Automatic modulation identification function implementation method based on TCP protocol

Abstract

The invention provides an automatic modulation recognition function implementation method based on a TCP (Transmission Control Protocol), belongs to the technical field of deep learning, and solves the problems that various factors of an intelligent receiver system are still missed by simulation data passing through channels and loading noise in the actual communication and signal acquisition process of an automatic modulation recognition algorithm, and the accuracy of the system is poor. The method comprises the following steps: carrying out iterative training on a modulation identification algorithm model through a modulation data set; judging the accuracy of the trained modulation recognition algorithm model; extracting the original data of the modulation data set and putting the original data back to the trained modulation recognition algorithm model for verification; establishing a vector signal generator, after verification of the modulation recognition algorithm model is completed, coupling deployment of the modulation algorithm model with a TCP receiver in the vector signal generator, and outputting real modulation signal data conforming to the deep neural network input format; and inputting real modulation signal data into the trained modulation identification algorithm model for signal identification.

Description

Technical Field

[0001] The present invention relates to a method for implementing an automatic modulation recognition function based on the TCP protocol, belonging to the technical field of deep learning. Background Art

[0002] With the wide application of deep learning methods and the optimization and improvement of deep neural networks, the types and attributes of signals in the communication modulation signal dataset are increasing and the details are being improved, and it is becoming more widely adaptable and inclusive of various modulation methods in actual application channel scenarios, which has led to the rapid development and progress of the research on adaptive modulation recognition. With the continuous improvement and optimization of the application of deep learning algorithms in adaptive modulation recognition technology, it has become easier to extract various signal features of communication modulation signal data. Different deep neural networks and optimization schemes have enriched the modulation recognition algorithms, and the recognition performance results have been efficiently improved. Most of the above-mentioned background automatic modulation recognition algorithms are deployed on a single computer device, mainly using application programs or simulation software to simulate the method of generating signal data to obtain modulation signal data for network model verification. Almost no actual communication and signal acquisition processes are involved. There are still omissions in various factors of the intelligent receiver system for the simulation data of passing through the channel and adding noise. Therefore, there may be a certain deviation in the verification of the network. Summary of the Invention

[0003] In order to solve the problem that the automatic modulation recognition algorithm goes through the process of actual communication and signal acquisition, and there are still omissions in various factors of the intelligent receiver system for the simulation data of passing through the channel and adding noise, resulting in a certain deviation in the verification of the network, the present invention further provides a method for implementing an automatic modulation recognition function based on the TCP protocol.

[0004] The technical solution adopted by the present invention to solve the above problems is as follows: The present invention includes the following steps:

[0005] Step 1: Iteratively train the modulation recognition algorithm model through a modulation dataset;

[0006] Step 2: Judge the accuracy rate of the trained modulation recognition algorithm model. If the accuracy rate is not less than 90%, go to Step 3. If the accuracy rate is less than 90%, repeat Step 1 until the accuracy rate is not less than 90%;

[0007] Step 3: Extract the original data of the modulation dataset and put it back into the trained modulation recognition algorithm model for verification;

[0008] Step 4: Establish a vector signal generator. After the modulation recognition algorithm model is verified, couple the deployment of the modulation algorithm model with the TCP receiver module in the vector signal generator to output real modulation signal data that conforms to the input format of the deep neural network;

[0009] Step 5: Input the real modulation signal data into the trained modulation recognition algorithm model for signal recognition.

[0010] Further, Step 1 specifically includes:

[0011] Step 1.1: Configure the training environment and set the deep learning architecture;

[0012] Step 1.2: Select the modulation dataset and preprocess the modulation dataset;

[0013] Step 1.3: Set the optimization parameters during the training process;

[0014] Step 1.4: Set the network structure of the modulation recognition algorithm model. A two-layer LSTM network structure can be adopted, with 64 recurrent neurons in each layer. Input the preprocessed modulation dataset into the modulation recognition algorithm model for iterative training.

[0015] Further, Step 2 specifically includes:

[0016] After the training of the current round is completed, draw a performance result graph and determine whether the accuracy rate in the performance result graph is not lower than 90%. If the accuracy rate is not lower than 90%, proceed to Step 3. If it is lower than 90%, improve the network structure of the modulation recognition algorithm model, and use the new optimization parameters for repeated training after adjusting the network structure until the accuracy rate is higher than 90%. Among them, during the improvement process of the network structure of the modulation recognition algorithm model, if the training effect is insufficient, the number of LSTM network layers and the number of neurons in one layer can be increased; if overfitting occurs during training, the number of neurons can be reduced.

[0017] Further, Step 3 specifically includes:

[0018] Step 3.1: Traverse the modulation dataset, randomly extract the labeled data in the modulation dataset, and reshape the extracted labeled data into an integer array, where the format of the integer array is the input format of the modulation recognition algorithm model;

[0019] Step 3.2: Input the labeled data after reshaping into the trained modulation recognition algorithm model and determine whether the input data is correctly recognized;

[0020] Step 3.3: Repeat Steps 3.1 - 3.2, perform multiple extractions and replays of the modulation dataset to complete the verification of the modulation recognition algorithm model.

[0021] Further, in step 4, the vector signal generating instrument includes a TCP transmitter module, a TCP receiver module, and a modulation signal generating module. Among them, the TCP transmitter module and the TCP receiver module build a TCP communication protocol transmission structure, and the IP address and port number are set to be the same;

[0022] The TCP transmitter module generates modulation signal data with 1024 sampling points according to the BPSK signal after ASCII encoding;

[0023] The TCP receiver module is used to decode and process the modulation signal data to obtain the real modulation signal data that conforms to the input format of the deep neural network;

[0024] The modulation signal generating module is used to output the real modulation signal data that conforms to the input format of the deep neural network.

[0025] Further, the TCP transmitter module includes a TCP data sending establishment sub-module, a TCP data node reading sub-module, a TCP node opening sub-module, and a TCP listener sub-module;

[0026] The TCP data sending establishment sub-module is responsible for sending the input data through the hardware device using the TCP protocol; the TCP data node reading sub-module is connected to the TCP data sending establishment sub-module and is responsible for reading the data to be sent into the TCP data sending establishment sub-module; the TCP node opening sub-module is responsible for opening the TCP protocol sending node; the TCP listener sub-module is responsible for detecting the port number and whether the data is transmitted correctly.

[0027] Further, the output of the real modulation signal data that conforms to the input format of the deep neural network in step 4 includes:

[0028] Step 4.1: Set the carrier center frequency, signal symbol rate, and the number of sampling points per symbol of the vector signal generating instrument, generate a 64-bit BPSK signal, and set the amplitude of the generated BPSK signal to +1 and -1;

[0029] Step 4.2: Perform ASCII encoding on the generated BPSK signal;

[0030] Step 4.3: Input the BPSK signal after ASCII encoding into the TCP transmitter module, set the IP address and port number of the TCP transmitter module, and generate modulation signal data with 1024 sampling points. Among them, both the real part and the imaginary part of the modulation signal data are represented by 1024 floating-point numbers;

[0031] Step 4.4: Use the TCP protocol to transmit the modulation signal data to python on the computer software side for array processing and modulation recognition to generate ASCII codes;

[0032] Step 4.5: The TCP receiver receives the ASCII code and decodes it into an array of floating-point types using the IEEE-754 standard, and compares whether the received data is the same as the transmitted data. If they are the same, go to Step 4.6. If they are not the same, repeat Steps 4.2 - 4.4 until the received data and the transmitted data are the same;

[0033] Step 4.6: Separate the real and imaginary parts of the data in the decoded array, respectively form arrays with a length of 1024, and vertically stack them together to obtain the real modulation signal data that conforms to the input format of the deep neural network and output it through the modulation signal generation module.

[0034] The beneficial effects of the present invention are:

[0035] (1) The method for realizing the function of the automatic modulation recognition algorithm proposed by the present invention has a simple idea and an easy-to-understand implementation method, realizes the complete information transfer process in the transmitter and receiver systems, generates real modulation signal data through the hardware device vector signal generator, and communicates with another computer in real time through the TCP communication protocol in the computer system that generates the data. The real communication modulation signal data received through the TCP communication protocol is used as the input data for system recognition, and it is input into the deep neural network training parameter model deployed in the automatic modulation recognition system in the format of IQ two-way signals for recognition, and a good recognition rate is obtained according to the performance of the network parameter model.

[0036] (2) The method for realizing the automatic modulation recognition function based on the TCP protocol proposed by the present invention has excellent performance, and the TCP communication process is stable and reliable, and the data recognition effect is accurate.

[0037] (3) Through the present invention, the parameter recognition model obtained by deep learning training can be actually and effectively verified, and a simple small signal transmission and reception recognition system can be realized through the method of TCP protocol communication, which improves the application process of adaptive modulation recognition. Description of the Drawings

[0038] Figure 1 It is a schematic flow chart of a method for realizing the automatic modulation recognition function based on the TCP protocol provided by the present invention;

[0039] Figure 2 It is a schematic diagram of the test result of the data set put-back provided by the present invention;

[0040] Figure 3 It is a schematic diagram of the language logic interface for generating modulation signal data provided by the present invention.

[0041] Figure 4Schematic diagram of the recognition result of the real modulation signal data received by the present invention. Detailed implementation mode

[0042] Combined with Figures 1 - 3 To illustrate this implementation mode, as Figure 1 shown, the steps of a method for realizing the automatic modulation recognition function based on the TCP protocol described in this implementation mode include:

[0043] S1: Train a modulation recognition algorithm model based on deep learning;

[0044] S101: Configure the training environment and set the deep learning architecture;

[0045] S102: Select a data set and preprocess the data set;

[0046] S103: Set the optimization parameters during training;

[0047] S104: Set the deep neural network structure. A two-layer LSTM network structure can be adopted, with 64 recurrent neurons in each layer. Input the preprocessed modulation data set into the modulation recognition algorithm model for iterative training;

[0048] S105: After training, draw a performance result graph, analyze the performance advantages and disadvantages, and repeat the training if the network and parameters need to be improved. Stop after achieving excellent performance. Among them, during the process of improving the network structure of the modulation recognition algorithm model, if the training effect is not sufficient, the number of LSTM network layers and the number of neurons in one layer can be increased; if overfitting occurs during training, the number of neurons can be reduced;

[0049] S106: Put the internal data of the data set back into the network parameter model for testing to further check the performance of the network model.

[0050] The first core of this implementation mode is the overall process of deep neural network algorithm simulation and the integrity and performance of the modulation recognition algorithm model. Finally, a modulation recognition algorithm model with a recognition accuracy of more than 95% in a high signal-to-noise ratio environment is obtained, which can identify and classify 11 types of modulation signals, and the overall recognition rate reaches about 85%. After the data set extraction data playback test, the effectiveness and stability of the modulation recognition algorithm model can be verified. Therefore, the modulation recognition algorithm model can be applied to actual communication in the next core content of this implementation mode.

[0051] S2: Judge whether the recognition rate of the model reaches more than 90%. If so, proceed to S3; if not, repeat S1 until the recognition rate of the model reaches more than 90%;

[0052] S3: Extract the original data of the modulation data set and put it back into the trained modulation recognition algorithm model for verification;

[0053] S301: Traverse the modulation data set, randomly extract the labeled data in the modulation data set, and reshape the extracted labeled data into an integer array, where the format of the integer array is the input format of the modulation recognition algorithm model;

[0054] S302: Input the labeled data after the integer array into the trained modulation recognition algorithm model to determine whether the input data is correctly recognized;

[0055] S303: Repeat S301 - S302 to perform multiple extractions and replays of the modulation data set, complete the verification of the modulation recognition algorithm model, and summarize the recognition accuracy and stability of the modulation recognition algorithm model obtained by simulating the automatic modulation recognition algorithm. As Figure 2 shown, Figure 2 For the test result of putting the data set back, the modulation method of 4PAM was selected for testing, and the result shows correct recognition.

[0056] S4: Build a TCP communication protocol transmission architecture on two computer platforms, namely the transmitter and the receiver, in the vector signal generator instrument;

[0057] The second core of this embodiment is to use a transmitter system with a vector signal generator module to generate real communication modulation signals. In the transmitter system, the generated signals are converted into IQ data format and sent to another receiver computer system using the TCP communication protocol. The real modulation signal data received through the TCP communication protocol is input into the adaptive modulation recognition algorithm parameter model for signal recognition and the results are displayed. The specific steps are as follows:

[0058] S401: First, set the carrier center frequency to 1 GHz, the signal symbol rate to 100 KHz, and 8 sampling points per symbol to generate a 64 - bit BPSK signal, and set the amplitude of the BPSK signal to +1 and -1;

[0059] S402: Next, perform ASCII encoding on the generated BPSK signal for easy transmission using the TCP communication protocol;

[0060] S403: Subsequently, create a TCP transmitter module in the data sending link, and add a TCP data sending establishment sub - module, a TCP data reading node sub - module, a TCP node opening module, and a TCP listener sub - module. Set the IP address of this machine and set the port number to 63400. As Figure 3 shown, Figure 3It is a language logic interface for generating modulated signal data. Among them, the TCP data sending establishment sub-module is responsible for sending the input data through a hardware device using the TCP protocol; the TCP data reading node sub-module is connected to the TCP data sending establishment sub-module and is responsible for reading the data to be sent into the TCP data sending establishment sub-module; the TCP node opening sub-module is responsible for opening the TCP protocol sending node; the TCP listener sub-module is responsible for detecting the port number and whether the data is correctly transmitted;

[0061] S404: Modulated signal data with 1024 sampling points can be generated. The real part and the imaginary part are each represented by 1024 floating-point numbers, and the total data length is 16384 bits. Next, it is transmitted to Python on the computer software side through the TCP protocol for array processing and modulation recognition, generating ASCII codes and sending them to the TCP receiver module;

[0062] S405: The TCP receiver module uses the same IP address and port number settings as the transmitter end, decodes the received ASCII codes into an array of floating-point types using the IEEE-754 standard, and compares whether the received data is the same as the transmitted data.

[0063] S406: After ensuring the data is correct, separate the real part and the imaginary part of the data, respectively form arrays with a length of 1024, and stack them vertically together to form modulated signal data that conforms to the input format of the deep neural network and output it through the modulated signal generation module.

[0064] The idea of this embodiment is concise and the implementation method is simple and easy to understand. It realizes the complete information transfer process in the transmitter and receiver systems, generates real modulated signal data through a hardware device vector signal generator, and conducts real-time communication with another computer in the computer system that generates the data through the TCP communication protocol. The real communication modulated signal data received through the TCP communication protocol is used as the input data for system recognition, and it is input into the deep neural network training parameter model deployed in the automatic modulation recognition system in the format of IQ two-channel signals for recognition, and a good recognition rate is obtained according to the performance of the network parameter model.

[0065] S5: Judge whether the vector signal generator transmits correctly. If so, proceed to S6; if not, adjust the structure of the vector signal generator;

[0066] S6: Input the real modulated signal data output by the vector signal generator into the trained modulation recognition algorithm model for signal recognition.

[0067] Finally, load the model and weights to perform classification detection, and the detection results are as Figure 4 shown. According to Figure 4The obtained results indicate that a method for realizing the automatic modulation recognition function based on the TCP protocol proposed by the present invention can be correctly implemented, and has a high recognition accuracy for signals, and has wide application value.

[0068] The above are only preferred embodiments of the present invention, and do not impose any form of limitation on the present invention. Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments by using the disclosed technical content within the scope of the technical solution of the present invention. However, as long as it does not depart from the content of the technical solution of the present invention, and based on the technical essence of the present invention, any simple modification, equivalent replacement, and improvement made to the above embodiments within the spirit and principles of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A method for implementing the automatic modulation recognition function based on the TCP protocol, characterized in that, The steps of the method for realizing the automatic modulation recognition function based on the TCP protocol include: Step 1: Iteratively train the modulation recognition algorithm model through the modulation data set; Step 2: Judge the accuracy rate of the trained modulation recognition algorithm model. If the accuracy rate is not lower than 90%, go to Step 3. If the accuracy rate is lower than 90%, repeat Step 1 until the accuracy rate is not lower than 90%; Step 3: Extract the original data of the modulation data set and put it back into the trained modulation recognition algorithm model for verification; Step 4: Establish a vector signal generator. After the verification of the modulation recognition algorithm model is completed, couple the deployment of the modulation algorithm model with the TCP receiver module in the vector signal generator to output real modulation signal data that conforms to the input format of the deep neural network; Step 5: Input the real modulation signal data into the trained modulation recognition algorithm model for signal recognition.

2. The method for implementing an automatic modulation recognition function based on the TCP protocol according to claim 1, wherein Step 1 specifically includes: Step 1.1: Configure the training environment and set the deep learning architecture; Step 1.2: Select the modulation data set and preprocess the modulation data set; Step 1.3: Set the optimization parameters during the training process; Step 1.4: Set the network structure of the modulation recognition algorithm model. A two-layer LSTM network structure can be adopted, with 64 recurrent neurons in each layer. Input the preprocessed modulation data set into the modulation recognition algorithm model for iterative training.

3. The method for implementing the automatic modulation recognition function based on the TCP protocol according to claim 1, characterized in that, Step 2 specifically includes: After the training of the current round is completed, draw a performance result graph and judge whether the accuracy rate in the performance result graph is not lower than 90%. If the accuracy rate is not lower than 90%, go to Step 3. If it is lower than 90%, improve the network structure of the modulation recognition algorithm model, and use the new optimization parameters for repeated training after adjusting the network structure until the accuracy rate is higher than 90%. Among them, during the improvement of the network structure of the modulation recognition algorithm model, if the training effect is not sufficient, the number of LSTM network layers and the number of neurons in one layer can be increased; if overfitting occurs during training, the number of neurons can be reduced.

4. The method for implementing the automatic modulation recognition function based on the TCP protocol according to claim 1, characterized in that, Step 3 specifically includes: Step 3.1: Traverse the modulation data set, randomly extract the data with labels in the modulation data set, and reshape the extracted data with labels into an integer array, where the format of the integer array is the input format of the modulation recognition algorithm model; Step 3.2: Input the data with labels after reshaping into the trained modulation recognition algorithm model and judge whether the input data is correctly recognized; Step 3.3: Repeat Steps 3.1 - 3.2, perform multiple extractions and replays of the modulation data set to complete the verification of the modulation recognition algorithm model.

5. The method for implementing the automatic modulation recognition function based on the TCP protocol according to claim 1, characterized in that In Step 4, the vector signal generator includes a TCP transmitter module, a TCP receiver module, and a modulation signal generation module. Among them, the TCP transmitter module and the TCP receiver module build a TCP communication protocol transmission structure, and the IP address and port number are set the same; The TCP transmitter module generates modulation signal data with 1024 sampling points according to the ASCII-encoded BPSK signal; The TCP receiver module is used to decode and process the modulated signal data to obtain the real modulated signal data that conforms to the input format of the deep neural network; The modulated signal generation module is used to output the real modulated signal data that conforms to the input format of the deep neural network.

6. The method for implementing the automatic modulation recognition function based on the TCP protocol according to claim 5, characterized in that, The TCP transmitter module includes a TCP data sending establishment sub-module, a TCP data node reading sub-module, a TCP node opening sub-module, and a TCP listener sub-module; The TCP data sending establishment sub-module is responsible for sending the input data through a hardware device using the TCP protocol; the TCP data node reading sub-module is connected to the TCP data sending establishment sub-module and is responsible for reading the data to be sent into the TCP data sending establishment sub-module; The TCP node opening sub-module is responsible for opening the TCP protocol sending node; the TCP listener sub-module is responsible for detecting the port number and whether the data is correctly transmitted.

7. A method for implementing an automatic modulation recognition function based on the TCP protocol according to claim 1, characterized in that, The real modulated signal data output in step 4 that conforms to the input format of the deep neural network includes: Step 4.1: Set the carrier center frequency, signal symbol rate, and the number of sampling points per symbol of the vector signal generator, generate a 64-bit BPSK signal, and set the amplitude of the generated BPSK signal to +1 and -1; Step 4.2: Perform ASCII encoding on the generated BPSK signal; Step 4.3: Input the ASCII-encoded BPSK signal into the TCP transmitter module, set the IP address and port number of the TCP transmitter module, and generate modulated signal data with 1024 sampling points. Among them, both the real part and the imaginary part of the modulated signal data are represented by 1024 floating-point numbers; Step 4.4: Use the TCP protocol to transmit the modulated signal data to python on the computer software side for array processing and modulation recognition, and generate ASCII codes; Step 4.5: The TCP receiver receives the ASCII code and decodes it into an array of floating-point types using the IEEE-754 standard, and compares whether the received data is the same as the transmitted data. If they are the same, proceed to step 4.

6. If they are not the same, repeat steps 4.2 - 4.4 until the received data and the transmitted data are the same; Step 4.6: Separate the real part and the imaginary part of the data in the decoded array, respectively form arrays with a length of 1024, and vertically stack them together to obtain the real modulated signal data that conforms to the input format of the deep neural network and output it through the modulated signal generation module.

Images