

n _i(t)、n_q (t) represents the noise component of the noise after the carrier frequency removal process, and still satisfies the same distribution as n (t).

As known from the MPSK modulation principle, each modulated symbol can be expressed in the form of a vector as:

where, δ _g is the signal energy, The vector corresponds to one constellation point on the constellation plane.

The constellation diagram pre-noise adding method based on ADMC comprises the following steps:

(a) L original MPSK modulated signal samples (without noise) are selected, all samples being taken as original elements S0 according to the constellation point set determined in (5).

(B) And generating a random Gaussian noise sequence, wherein the random Gaussian noise sequence accords with N (0,I), the mean value is 0, the variance is I, the I is an L-dimensional vector, and the element is 1. And taking the generated noise sequence as a label of the pre-noise training.

(C) An integer N is randomly selected in the range of 0-10000 as a step selection for training epoch once.

(D) Starting from S0, dividing into N steps, and successively adding noise to S0 in the following mode:

η _t is 0 to 1, η _t is smaller and smaller as t increases, S _t represents a modulated signal sample obtained after adding noise t times, ε _t represents noise added every time, and all t are subjected to the same distribution, namely, N (0,I) is met, the mean value is 0, the variance is I, the I is an L-dimensional vector, and the element is 1.

Training ADMC a convolutional neural network by using the denoised training data set samples (and the labels generated in the step b), and learning ADMC a reverse denoising process to obtain a trained ADMC applied to the BPSK modulation signal;

Ideal QPSK, 8PSK modulated signal constellation samples are generated to determine trained ADMC that is applied to the QPSK modulated signal and the 8PSK modulated signal.

And S102, respectively sending the three channels into a trained ADMC model for denoising the BPSK, QPSK and 8PSK modulation signals, performing a reverse denoising process, and generating denoised constellation pattern point coordinates by gradually predicting and removing noise.

And S103, performing MSE-based feature matching on the denoised modulation signal constellation and ideal constellations of different MPSK modulation modes to obtain a recognition result of the modulation modes.

As shown in fig. 3, S103 specifically includes:

judging the similarity degree of the denoised and recovered modulation signal constellation diagram and the matching template according to the comparison result;

For three modulation signals of BPSK, QPSK and 8PSK, the experimental parameters are set as follows, the number of code elements of the signal data is set to 1000, the sampling frequency f _s =10 kHz, the carrier frequency f _c =1.5 kHz, and the channel is a Gaussian white noise channel. The signal-to-noise ratio range is set to be-5-10 db, and 300 Monte Carlo simulation experiments are carried out on the signals under each signal-to-noise ratio by taking 1db as a step to obtain the identification accuracy of each modulation signal, and the result is shown in fig. 4.

As shown in fig. 4, as the signal-to-noise ratio increases, the recognition accuracy of the three modulation signals tends to steadily increase. The identification accuracy of the invention reaches 100.0% when the signal-to-noise ratio of the BPSK signal is 0db, when the QPSK signal is 4db and when the 8PSK signal is 8 db.

In order to verify the effective improvement of the invention on the recognition effect, the invention is compared with three methods based on high-order accumulation, convolutional neural network and ISPP-ResNet aiming at the conditions of 0db, 5db and 10db of signal to noise ratio. The recognition results are summarized in tables 1, 2 and 3, respectively, in order to intuitively demonstrate the performance of the different methods at low signal-to-noise ratios.

Table 1 comparison of identification properties of three modulated signals under different methods (0 db)

Table 2 comparison of identification properties of three modulated signals under different methods (5 db)

Table 3 comparison of identification properties of three modulated signals under different methods (10 db)

According to the data in table 1, the method presented herein works better at identifying BPSK signals when the signal-to-noise ratio is 0db than the other three methods. Although the recognition of QPSK signals is slightly inferior to ISPP-ResNet < 18 >, and the recognition effect of 8PSK signals is slightly lower than that of convolutional neural networks, the overall average recognition rate is still significantly superior to the other three methods. When the data in the table 2 are observed and the signal to noise ratio is improved to 5db, the identification effect of the invention on three modulation signals is better than that of other three methods, and the identification rate of BPSK signals and QPSK signals reaches 100%. As the signal-to-noise ratio is further increased to 10db, as shown in Table 3, it can be seen that all methods exhibit excellent recognition effects under the condition of high signal-to-noise ratio, and the recognition rate of the invention for three modulation signals reaches 100%. Therefore, the invention obtains good recognition effect under the 0db signal-to-noise ratio and obtains the best recognition effect under the 5db signal-to-noise ratio.

In addition, the signal constellation pattern points are used as data to be processed, so that the calculation complexity is low.

Example 2

A computer device comprising a memory, a processor to store a computer program on the memory and executable on the processor, the processor executing the computer program to implement a MPSK signal modulation mode identification method of embodiment 1.

Example 3

A computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements a MPSK signal modulation mode identification method of embodiment 1.

Example 4

A computer program product comprising a computer program which when executed by a processor implements a MPSK signal modulation mode identification method of embodiment 1.

Example 5

A computer device may be a database. The computer device includes a processor, a memory, an Input/Output interface (I/O) and a communication interface. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface is connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used to store the pending transactions. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a MPSK signal modulation mode identification method in embodiment 1.

It should be noted that, the object information (including, but not limited to, object device information, object personal information, etc.) and the data (including, but not limited to, data for analysis, stored data, presented data, etc.) related to the present invention are both information and data authorized by the object or sufficiently authorized by each party, and the collection, use and processing of the related data need to comply with the related laws and regulations and standards of the related countries and regions.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magneto-resistive random access Memory (Magnetoresistive RandomAccess Memory, MRAM), ferroelectric Memory (Ferroelectric RandomAccess Memory, FRAM), phase change Memory (PHASE CHANGE Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (RandomAccess Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in various forms such as Static RandomAccessMemory, SRAM or dynamic Dynamic RandomAccess Memory (DRAM) and the like. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present invention may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The principles and embodiments of the present invention have been described herein with reference to specific examples, which are intended to facilitate an understanding of the principles and concepts of the invention and are to be varied in scope and detail by persons of ordinary skill in the art based on the teachings herein. In view of the foregoing, this description should not be construed as limiting the invention.

Claims

1. A method for identifying an MPSK signal modulation mode, characterized in that the method comprises:

Obtain a certain number of modulation signal samples and extract the modulation signal constellation diagram according to the modulation signal;

Denoising the modulated signal constellation using a trained constellation pre-noising and denoising model ADMC; the trained ADMC is a one-dimensional convolutional neural network and includes an embedding layer; the ADMC is a constellation pre-noising and denoising model; the embedding layer is an embedding layer;

The denoised modulation signal constellation is matched with the ideal constellation of different MPSK modulation modes based on MSE to obtain the modulation mode recognition result; MPSK modulation mode is multi-phase shift keying; MSE is mean square error;

The obtaining of a certain number of modulation signal samples and extracting a modulation signal constellation diagram according to the modulation signal specifically includes:

Carrier frequency estimation of modulated signal is performed based on Welch power spectrum method, which is a modern spectrum analysis method.

Symbol rate estimation based on complex envelope detection method;

Symbol synchronization is achieved using a self-synchronization method based on the Gardner algorithm; the Gardner algorithm is a non-data-assisted timing error detection algorithm;

Carrier synchronization is achieved based on the Costas in-phase orthogonal loop to obtain the modulation signal constellation diagram;

Costas is a Costas loop, a feedback control system used for carrier synchronization;

The process of determining the trained ADMC is:

Construct a training data set based on the ideal BPSK modulated signal constellation sample;

Construct an ADMC convolutional neural network; the input layer of the ADMC convolutional neural network is used to receive the modulation signal constellation sample, and then the modulation signal constellation sample passes through the convolution layer, BN layer, ReLU activation function layer and fully connected layer respectively, and finally outputs the predicted noise data; the BN layer is a normalization layer;

Noise addition to training dataset samples based on ADMC constellation diagram noising process;

The ADMC convolutional neural network is trained using the noisy training dataset samples to learn the ADMC denoising process and obtain the trained ADMC applied to the BPSK modulated signal.

Based on the ideal QPSK modulated signal constellation diagram samples and the ideal 8PSK modulated signal constellation diagram samples, the ADMC convolutional neural network is trained respectively to learn the ADMC denoising process and determine the trained ADMC applied to the QPSK modulated signal and the 8PSK modulated signal;

The modulation signal constellation after denoising is matched with the ideal constellation of different MPSK modulation modes based on MSE features to obtain the recognition result of the modulation mode, specifically including:

The ideal constellation sample points of BPSK, QPSK and 8PSK modulation signals are used as matching templates;

The denoised modulation signal constellation is matched and compared with the matching template based on MSE;

The similarity between the denoised modulation signal constellation and the matching template is determined based on the comparison result;

According to the minimum MSE rule, the most matching modulation mode is selected as the modulation mode of the modulated signal.

2. A computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the MPSK signal modulation mode identification method as claimed in claim 1.

3. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for identifying the MPSK signal modulation mode as claimed in claim 1 is implemented.

4. A computer program product, comprising a computer program, characterized in that when the computer program is executed by a processor, it implements the MPSK signal modulation mode identification method described in claim 1.