A method and system for identifying the modulation mode of a radiation source signal

Abstract

The present invention belongs to the technical field of unknown radiation source signal analysis, and its purpose is to provide a method and system for identifying the modulation mode of radiation source signals. The present invention proposes a method for identifying the modulation mode of unknown radiation source using the characteristic parameters of the signal itself in conjunction with a global optimization neural network. During the implementation process, the characteristic parameters of the radiation source signal are first obtained, that is, the shallow characteristics of the radiation source signal itself are estimated, and then the neural network is constructed and the optimal network structure is obtained. Based on the superior performance of the deep neural network in nonlinear approximation, the characteristic parameters of the radiation source signal are input into the optimal network structure after multiple layers of individual coding changes, so as to mine and capture the potential characteristics of the radiation source signal, thereby making the modulation mode output result more stable and efficient. Compared with the recognition of ordinary neural networks, the noise resistance and recognition accuracy of the modulation mode recognition of the radiation source signal in a low signal-to-noise ratio environment of the present invention are significantly improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of unknown radiation source signal analysis, and particularly relates to a method and system for identifying the modulation mode of a radiation source signal. Background Art

[0002] Automatic modulation mode recognition is a technology between signal detection and signal demodulation. The modulation mode is an important feature for distinguishing different types of communication signals. The recognition of the modulation mode requires correctly judging the signal modulation mode without knowing the modulation information content of the unknown radiation source, laying a foundation for subsequent demodulation and decoding, so as to obtain useful information content.

[0003] In the prior art, the research on modulation recognition methods is generally divided into the maximum likelihood method and the pattern recognition method. Among them, the maximum likelihood method uses probability theory and hypothesis testing theory to make predictions by maximizing the likelihood function, analyzes the statistical characteristics of the signal and derives the test statistic, and realizes the automatic recognition of the modulation mode by the decision criterion. The pattern recognition method extracts the characteristic parameters containing the modulation mode information from the modulation signal through feature extraction, and then realizes the automatic recognition of the modulation mode through pattern matching. However, in the process of using the prior art, the inventor found that there are at least the following problems in the prior art:

[0004] In the process of using the prior art for recognition, the noise in the environment will greatly affect the construction of the likelihood function and the calculation of the characteristic parameters, and thus when the above two methods are applied to a low signal-to-noise ratio environment, the recognition accuracy is reduced. Summary of the Invention

[0005] The present invention aims to solve the above technical problems at least to a certain extent, and provides a method and system for identifying the modulation mode of a radiation source signal.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a method for identifying the modulation mode of a radiation source signal, including:

[0008] Receiving a radiation source signal, and extracting instantaneous parameter information from the radiation source signal;

[0009] Calculating the characteristic parameters of the radiation source signal from the instantaneous parameter information;

[0010] Simulating the analog signals of each modulation mode and calculating the characteristic parameters corresponding to the analog signals of each modulation mode, and generating a labeled data set according to the analog signals of each modulation mode and their corresponding characteristic parameters;

[0011] Constructing an initial deep neural network, and encoding all weights and thresholds in the initial deep neural network to obtain multiple individual codes;

[0012] Performing code point crossover and code point transformation operations on the designated individual code, and then training the initial deep neural network using the labeled data set to obtain a trained neural network;

[0013] Calculating the partial variance of the trained neural network, and inferring the structural approximation of the trained neural network based on the partial variance;

[0014] Obtaining an optimal network structure according to the structural approximation and a preset approximation threshold;

[0015] The characteristic parameters of the radiation source signal are input into the optimal network structure as the characteristic parameters of the signal to be identified, so as to obtain the maximum probability result of the modulation mode of the radiation source signal.

[0016] The present invention can improve the accuracy and stability of the modulation mode identification of the radiation source signal in a low signal-to-noise ratio environment. Specifically, the present invention proposes a method for identifying the modulation mode of an unknown radiation source by using the characteristic parameters of the signal itself in conjunction with a global optimization neural network. During the implementation process, the characteristic parameters of the radiation source signal are first obtained, that is, the shallow characteristics of the radiation source signal itself are estimated, and then the neural network is constructed and the optimal network structure is obtained. Based on the superior performance of the deep neural network in nonlinear approximation, the characteristic parameters of the radiation source signal are input into the optimal network structure after multiple layers of individual coding changes, so as to explore and capture the potential characteristics of the radiation source signal, thereby making the modulation mode output result more stable and efficient. Compared with the recognition of ordinary neural networks, the noise resistance and recognition accuracy of the modulation mode recognition of the radiation source signal in a low signal-to-noise ratio environment of the present invention are significantly improved.

[0017] In a possible design, the instantaneous parameter information extracted from the radiation source signal includes:

[0018]

[0019] In formula 1, N s is the number of sampling points of the radiation source signal, a(i) is the instantaneous amplitude of the radiation source signal at the i-th sampling point, m a is the instantaneous amplitude mean value of the radiation source signal; a n (i) is the normalized instantaneous amplitude of the radiation source signal at the i-th sampling point, a cn (i) is the zero-centered normalized instantaneous amplitude of the radiation source signal at the i-th sampling point; is the normalized instantaneous amplitude mean of the radiation source signal; f(i) is the instantaneous frequency of the radiation source signal at the i-th sampling point, is the normalized instantaneous frequency of the radiation source signal at the i-th sampling point; R s is the symbol rate of the radiation source signal; is the instantaneous phase of the radiation source signal at the i-th sampling point, f s is the signal frequency, f c is the carrier frequency, is the non-linear phase component of the radiation source signal at the i-th sampling point.

[0020] In a possible design, the characteristic parameters of the radiation source signal deduced from the instantaneous parameter information include:

[0021]

[0022] In Equation 2, Z1 is the maximum value of the amplitude spectral density, Z2 is the instantaneous amplitude variance, Z3 is the instantaneous frequency variance, Z4 is the instantaneous phase variance, is the Fourier operator, c is the number of non-weak signal values among the N s sampling points of the radiation source signal.

[0023] In a possible design, the analog signals of each modulation method include multi-amplitude shift keying modulation signals, multi-frequency shift keying modulation signals, multi-phase shift keying modulation signals, amplitude modulation signals, linear frequency modulation signals, and continuous wave modulation signals; among them,

[0024] The multi-amplitude shift keying modulation signal is as follows:

[0025]

[0026] In Equation 3, e MASK (t) is the multi-amplitude shift keying signal obtained by simulation and varying with time t, a n is the level of the n-th symbol, g() is the window function, T s is the time interval of g(), ω c is the angular frequency of the carrier signal;

[0027] The multi-frequency shift keying modulation signal is as follows:

[0028]

[0029]

[0030] In Equation 4, e MFSK (t) is the multi-level frequency shift keying signal obtained by simulation and varying with time t, A is the signal amplitude, ω iis the angular frequency of the i-th carrier signal;

[0031] The multi-phase shift keying modulation signal is as follows:

[0032]

[0033] In Equation 5, e MPSK (t) is the multi-level phase shift keying signal obtained by simulation and varying with time t, and φ n is the initial phase of the n-th carrier signal;

[0034] The amplitude modulation signal is as follows:

[0035] e AM (t) = a n (1 + m c v Ω (t)) cos ω c t; Equation 6

[0036] In Equation 6, e AM (t) is the amplitude modulation signal obtained by simulation and varying with time t, and m c is the modulation index, ranging from [0, 1], and ν Ω (t) is the modulation signal;

[0037] The chirp signal is as follows:

[0038]

[0039] In Equation 7, e LFM (t) is the chirp signal obtained by simulation and varying with time t, rect is the rectangular function transformation symbol, K is the chirp rate, exp() is the exponential function with the natural constant e as the base, and j is the imaginary symbol;

[0040] The continuous wave modulation signal is as follows:

[0041] e CW (t) = a(t) cos[ω c t + θ(t)]; Equation 8

[0042] In Equation 8, e CW (t) is the continuous wave modulation signal obtained by simulation and varying with time t, a(t) is the amplitude function, and θ(t) is the phase function.

[0043] In a possible design, an initial deep neural network is constructed, and all weights and thresholds in the initial deep neural network are encoded to obtain a plurality of individual encodings, including:

[0044] Construct an initial deep neural network; wherein, the initial deep neural network adopts a three-layer neural network, and the initial deep neural network includes an input layer, a hidden layer, and an output layer connected in sequence;

[0045] Initialize all the weights and thresholds in the initial deep neural network; wherein, the weights in the initial deep neural network include the weights between the input layer and the hidden layer and the weights between the hidden layer and the output layer, and the thresholds in the initial deep neural network include the hidden layer threshold and the output layer threshold;

[0046] Encode all the weights and thresholds to obtain multiple individual encodings.

[0047] In a possible design, perform bit crossover and bit transformation operations on a specified individual encoding, and then use the labeled dataset to train the initial deep neural network to obtain a trained neural network, including:

[0048] Perform a random bit crossover operation on two adjacent individual encodings in the same layer of the initial deep neural network;

[0049] After performing the crossover operation on all the individual encodings, randomly select an individual encoding in each layer of the initial deep neural network for bit transformation operation;

[0050] Use the analog signals of each modulation method in the labeled dataset as the input data of the initial deep neural network, and use the characteristic parameters corresponding to the analog signals of each modulation method as the output data of the initial deep neural network to train the initial deep neural network.

[0051] In a possible design, when the j-th bit \(a\) of the k-th individual encoding kj and the j-th bit \(a\) of the l-th individual encoding lj perform a bit crossover operation, the following conditions are satisfied:

[0052] \(a\) kj =\(a\) kj (1 - b)+\(a\) lj b; Equation 9

[0053] In Equation 9, b is a random number in the interval [0, 1];

[0054] When the j-th bit \(a\) of the i-th individual encoding ij performs a bit transformation operation, the following conditions are satisfied:

[0055]

[0056] In Equation 10, \(a\) max and \(a\) min are respectively the encoding \(a\)ij The maximum and minimum values that each digit can take, g is the current iteration number, r is a random number in the interval [0, 1], r2 is a random number, g max is the preset maximum tolerable iteration number.

[0057] In a possible design, the partial variance of the trained neural network is:

[0058]

[0059] In Equation 11, m is the number of output samples of the trained neural network, each sample is an n-dimensional vector, the actual output is x j , and the expected output is

[0060] The structural approximation degree of the trained neural network is the reciprocal of (MSE f +1), and

[0061] In a possible design, according to the structural approximation degree and the preset approximation degree threshold, an optimal network structure is obtained, including:

[0062] Set an approximation degree threshold for the structural approximation degree;

[0063] When the structural approximation degree is greater than the approximation degree threshold, or the current iteration number exceeds the preset maximum iteration number, use the trained neural network as the optimal network structure;

[0064] Otherwise, re-perform bit crossover and bit transformation operations on the specified individual coding until an optimal network structure is obtained.

[0065] In a second aspect, the present invention provides a radiation source signal modulation mode recognition system for implementing the radiation source signal modulation mode recognition method as described in any one of the above; the radiation source signal modulation mode recognition system includes:

[0066] A radiation source signal information extraction module, configured to receive a radiation source signal, extract instantaneous parameter information from the radiation source signal; and further configured to deduce characteristic parameters of the radiation source signal through the instantaneous parameter information;

[0067] A neural network construction module, which is used to simulate analog signals of various modulation modes, calculate characteristic parameters corresponding to the analog signals of various modulation modes, and generate a labeled data set according to the analog signals of various modulation modes and their corresponding characteristic parameters; it is used to construct an initial deep neural network, encode all weights and thresholds in the initial deep neural network to obtain multiple individual encodings; it is used to perform bit crossover and bit transformation operations on a specified individual encoding, and then use the labeled data set to train the initial deep neural network to obtain a trained neural network; it is used to calculate the partial variance of the trained neural network, and infer the structure approximation degree of the trained neural network according to the partial variance; it is also used to obtain an optimal network structure according to the structure approximation degree and a preset approximation degree threshold;

[0068] A modulation mode prediction module, which is respectively communicatively connected to the radiation source signal information extraction module and the neural network construction module, and is used to input the characteristic parameters of the radiation source signal as the characteristic parameters of the signal to be recognized into the optimal network structure to obtain the maximum probability result of the modulation mode of the radiation source signal.

[0069] In a third aspect, the present invention provides an electronic device, including:

[0070] A memory, which is used to store computer program instructions; and,

[0071] A processor, which is used to execute the computer program instructions to complete the operations of the radiation source signal modulation mode recognition method as described in any one of the above.

[0072] In a fourth aspect, the present invention provides a computer-readable storage medium, which is used to store computer-readable computer program instructions, and the computer program instructions are configured to execute the operations of the radiation source signal modulation mode recognition method as described in any one of the above when running. Description of the Drawings

[0073] Figure 1 is a flowchart of a method for recognizing the modulation mode of a radiation source signal in Embodiment 1;

[0074] Figure 2 is a schematic structural diagram of the initial deep neural network in Embodiment 1;

[0075] Figure 3 is the recognition result of the optimal network structure under the condition of no noise in Embodiment 1;

[0076] Figure 4 is the recognition result under different signal-to-noise ratios before and after optimizing the neural network by using individual encoding in Embodiment 1;

[0077] Figure 5It is a block diagram of a module of a radiation source signal modulation method recognition system in Embodiment 2. Detailed implementation manners

[0078] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in combination with the accompanying drawings and the descriptions of the embodiments or the prior art. Obviously, the following descriptions of the structures of the accompanying drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other accompanying drawings can be obtained based on these accompanying drawings without creative efforts. It should be noted here that the descriptions of these embodiment manners are used to help understand the present invention, but do not constitute a limitation to the present invention.

[0079] Embodiment 1:

[0080] This embodiment discloses a method for recognizing the modulation mode of a radiation source signal, which can be, but is not limited to, executed by a computer device or virtual machine with certain computing resources, such as executed by an electronic device such as a personal computer, a smart phone, a personal digital assistant or a wearable device, or executed by a virtual machine.

[0081] As Figure 1 shown, a method for recognizing the modulation mode of a radiation source signal can, but is not limited to, include the following steps:

[0082] S1. Receive the radiation source signal and extract the instantaneous parameter information from the radiation source signal.

[0083] Specifically, in step S1, the instantaneous parameter information extracted from the radiation source signal includes:

[0084]

[0085] In Equation 1, N s is the number of sampling points of the radiation source signal, that is, there are a total of N s sampling points, a(i) is the instantaneous amplitude of the radiation source signal at the i-th sampling point, and m a is the mean value of the instantaneous amplitude of the radiation source signal; a n (i) is the normalized instantaneous amplitude of the radiation source signal at the i-th sampling point, and a cn (i) is the zero-centered normalized instantaneous amplitude of the radiation source signal at the i-th sampling point; is the mean value of the normalized instantaneous amplitude of the radiation source signal; f(i) is the instantaneous frequency of the radiation source signal at the i-th sampling point, is the normalized instantaneous frequency of the radiation source signal at the i-th sampling point; R s is the symbol rate of the radiation source signal; is the instantaneous phase of the radiation source signal at the i-th sampling point, f s is the signal frequency, f c is the carrier frequency is the non-linear phase component of the radiation source signal at the i-th sampling point.

[0086] S2. Deduce the characteristic parameters of the radiation source signal from the instantaneous parameter information; wherein, the characteristic parameters include amplitude spectral density, instantaneous amplitude variance, instantaneous frequency variance, and / or instantaneous phase variance.

[0087] In step S2, the characteristic parameters of the radiation source signal deduced from the instantaneous parameter information include:

[0088]

[0089] In Equation 2, Z1 is the maximum value of the amplitude spectral density, Z2 is the instantaneous amplitude variance, Z3 is the instantaneous frequency variance, Z4 is the instantaneous phase variance, is the Fourier operator, c is the number of non-weak signal values among the N s sampling points of the radiation source signal. It should be noted that in this embodiment, the non-weak signal value is a signal value greater than the normalized instantaneous amplitude mean.

[0090] In this embodiment, it is set that Z = [Z1, Z2, Z3, Z4], and Z is the self-characteristic vector of the radiation source signal, which is used as the input value of the subsequent optimal network structure.

[0091] S3. Simulate the analog signals of each modulation method and calculate the characteristic parameters corresponding to the analog signals of each modulation method, and generate a labeled data set according to the analog signals of each modulation method and their corresponding characteristic parameters.

[0092] In step S3, the analog signals of each modulation method simulated include multi-amplitude shift keying modulation signals, multi-frequency shift keying modulation signals, multi-phase shift keying modulation signals, amplitude modulation signals, linear frequency modulation signals, and continuous wave modulation signals; wherein,

[0093] The multi-amplitude shift keying modulation signal is as follows:

[0094]

[0095] In Equation 3, e MASK (t) is the multi-amplitude shift keying signal obtained by simulation and varying with time t, a n is the level of the n-th symbol, g() is the window function, generally a rectangular window, T s is the time interval of g(), ω c is the angular frequency of the carrier signal;

[0096] The multi - frequency shift keying modulation signal is as follows:

[0097]

[0098] In Equation 4, e MFSK (t) is the multi - level frequency - shift keying signal obtained by simulation and varying with time t, A is the signal amplitude, ω i is the angular frequency of the i - th carrier signal;

[0099] The multi - phase shift keying modulation signal is as follows:

[0100]

[0101] In Equation 5, e MPSK (t) is the multi - level phase - shift keying signal obtained by simulation and varying with time t, φ n is the initial phase of the n - th carrier signal;

[0102] The amplitude - modulation signal is as follows:

[0103] e AM (t)=a n (1 + m c v Ω (t))cosω c t; Equation 6

[0104] In Equation 6, e AM (t) is the amplitude - modulation signal obtained by simulation and varying with time t, m c is the modulation index, ranging from [0, 1], ν Ω (t) is the modulation signal, which reflects the envelope change law of the modulated e AM (t) signal;

[0105] The chirp signal is as follows:

[0106]

[0107] In Equation 7, e LFM (t) is the chirp signal obtained by simulation and varying with time t, rect is the rectangular function transformation symbol, K is the chirp rate, exp() is the exponential function with the natural constant e as the base, and j is the imaginary symbol;

[0108] The continuous - wave modulation signal is as follows:

[0109] e CW (t)=a(t)cos[ω c t + θ(t)]; Equation 8

[0110] In Equation 8, e CW(t) is a continuous wave modulation signal obtained by simulation and varying with time t, a(t) is an amplitude function, and θ(t) is a phase function.

[0111] S4. Construct an initial deep neural network, and encode all weights and thresholds in the initial deep neural network to obtain multiple individual encodings. In this embodiment, the structural schematic diagram of the initial deep neural network is as Figure 2 shown.

[0112] In step S4, constructing an initial deep neural network and encoding all weights and thresholds in the initial deep neural network to obtain multiple individual encodings includes:

[0113] S401. Construct an initial deep neural network; wherein, the initial deep neural network adopts a three-layer neural network, and the initial deep neural network includes an input layer, a hidden layer, and an output layer connected in sequence; in this embodiment, the activation function of the initial deep neural network adopts the Sigmoid function

[0114] S402. Initialize all weights and thresholds in the initial deep neural network; wherein, the weights in the initial deep neural network include the weights between the input layer and the hidden layer and the weights between the hidden layer and the output layer, and the thresholds in the initial deep neural network include the hidden layer threshold and the output layer threshold;

[0115] S403. Encode all weights and thresholds to obtain multiple individual encodings. It should be noted that in this embodiment, each individual encoding has a corresponding relationship with the corresponding weight or threshold.

[0116] It should be noted that in this embodiment, the individual encoding can but is not limited to binary encoding and real number encoding. When the individual encoding adopts binary encoding, each individual is a binary number string, which is composed of four parts: the weight between the input layer and the hidden layer, the hidden layer threshold, the weight between the hidden layer and the output layer, and the output layer threshold. When the individual encoding adopts real number encoding, each individual is a real number string, which is composed of four parts: the weight between the input layer and the hidden layer, the hidden layer threshold, the weight between the hidden layer and the output layer, and the output layer threshold.

[0117] S5. Perform bit crossover and bit transformation operations on the specified individual encoding, and then use the labeled data set to train the initial deep neural network to obtain a trained neural network; in this embodiment, through this step, the role of optimizing the network weight structure based on the individual encoding algorithm is achieved, which is beneficial to realizing high-precision recognition of the modulation mode of the radiation source signal in a low signal-to-noise ratio environment.

[0118] In step S5, perform bit position crossover and bit position transformation operations on the specified individual codes, and then use the labeled dataset to train the initial deep neural network to obtain a trained neural network, including:

[0119] S501. Perform a random bit position crossover operation on two adjacent individual codes in the same layer of the initial deep neural network; wherein, when the j-th bit a kj of the k-th individual code and the j-th bit a lj of the l-th individual code perform a bit position crossover operation, the following conditions are satisfied:

[0120] a kj = a kj (1 - b)+a lj b; Equation 9

[0121] In Equation 9, b is a random number in the interval [0, 1];

[0122] After all individual codes have been subjected to the crossover operation, randomly select an individual code in each layer of the initial deep neural network for bit position transformation operation; wherein, when the j-th bit a ij of the i-th individual code performs a bit position transformation operation, the following conditions are satisfied:

[0123]

[0124] In Equation 10, a max and a min are respectively the maximum and minimum values that each bit of the code a ij can take, g is the current iteration number, r is a random number in the interval [0, 1], r2 is a random number, g max is the preset maximum tolerable iteration number;

[0125] S503. Use the analog signals of each modulation method in the labeled dataset as the input data of the initial deep neural network, and use the characteristic parameters corresponding to the analog signals of each modulation method as the output data of the initial deep neural network to train the initial deep neural network.

[0126] S6. Calculate the partial variance of the trained neural network, and estimate the structure approximation degree of the trained neural network based on the partial variance; it should be noted that the partial variance is used to quantify the error between the actual output and the expected output.

[0127] In step S6, the partial variance of the trained neural network is:

[0128]

[0129] In Equation 11, m is the number of samples output by the trained neural network, each sample being an n-dimensional vector, and the actual output being x j , and the expected output is

[0130] The structural approximation degree of the trained neural network is the reciprocal of (MSE f +1), and

[0131] In this embodiment, considering the case where the mean square error is equal to 0, the reciprocal of (MSE f +1) is called the structural approximation degree and is denoted as Obviously When is closer to 1, the output effect is closer to the expected value, that is, the weights and threshold performance corresponding to the network structure are also better at this time.

[0132] S7. Obtain the optimal network structure according to the structural approximation degree and a preset approximation degree threshold.

[0133] In step S7, obtaining the optimal network structure according to the structural approximation degree and a preset approximation degree threshold includes:

[0134] S701. Set an approximation degree threshold for the structural approximation degree;

[0135] S702. When the structural approximation degree is greater than the approximation degree threshold, or the current iteration number exceeds a preset maximum iteration number, use the trained neural network as the optimal network structure;

[0136] Otherwise, re-perform bit crossover and bit transformation operations on the specified individual encoding, that is, re-execute steps S5, S6, and S7 until the optimal network structure is obtained.

[0137] S8. Input the characteristic parameters of the radiation source signal as the characteristic parameters of the signal to be recognized into the optimal network structure to obtain the maximum probability result of the modulation mode of the radiation source signal.

[0138] As an example, in this embodiment, the recognition result of the optimal network structure under the condition of no noise is as Figure 3 shown. The recognition results at different signal-to-noise ratios before and after optimizing the neural network using individual encoding are as Figure 4 shown. Obviously, in this embodiment, due to performing bit crossover and bit transformation operations on the specified individual encoding, the accuracy of the recognition results of the modulation mode of the finally obtained optimal network structure at different signal-to-noise ratios can be improved.

[0139] This embodiment can improve the accuracy and stability of the modulation mode identification of the radiation source signal in a low signal-to-noise ratio environment. Specifically, this embodiment proposes a method for identifying the modulation mode of an unknown radiation source using the characteristic parameters of the signal itself in conjunction with a global optimization neural network. During the implementation process, the characteristic parameters of the radiation source signal are first obtained, that is, the shallow characteristics of the radiation source signal itself are estimated, and then the neural network is constructed and the optimal network structure is obtained. Based on the superior performance of the deep neural network in nonlinear approximation, the characteristic parameters of the radiation source signal can be input into the optimal network structure after multiple layers of individual coding changes, so as to explore and capture the potential characteristics of the radiation source signal, thereby making the modulation mode output result more stable and efficient. Compared with the recognition of ordinary neural networks, the noise resistance and recognition accuracy of the modulation mode recognition of the radiation source signal in a low signal-to-noise ratio environment in this embodiment are significantly improved.

[0140] Embodiment 2:

[0141] This embodiment discloses a radiation source signal modulation mode identification system, which is used to implement the radiation source signal modulation mode identification method in embodiment 1; Figure 5 As shown, the radiation source signal modulation mode identification system includes:

[0142] A radiation source signal information extraction module, used to receive a radiation source signal and extract instantaneous parameter information from the radiation source signal; and also used to calculate characteristic parameters of the radiation source signal through the instantaneous parameter information;

[0143] A neural network construction module is used to simulate analog signals of various modulation modes and calculate characteristic parameters corresponding to the analog signals of various modulation modes, and generate a labeled data set according to the analog signals of various modulation modes and their corresponding characteristic parameters; to construct an initial deep neural network, and encode all weights and thresholds in the initial deep neural network to obtain multiple individual codes; to perform code bit crossover and code bit transformation operations on designated individual codes, and then train the initial deep neural network using the labeled data set to obtain a trained neural network; to calculate the partial variance of the trained neural network, and to infer the structural approximation of the trained neural network according to the partial variance; and to obtain an optimal network structure according to the structural approximation and a preset approximation threshold;

[0144] The modulation mode prediction module is respectively communicated with the radiation source signal information extraction module and the neural network construction module, and is used to input the characteristic parameters of the radiation source signal as the characteristic parameters of the signal to be identified into the optimal network structure to obtain the maximum probability result of the modulation mode of the radiation source signal.

[0145] Embodiment 3:

[0146] Based on Embodiment 1 or 2, this embodiment discloses an electronic device, which may be a smart phone, a tablet computer, a laptop computer, a desktop computer, or the like. The electronic device may be referred to as a terminal device, a portable terminal device, a desktop terminal device, etc. The electronic device includes:

[0147] A memory for storing computer program instructions; and,

[0148] A processor for executing the computer program instructions to complete the operations of the radiation source signal modulation mode recognition method as described in any one of Embodiment 1.

[0149] Embodiment 4:

[0150] Based on any one of Embodiments 1 to 3, this embodiment discloses a computer-readable storage medium for storing computer-readable computer program instructions, and the computer program instructions are configured to perform the operations of the radiation source signal modulation mode recognition method as described in Embodiment 1 when running.

[0151] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device, so that they can be stored in a storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module to implement. In this way, the present invention is not limited to any specific combination of hardware and software.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying a modulation mode of a radiation source signal, characterized in that: Including: Receiving a radiation source signal and extracting instantaneous parameter information from the radiation source signal; Calculating characteristic parameters of the radiation source signal based on the instantaneous parameter information; Simulating analog signals of various modulation modes, calculating characteristic parameters corresponding to the analog signals of various modulation modes, and generating a labeled data set according to the analog signals of various modulation modes and their corresponding characteristic parameters; Constructing an initial deep neural network and encoding all weights and thresholds in the initial deep neural network to obtain a plurality of individual encodings; Performing bit crossover and bit transformation operations on a specified individual encoding, and then training the initial deep neural network using the labeled data set to obtain a trained neural network; Calculating the partial variance of the trained neural network and inferring the structure approximation degree of the trained neural network according to the partial variance; Obtaining an optimal network structure according to the structure approximation degree and a preset approximation degree threshold; Taking the characteristic parameters of the radiation source signal as the characteristic parameters of the signal to be recognized and inputting them into the optimal network structure to obtain the maximum probability result of the modulation mode of the radiation source signal; Performing bit crossover and bit transformation operations on a specified individual encoding, and then training the initial deep neural network using the labeled data set to obtain a trained neural network, including: Performing a random bit crossover operation on two adjacent individual encodings in the same layer of the initial deep neural network; After performing the crossover operation on all individual encodings, randomly selecting an individual encoding in each layer of the initial deep neural network for bit transformation operation; Using the analog signals of various modulation modes in the labeled data set as the input data of the initial deep neural network, and using the characteristic parameters corresponding to the analog signals of various modulation modes as the output data of the initial deep neural network to train the initial deep neural network; When the j-th bit a of the k-th individual code kj and the j-th bit a of the l-th individual code lj perform a bit crossover operation, the following conditions are met: a kj = a kj (1 - b)+a lj b; Equation 9 In Equation 9, b is a random number in the interval [0, 1]; When the j-th bit a of the i-th individual code ij performs a code bit transformation operation, the following conditions are satisfied: In Equation 10, a max and a min are respectively the maximum and minimum values that each digit of the encoded a ij can take, g is the current iteration number, r is a random number in the interval [0, 1], r2 is a random number, and g max is the preset maximum tolerable number of iterations.

2. The method for identifying a modulation mode of a radiation source signal according to claim 1, wherein: The instantaneous parameter information extracted from the radiation source signal includes: In Equation 1, N s is the number of sampling points of the radiation source signal, a(i) is the instantaneous amplitude of the radiation source signal at the i-th sampling point, and m a is the mean value of the instantaneous amplitude of the radiation source signal; a n (i) is the normalized instantaneous amplitude of the radiation source signal at the i-th sampling point, and a cn (i) is the zero-centered normalized instantaneous amplitude of the radiation source signal at the i-th sampling point; is the mean value of the normalized instantaneous amplitude of the radiation source signal; f(i) is the instantaneous frequency of the radiation source signal at the i-th sampling point, is the normalized instantaneous frequency of the radiation source signal at the i-th sampling point; R s is the symbol rate of the radiation source signal; is the instantaneous phase of the radiation source signal at the i-th sampling point, and f s is the signal frequency, and f c is the carrier frequency, is the non-linear phase component of the radiation source signal at the i-th sampling point.

3. The method for identifying a modulation mode of a radiation source signal according to claim 2, characterized in that: The characteristic parameters of the radiation source signal calculated based on the instantaneous parameter information include: In Equation 2, Z1 is the maximum value of the amplitude spectral density, Z2 is the instantaneous amplitude variance, Z3 is the instantaneous frequency variance, and Z4 is the instantaneous phase variance. is the Fourier operator, and c is the number of non-weak signal values among the N s sampling points of the radiation source signal.

4. The method for identifying a radiation source signal modulation method according to claim 1, wherein: The analog signals of various modulation modes simulated include multiple amplitude shift keying modulation signals, multiple frequency shift keying modulation signals, multiple phase shift keying modulation signals, amplitude modulation signals, linear frequency modulation signals, and continuous wave modulation signals; where The multiple amplitude shift keying modulation signal is as follows: In Equation 3, e MASK (t) is the multiple shift keying signals obtained by simulation and varying with time t, a n is the level of the nth symbol, g() is the window function, T s is the time interval of g(), ω c is the angular frequency of the carrier signal; The multiple frequency shift keying modulation signal is as follows: In Equation 4, e MFSK (t) is the multi - level frequency - shift keying signal obtained by simulation and varying with time t, A is the signal amplitude, ω i is the angular frequency of the i - th carrier signal; The multiple phase shift keying modulation signal is as follows: In Equation 5, e MPSK (t) is a multi - level phase - shift keying signal obtained by simulation and varying with time t, and φ n is the initial phase of the nth carrier signal; The amplitude modulation signal is as follows: e AM e(t) = a n (1 + m c v Ω (t)) cos ω c t; Equation 6 In Equation 6, e AM (t) is the amplitude-modulated signal obtained by simulation and varying with time t, and m c is the modulation index, with a range between [0, 1], and ν Ω (t) is the modulation signal; The linear frequency modulation signal is as follows: In Equation 7, e LFM (t) is the chirp signal obtained by simulation and varying with time t, rect is the rectangular function transformation symbol, K is the chirp rate, exp() is the exponential function with the natural constant e as the base, and j is the imaginary symbol; The continuous wave modulation signal is as follows: e CW (t) = a(t) cos[ω c t + θ(t)]; Equation 8 In Equation 8, e CW (t) is the continuous wave modulation signal obtained by simulation and varying with time t, a(t) is the amplitude function, and θ(t) is the phase function.

5. A method for identifying a modulation method of a radiation source signal according to claim 1, characterized in that: Constructing an initial deep neural network and encoding all weights and thresholds in the initial deep neural network to obtain a plurality of individual encodings, including: Constructing an initial deep neural network; where the initial deep neural network adopts a three-layer neural network, and the initial deep neural network includes an input layer, a hidden layer, and an output layer connected in sequence; Initialize all the weights and thresholds in the initial deep neural network; wherein, the weights in the initial deep neural network include the weights between the input layer and the hidden layer and the weights between the hidden layer and the output layer, and the thresholds in the initial deep neural network include the hidden layer threshold and the output layer threshold; Encode all the weights and thresholds to obtain a plurality of individual encodings.

6. The method for identifying a modulation mode of a radiation source signal according to claim 1, wherein: The partial variance of the trained neural network is: In Equation 11, m is the number of output samples of the trained neural network, and the actual output is x j , and the expected output is (j = 1, 2, 3... m); The structural approximation degree of the trained neural network is the reciprocal of (MSE f + 1), and 7. A method for identifying a modulation method of a radiation source signal according to claim 1, characterized in that: According to the structural approximation degree and a preset approximation degree threshold, obtain the optimal network structure, including: Set an approximation degree threshold for the structural approximation degree; When the structural approximation degree is greater than the approximation degree threshold, or the current iteration number exceeds the preset maximum iteration number, use the trained neural network as the optimal network structure; Otherwise, re-perform bit crossover and bit transformation operations on the specified individual encoding until the optimal network structure is obtained.

8. A radiation source signal modulation mode recognition system, characterized in that: For implementing the radiation source signal modulation mode recognition method according to any one of claims 1 to 7; the radiation source signal modulation mode recognition system includes: A radiation source signal information extraction module, configured to receive a radiation source signal, extract instantaneous parameter information from the radiation source signal; and further configured to deduce the characteristic parameters of the radiation source signal through the instantaneous parameter information; A neural network construction module, configured to simulate the analog signals of each modulation mode and calculate the characteristic parameters corresponding to the analog signals of each modulation mode, and generate a labeled data set according to the analog signals of each modulation mode and their corresponding characteristic parameters; configured to construct an initial deep neural network, encode all the weights and thresholds in the initial deep neural network to obtain a plurality of individual encodings; configured to perform bit crossover and bit transformation operations on the specified individual encoding, and then use the labeled data set to train the initial deep neural network to obtain a trained neural network; configured to calculate the partial variance of the trained neural network, and deduce the structural approximation degree of the trained neural network according to the partial variance; and further configured to obtain the optimal network structure according to the structural approximation degree and a preset approximation degree threshold; A modulation mode prediction module, communicatively connected to the radiation source signal information extraction module and the neural network construction module respectively, configured to input the characteristic parameters of the radiation source signal as the characteristic parameters of the signal to be recognized into the optimal network structure to obtain the maximum probability result of the modulation mode of the radiation source signal.

Images