CN111884962A - A method and system for classification of signal modulation types based on convolutional neural network - Google Patents

Abstract

The invention discloses a signal modulation type classification method and system based on a convolutional neural network, comprising: extracting modulation features from a received electromagnetic signal, the modulation features including an amplitude feature and a first phase feature; The carrier frequency of one phase feature is used to obtain the second phase feature; the constellation map is performed on the amplitude feature and the second phase feature to obtain the feature image; the pre-built convolutional neural network model is used to classify the feature image to obtain the feature image to which the feature image belongs. Modulation type, the electromagnetic signal is demodulated according to the modulation type. The modulation features of electromagnetic signals are extracted, and carrier removal is performed on the modulation features to reduce the influence of frequency offset; the processed modulation features are converted into feature images by constellation map mapping method, and the feature images are classified by convolutional neural network. The recognition is converted into image recognition, which realizes the classification and recognition of the modulation type of electromagnetic signals in the communication system.

Description

A method and system for classification of signal modulation types based on convolutional neural network

technical field

The present invention relates to the technical field of wireless communication, in particular to a method and system for classifying signal modulation types based on a convolutional neural network.

Background technique

The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

In the field of communication, in the analysis and processing of communication signals, it is necessary to first identify the modulation type of the signal, correctly identify the modulation type, and then demodulate the signal according to the modulation type; there are currently two methods for signal modulation identification: that is, based on similar The former is more sensitive to parameter deviation and model mismatch, and it is difficult to form correct assumptions, and it is difficult to determine the correct decision threshold; the key of the latter is The selection of feature parameters, the recognition results are easily disturbed by noise. Therefore, the inventor believes that the signal modulation identification method in the communication system has problems such as low accuracy and great influence by interference such as frequency offset.

SUMMARY OF THE INVENTION

In order to solve the above problems, the present invention proposes a signal modulation type classification method and system based on convolutional neural network, which extracts the modulation characteristics of electromagnetic signals, removes the carrier wave from the modulation characteristics, and reduces the influence of frequency offset; The modulation feature is converted into a feature image by constellation map mapping method, and the feature image is classified by convolutional neural network, and the modulation recognition is converted into image recognition, so as to realize the classification and recognition of the electromagnetic signal modulation type in the communication system.

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

In a first aspect, the present invention provides a method for classifying signal modulation types based on a convolutional neural network, comprising:

extracting modulation features from the received electromagnetic signal, the modulation features including amplitude features and first phase features;

The carrier frequency of the first phase characteristic is filtered by the carrier estimation algorithm to obtain the second phase characteristic;

Perform constellation map mapping on the amplitude feature and the second phase feature to obtain a feature image;

A pre-built convolutional neural network model is used to classify the characteristic images, and the modulation type to which the characteristic images belong is obtained, and the electromagnetic signal is demodulated according to the modulation type.

In a second aspect, the present invention provides a signal modulation type classification system based on a convolutional neural network, comprising:

a feature extraction module, configured to extract modulation features from the received electromagnetic signal, where the modulation features include amplitude features and first phase features;

a carrier processing module, used for filtering the carrier frequency of the first phase characteristic by adopting a carrier estimation algorithm to obtain the second phase characteristic;

The constellation map mapping module is used to perform constellation map mapping on the amplitude feature and the second phase feature to obtain a feature image;

The classification module is used for classifying the characteristic images by using a pre-built convolutional neural network model, obtaining the modulation type to which the characteristic images belong, and demodulating the electromagnetic signal according to the modulation type.

In a third aspect, the present invention provides an electronic device, comprising a memory, a processor, and computer instructions stored in the memory and executed on the processor, and when the computer instructions are executed by the processor, the method described in the first aspect is completed .

In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, and when the computer instructions are executed by a processor, the method described in the first aspect is completed.

Compared with the prior art, the beneficial effects of the present invention are:

The invention adopts the convolutional neural network to automatically extract the characteristics of the characteristic picture of the modulated signal, and realizes the identification of the modulation type. The components are eliminated, which greatly reduces the influence of frequency offset on signal feature extraction, effectively improves the purity of signal modulation features, and extracts signal differences to the greatest extent.

The invention converts modulation features into feature images, expresses differences intuitively, converts modulation recognition problems into image recognition problems, and finally uses a convolutional neural network model to classify feature images, thereby effectively improving the classification accuracy.

Description of drawings

The accompanying drawings forming a part of the present invention are used to provide further understanding of the present invention, and the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, and do not constitute an improper limitation of the present invention.

FIG. 1 is a flowchart of processing a received signal according to Embodiment 1 of the present invention;

2(a) is a constellation diagram of a BPSK signal provided by Embodiment 1 of the present invention;

FIG. 2(b) is a constellation diagram of a QPSK signal provided by Embodiment 1 of the present invention;

Figure 2 (c) is a constellation diagram of an 8PSK signal provided by Embodiment 1 of the present invention;

FIG. 2(d) is a constellation diagram of a 16QAM signal provided by Embodiment 1 of the present invention;

FIG. 2(e) is a constellation diagram of a 32QAM signal provided by Embodiment 1 of the present invention;

2(f) is a constellation diagram of a 64QAM signal provided by Embodiment 1 of the present invention;

3 is a frame diagram of a convolutional neural network modulation identification provided in Embodiment 1 of the present invention;

FIG. 4 is a model loss diagram used in the training process of the convolutional neural network modulation recognition framework provided by Embodiment 1 of the present invention;

FIG. 5 is a model accuracy diagram used in the training process of the convolutional neural network modulation and identification framework provided by Embodiment 1 of the present invention;

FIG. 6 is a confusion matrix of a prediction result of a convolutional neural network identification framework provided in Embodiment 1 of the present invention.

Detailed ways:

The present invention will be further described below with reference to the accompanying drawings and embodiments.

It should be noted that the following detailed description is exemplary and intended to provide further explanation of the invention. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

It should be noted that the terminology used herein is for the purpose of describing specific embodiments only, and is not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly dictates otherwise, the singular is intended to include the plural as well, furthermore, it is to be understood that the terms "including" and "having" and any conjugations thereof are intended to cover the non-exclusive A process, method, system, product or device comprising, for example, a series of steps or units is not necessarily limited to those steps or units expressly listed, but may include those steps or units not expressly listed or for such processes, methods, Other steps or units inherent in the product or equipment.

Embodiments of the invention and features of the embodiments may be combined with each other without conflict.

Example 1

As shown in FIG. 1 , this embodiment provides a signal modulation classification method based on a convolutional neural network. This embodiment is applied to technical fields such as wireless communication technology, communication countermeasures, and electromagnetic spectrum management. Network, classify 6 types of modulation signal types such as BPSK, QPSK, 8PSK, 16QAM, 32QAM, 64QAM; specifically include:

S1: extract modulation features from the received electromagnetic signal, where the modulation features include amplitude features and initial phase features;

S2: Use the carrier estimation algorithm to filter the carrier frequency of the initial phase characteristic to obtain the phase characteristic;

S3: perform constellation map mapping on the amplitude feature and the phase feature to obtain a feature image;

S4: Use a pre-built convolutional neural network model to classify the feature image, obtain the modulation type to which the feature image belongs, and demodulate the electromagnetic signal according to the modulation type.

In described step S1, use signal generator to set the parameter that test signal produces, produce PSK and QAM signal of different modulation order;

In this embodiment, the signal generator adopts SMW200A; modulation signals of different modulation orders include: BPSK, QPSK, 8PSK, 16QAM, 32QAM, 64QAM and other six types.

The antenna is used to receive the electromagnetic signal to be modulated, and the signal IQ acquisition toolbox is used at the terminal to receive and sample the electromagnetic signal connected to the antenna of the PC terminal, and save the I/Q two-way data of the sampled electromagnetic signal sample in .mat format file.

具体为：In the step S1, according to the modulation principle of the PSK and QAM signals, the modulation characteristics related to the modulation type, that is, the amplitude characteristics a(n) and the phase characteristics, are extracted from the electromagnetic signals in the received IQ samples.

Specifically:

S101: The I/Q data expression of the received electromagnetic signal is:

为初始相位；f_s为采样频率，n＝1,2,3,…,N，采样时间t＝N/f_s；where a(n) is the amplitude of the electromagnetic signal sample, f(n) is the instantaneous frequency,

is the initial phase; f _s is the sampling frequency, n=1, 2, 3, ..., N, and the sampling time t=N/f _s ;

S102: The amplitude characteristic is:

S103: The phase characteristics are:

的取值范围在[-π,π]内，需要根据I(n)、Q(n)的正负确定

在第几象限，相位特征具体的求解公式为：S104: Since the value range of the arctangent function is [-π/2, π/2], and

The value range of is within [-π,π], which needs to be determined according to the positive and negative of I(n) and Q(n).

In which quadrant, the specific solution formula of the phase characteristic is:

Determine the positive and negative score conditions of the IQ sequence, and solve according to this formula

中的载波频率分量，减少频偏对调制识别的影响，得到基带信号的相位信息

In the step S2, since the extracted phase features contain carrier frequency components, this embodiment adopts a carrier estimation algorithm to estimate the carrier of the signal, and filter the phase features of the electromagnetic signal.

The carrier frequency component in the baseband signal reduces the influence of frequency offset on modulation identification, and obtains the phase information of the baseband signal

In this embodiment, the carrier estimation algorithm adopts the Rife algorithm, that is, the bispectral method, and the steps of realizing the carrier frequency estimation according to the Rife algorithm are:

S201: Import electromagnetic signal IQ samples into MATLAB, count the sampling rate f _s of the electromagnetic signal IQ samples, and the number of points N of the fast Fourier transform FFT, and calculate Δf=f _s /N;

S202: Perform N-point FFT transformation on the electromagnetic signal IQ samples to obtain the maximum value M ₁ of the spectrum, and record its subscript index value x ₀ ;

S203: According to x ₂ =x ₀ ±1, obtain the next largest value M ₂ of the spectrum, and record its subscript index value x ₂ ;

S204: Calculate the absolute value |δ| of the relative deviation between the frequency of the electromagnetic signal and the corresponding frequency at the maximum value of the FFT as:

S205: Compare the sizes of x ₀ and x ₂ ;

若x₂>x₀，该式取加号；若x₂<x₀，该式取减号，对FFT中频谱最大处的频率值x₀进行插值，得到准确的频率估计值；S206: According to

If x ₂ >x ₀ , the formula takes the plus sign; if x ₂ <x ₀ , the formula takes the minus sign, and the frequency value x ₀ at the maximum frequency spectrum in the FFT is interpolated to obtain an accurate frequency estimate;

后，计算基带信号的初始相位信息

其计算公式为：S207: Use the Rife algorithm to estimate the carrier frequency

After that, calculate the initial phase information of the baseband signal

Its calculation formula is:

In the step S3, the amplitude feature a(n) and the phase feature obtained above are mapped to a constellation map, and then converted into a two-dimensional feature constellation map, and the constellation map calculation formula is:

where j is an imaginary unit.

As shown in Figure 2(a)-Figure 2(f), the constellation diagrams of six types of modulated signals are respectively. After the modulation characteristic parameters of the signal are mapped to the constellation diagram, the horizontal X axis of the constellation diagram represents the in-phase component of the signal, and the vertical X axis represents the in-phase component of the signal. The Y-axis represents the quadrature component of the signal; the projection of the point on the X-axis is the maximum amplitude in the in-phase component, the projection of the point on the Y-axis is the maximum amplitude in the quadrature component, and the length of the line connecting the point to the origin is the The peak amplitude of the signal, the angle between the line and the X-axis is the phase of the signal.

In the step S4: constructing a modulation recognition framework of the convolutional neural network model, classifying the modulation type using the constructed convolutional neural network model for the data set composed of the characteristic constellation diagram of the signal, and demodulating the electromagnetic signal according to the modulation type.

As shown in FIG. 3 , the convolutional neural network model constructed in this embodiment includes three convolutional layers;

The feature map output by each convolutional layer is activated by ReLu and then subjected to down-sampling maximum pooling;

The features obtained from the previous convolutional layers are combined through 3 fully connected layers, and the dropout layer is used to reduce the risk of overfitting;

The final output is classified by the softmax function for modulation type.

The network model imitates the VGG network, by repeatedly stacking 3*3 small convolution kernels and 2*2 maximum pooling layers, deepening the network to improve performance, making the model's generalization ability strong, on different image datasets All performed well.

For the training of this model, this embodiment uses Keras, an advanced neural network application program interface under the tensorflow framework, which supports rapid experiments and can quickly convert ideas into results; using Keras to build a CNN framework, when compiling the model, the loss function adopts category crossover Entropy function categorical_crossentropy, which is used for multivariate classification problems; when calculating backpropagation, the optimizer uses adaptive momentum estimation Adam, which is not easy to fall into local advantages, faster, and more effective in learning effect;

During the training process, the loss change process of the training set is shown in Figure 4. After iterating the data of the training set and the validation set for 32 rounds, as shown in Figure 5, the accuracy of the training set is stable at more than 97%, and the The accuracy rate is stable at 97.6%.

In this embodiment, when the convolutional neural network CNN is used for signal modulation classification, there is no need for any precise mathematical expression between the input signal and the output category. As long as the convolutional network is trained with a known pattern, the network will It has the ability to map between input and output pairs, and it automatically obtains the ability to map input to output during the learning process of training data; secondly, the convolutional layer of CNN automatically extracts features through convolution operations, avoiding explicit feature extraction. , implicitly learns from the training data; compared with other deep learning models, the local connection and weight sharing characteristics of CNN reduce the complexity of the network and reduce a large number of training parameters.

In this embodiment, the signal generator SMW200A is used to generate PSK and QAM signals of different modulation orders, including: BPSK, QPSK, 8PSK, 16QAM, 32QAM, 64QAM, etc. The experimental verification depends on the parallel operation of the computer GPU, and the computer software Program MATLAB and Python;

The image data set used in this example has a total of 6480 images, including 1080 constellation maps for BPSK, QPSK, 8PSK, 16QAM, 32QAM, and 64QAM; the size of the image input to the neural network is 128×128 pixels; The IQ sample points are 5000, and the size of each image is 128 × 128 pixels; the images of the training set, validation set, and test set are divided in a ratio of 6:2:2 and saved in a file in hdf5 format; the trained model is the same It is saved in the h5 file. In order to test the usability of the model, the test set is predicted. The confusion matrix of the prediction result is shown in Figure 6. The overall accuracy rate is 97.38%.

Example 2

This embodiment provides a signal modulation type classification system based on a convolutional neural network, including:

a feature extraction module, configured to extract modulation features from the received electromagnetic signal, where the modulation features include amplitude features and first phase features;

a carrier processing module, used for filtering the carrier frequency of the first phase characteristic by adopting a carrier estimation algorithm to obtain the second phase characteristic;

The constellation map mapping module is used to perform constellation map mapping on the amplitude feature and the second phase feature to obtain a feature image;

The classification module is used for classifying the characteristic images by using a pre-built convolutional neural network model, obtaining the modulation type to which the characteristic images belong, and demodulating the electromagnetic signal according to the modulation type.

It should be noted here that the foregoing modules correspond to steps S1 to S4 in Embodiment 1, and the examples and application scenarios implemented by the foregoing modules and corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1 above. It should be noted that the above modules may be executed in a computer system such as a set of computer-executable instructions as part of the system.

In further embodiments, there is also provided:

An electronic device includes a memory, a processor, and computer instructions stored on the memory and executed on the processor, and when the computer instructions are executed by the processor, the method described in Embodiment 1 is completed. For brevity, details are not repeated here.

It should be understood that, in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate array FPGA or other programmable logic devices , discrete gate or transistor logic devices, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory may include read-only memory and random access memory and provide instructions and data to the processor, and a portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information.

A computer-readable storage medium for storing computer instructions, when the computer instructions are executed by a processor, the method described in Embodiment 1 is completed.

The method in Embodiment 1 may be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor. The software modules may be located in random access memory, flash memory, read-only memory, programmable read-only memory or electrically erasable programmable memory, registers and other storage media mature in the art. The storage medium is located in the memory, and the processor reads the information in the memory, and completes the steps of the above method in combination with its hardware. To avoid repetition, detailed description is omitted here.

Those of ordinary skill in the art can realize that the unit, that is, the algorithm step of each example described in conjunction with this embodiment, can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Skilled artisans may implement the described functionality using different methods for each particular application, but such implementations should not be considered beyond the scope of this application.

The above are only preferred embodiments of the present invention, and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Although the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, they do not limit the scope of protection of the present invention. Those skilled in the art should understand that on the basis of the technical solutions of the present invention, those skilled in the art do not need to pay creative work. Various modifications or deformations that can be made are still within the protection scope of the present invention.

Claims (10)

1. A signal modulation type classification method based on a convolutional neural network is characterized by comprising the following steps:

extracting modulation characteristics from a received electromagnetic signal, the modulation characteristics including an amplitude characteristic and a first phase characteristic;

filtering the carrier frequency of the first phase characteristic by adopting a carrier estimation algorithm to obtain a second phase characteristic;

carrying out constellation map mapping on the amplitude characteristic and the second phase characteristic to obtain a characteristic image;

and classifying the characteristic images by adopting a pre-constructed convolutional neural network model to obtain modulation types to which the characteristic images belong, and demodulating the electromagnetic signals according to the modulation types.

2. The convolutional neural network-based signal modulation type classification method as claimed in claim 1, wherein the amplitude characteristic is:

the first phase characteristic is:

wherein f (n) is an instantaneous frequency,

is the initial phase; f. of_sFor the sampling frequency, the sampling time t is N/f_s(ii) a I (n), Q (n) are the signal characteristics of the I path and the Q path of the electromagnetic signal respectively.

3. The convolutional neural network-based signal modulation type classification method as claimed in claim 1, wherein the carrier estimation algorithm is:

performing fast Fourier transform on the electromagnetic signal to obtain a maximum value M of a frequency spectrum₁And its subscript index value x₀；

According to x₂＝x₀Plus or minus 1 to obtain a frequency spectrum sub-maximum value M₂And its subscript index value x₂；

Frequency value x at spectral maximum in fast Fourier transform₀Interpolation is performed according to x₀、x₂Calculating a frequency estimation value;

and subtracting the frequency spectrum estimation value according to the first phase characteristic to obtain a second phase characteristic.

4. The convolutional neural network-based signal modulation type classification method of claim 1, wherein the constellation map is mapped as:

wherein a (n) is an amplitude characteristic,

is the initial phase.

5. The signal modulation type classification method based on the convolutional neural network as claimed in claim 1, wherein the convolutional neural network model comprises 3 convolutional layers, and a feature map output by each convolutional layer is subjected to maximum pooling processing of downsampling after ReLu activation;

combining the characteristics obtained by the convolution layer by using 3 full-connection layers, and reducing overfitting by using a dropout layer;

the output layer performs the classification of the modulation type by means of a softmax function.

6. The signal modulation type classification method based on the convolutional neural network as claimed in claim 1, wherein in the training of the convolutional neural network model, the loss function adopts a class cross entropy function, and in the calculation of the backward propagation, an adaptive momentum estimation optimizer is adopted.

7. The method of claim 6, wherein training the convolutional neural network model to generate PSK and QAM signals of different modulation orders using a signal generator, comprises: BPSK, QPSK, 8PSK, 16QAM, 32QAM, 64 QAM.

8. A convolutional neural network-based signal modulation type classification system, comprising:

the characteristic extraction module is used for extracting modulation characteristics from the received electromagnetic signals, wherein the modulation characteristics comprise amplitude characteristics and first phase characteristics;

the carrier processing module is used for filtering the carrier frequency of the first phase characteristic by adopting a carrier estimation algorithm to obtain a second phase characteristic;

the constellation map mapping module is used for carrying out constellation map mapping on the amplitude characteristic and the second phase characteristic to obtain a characteristic image;

and the classification module is used for classifying the characteristic images by adopting a pre-constructed convolutional neural network model to obtain modulation types to which the characteristic images belong, and demodulating the electromagnetic signals according to the modulation types.

9. An electronic device comprising a memory and a processor and computer instructions stored on the memory and executed on the processor, the computer instructions when executed by the processor performing the method of any of claims 1-7.

10. A computer-readable storage medium storing computer instructions which, when executed by a processor, perform the method of any one of claims 1 to 7.

Images