A MPSK-type burst signal modulation recognition method and system based on AlexNet network

Abstract

The invention discloses a method and system for MPSK-type burst signal modulation recognition based on AlexNet network, which detects the start and end positions of signals based on delayed addition double sliding window energy detection method, then generates five complex baseband signals after down-conversion and matched filtering, and blindly estimates and compensates the complex baseband signals based on fast Fourier transform and spectrum analysis method to generate training data set and test data set; extracts signal features from the five signals of the training data set to generate constellation diagrams, which are input into the AlexNet network training recognition model; then generates differential high-order cumulants of the signals separately, which are input into the DNN network training recognition model; finally, extracts signal features from the five signals of the test data set to generate constellation diagrams, which are sent to the AlexNet network for coarse classification and the DNN network for fine classification. The invention can realize the start and end position detection and modulation mode recognition tasks of low signal-to-noise ratio burst signals under non-cooperative communication without any prior information, and ensures that the recognition accuracy is above 90%.

Description

Technical Field

[0001] The invention belongs to the technical field of signal modulation recognition, and in particular relates to an MPSK type burst signal modulation recognition method and system based on an AlexNet network. Background Art

[0002] Communication systems are widely used in both civil and military fields. In order to ensure the accuracy of transmission, different modulation methods are used, which can occupy different bandwidths and frequencies. Therefore, the modulation recognition and classification of communication signals are of great significance in both military and civilian fields. Three methods are often mentioned in the field of signal modulation recognition. The first method is an algorithm based on maximum likelihood value, the second method is a feature-based recognition algorithm, the most common method is to use the accumulated features of different signals to distinguish signals, and another effective method is to use deep learning for feature extraction.

[0003] However, the actual communication environment is changeable and complex, which brings many challenges to the recognition of communication signals. Due to the Doppler effect and the difference in transceiver oscillators, the received signal will have a frequency offset. The frequency offset has a great influence on the demodulation of wireless communication and may cause inter-symbol interference. Therefore, it is necessary to accurately estimate the frequency offset of the received signal so that the influence of the frequency offset can be eliminated through corresponding compensation, so as to further accurately perform the modulation recognition of the signal. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide an MPSK-type burst signal modulation recognition method and system based on the AlexNet network in view of the deficiencies in the above-mentioned prior art, which is used to solve the technical problem that the modulation mode of the burst signal cannot be recognized under the condition of frequency offset.

[0005] The present invention adopts the following technical solutions:

[0006] A method for identifying MPSK-like burst signal modulation based on AlexNet network comprises the following steps:

[0007] S1. The receiver obtains a predetermined signal and determines the start and end positions of the burst signal based on a delayed-addition double sliding window energy detection method;

[0008] S2, segmenting the burst signal according to the start and end positions obtained in step S1, performing carrier frequency estimation and bandwidth estimation, and obtaining a complex baseband signal after bandpass filtering and down-conversion;

[0009] S3, performing matched filtering on the characteristic parameters of the complex baseband signal obtained in step S2, and performing frequency offset estimation and frequency offset compensation on the complex baseband signal based on fast Fourier transform, and generating a training data set and a test data set in proportion;

[0010] S4, extracting the in-phase component and the orthogonal component of the training data set signal obtained in step S3 to generate a constellation diagram, and designing a modulation mode recognition method based on the deep learning AlexNet network idea to train the AlexNet recognition model;

[0011] S5, calculating the differential high-order cumulants of the signals in the training data set obtained in step S3, and designing a modulation mode recognition method based on the deep learning DNN network concept to train a DNN recognition model;

[0012] S6. Extract the in-phase components and orthogonal components of the five signals in the test data set obtained in step S3 to generate a constellation diagram, input the AlexNet model obtained in step S4 for rough identification and classification, calculate the differential high-order cumulants of the signals in the classification results, and input the DNN model obtained in step S5 for fine identification and classification.

[0013] Specifically, in step S1, the start and end positions of the burst signal are determined as follows:

[0014] S101, convert the received signal sampling into a discrete sequence r(n), delay the discrete sequence r(n) by one sampling point to obtain r(n-1), perform instantaneous energy calculation on the data after adding the discrete sequences r(n) and r(n-1) to obtain the power and Z of the effective signal and noise in the received signal n ;

[0015] S102, using the principle of double sliding windows to calculate the energy of the sampling points of windows A and B and A n and B n , calculate A n and B n The ratio is M n ;

[0016] S103, calculate the M obtained in step S102 n The maximum value M nmax ; Set the reference threshold β, when M nmax >β indicates there is a signal;

[0017] S104: After performing energy detection on the signal obtained in step S103 to obtain the signal starting point, use the reverse window energy ratio M nr =B n / A n Find M nr The minimum value gives the end position of the signal.

[0018] Specifically, in step S2, the complex baseband signal is obtained as follows:

[0019] S201, estimating the carrier frequency f of the signal r(n) according to the signal spectrum diagram c and bandwidth Bw ;

[0020] S202, processing the signal r(n) to obtain a complex baseband signal, as follows:

[0021]

[0022] Where h[n] is the unit impulse response of the bandpass filter, r(n) is the discrete sequence, and T is the sampling period.

[0023] Specifically, in step S3, frequency offset estimation and frequency offset compensation for the complex baseband signal based on fast Fourier transform are specifically performed as follows:

[0024] S301, fixing parameters and performing matched filtering on the complex baseband signal;

[0025] S302, performing nonlinear transformation on the complex baseband signal to remove initial phase information, and performing fast Fourier transform transformation;

[0026] S303, searching for the maximum peak based on the spectrum diagram transformed by fast Fourier transform, where the frequency corresponding to the maximum peak is the frequency deviation of the complex baseband signal;

[0027] S304: Perform frequency offset compensation on the complex baseband signal.

[0028] Furthermore, step S304 is specifically as follows:

[0029] S3041, determine the burst signal w n is additive Gaussian white noise, φ n is the initial phase, θ is the phase offset, and T is the symbol period;

[0030] S3042, performing N-point Fourier transform on the signal to obtain a signal spectrum;

[0031] S3043, Search for the highest peak in the signal spectrum | F k | max Get the corresponding k max , and then according to the maximum likelihood estimation theory, the estimated expression of the frequency offset is obtained as follows:

[0032]

[0033] S3044, the estimated frequency deviation is The compensated signal r'(n) is determined as follows:

[0034]

[0035] Specifically, in step S4, the modulation mode recognition method based on the deep learning AlexNet idea is as follows:

[0036] S401, establish an AlexNet network, and determine the loss function, learning rate, and update method of the AlexNet network;

[0037] S402, input the constellation diagrams of five frequency offset compensated signals, namely BPSK, QPSK, OQPSK, PI4QPSK and 8PSK, into the AlexNet network, give classification labels and train the AlexNet recognition model.

[0038] Specifically, in step S5, the modulation mode recognition method based on the deep learning DNN idea is specifically:

[0039] S501, establish a DNN network, determine the network's loss function, learning rate, and update method;

[0040] S502, input the differential high-order cumulants of PI4QPSK and 8PSK into the DNN network, give classification labels and train the recognition model.

[0041] Furthermore, in step S502, the difference high-order cumulant is calculated as follows:

[0042] S5021. Calculate the differential characteristics of the signal

[0043] S5022, calculate the eighth-order difference high-order cumulant of the signal The specific calculation is as follows:

[0044]

[0045] Where N is the length of the signal, is the differential feature quantity, C 20 is the second-order difference cumulant, C 40 is the fourth-order difference cumulant, C 60 is the sixth-order difference cumulant.

[0046] In a second aspect, an embodiment of the present invention provides an MPSK-type burst signal modulation recognition system based on an AlexNet network, characterized in that it includes:

[0047] The receiving module, the receiver obtains the predetermined signal and determines the start and end positions of the burst signal based on the delayed addition double sliding window energy detection method;

[0048] The processing module segments the burst signal according to the start and end positions obtained by the receiving module, performs carrier frequency estimation and bandwidth estimation, and obtains a complex baseband signal after bandpass filtering and down-conversion;

[0049] The data module calculates and processes the characteristic parameters of the complex baseband signal obtained by the processing module for matched filtering, and performs frequency offset estimation and frequency offset compensation on the complex baseband signal based on fast Fourier transform, and generates a training data set and a test data set in proportion;

[0050] AlexNet module, extracts the in-phase component and orthogonal component of the training data set signal obtained by the data module to generate a constellation diagram, and designs a modulation mode recognition method based on the deep learning AlexNet network idea to train the AlexNet recognition model;

[0051] The DNN module calculates the differential high-order cumulants of the signals in the training data set obtained by the data module, and designs a modulation mode recognition method based on the deep learning DNN network concept to train the DNN recognition model;

[0052] The recognition module extracts the in-phase components and orthogonal components of the five signals in the test data set obtained by the data module to generate a constellation diagram, inputs the AlexNet model obtained by the AlexNet module for rough recognition and classification, and then calculates the differential high-order cumulant of the signal in the classification result, and inputs the DNN model obtained by the DNN module for fine recognition and classification.

[0053] Compared with the prior art, the present invention has at least the following beneficial effects:

[0054] The invention discloses an MPSK type burst signal modulation recognition method based on AlexNet network, which determines the start and end positions of the burst signal, performs parameter estimation and frequency offset compensation after obtaining the effective signal, and then inputs the signal into a neural network for training, so as to design a modulation mode recognition algorithm with maximum recognition accuracy and maximum practicality. The invention can realize accurate recognition of burst signals of five PSK modulation forms by combining two signal features of constellation diagram and differential high-order cumulant without any prior information.

[0055] Furthermore, detecting the start and end positions of the burst signal can determine the effective signal length, thereby extracting the signal containing the modulation information from the chaotic noise.

[0056] Furthermore, down-conversion to obtain a baseband signal can effectively remove the carrier information in the signal to prevent it from interfering with subsequent signal processing.

[0057] Furthermore, the signal is affected by the complex geographical environment during the channel propagation process, and the signal at the receiving end will have a frequency offset phenomenon. The characteristic information of the signal with frequency offset will have a huge deviation, so the modulation mode cannot be identified. Frequency offset estimation and frequency offset compensation can restore the frequency information of the signal to a certain extent, eliminate the adverse effects of frequency offset on signal identification, and improve the accuracy of signal identification.

[0058] Furthermore, frequency offset compensation cannot completely eliminate the interference information of signal recognition, so it is necessary to use deep learning methods to replace traditional signal processing methods for modulation recognition. Deep learning methods can more accurately distinguish the characteristic differences of different signals by training neural networks, thereby performing accurate recognition. This method uses the constellation diagram of the signal as the recognition feature, so choosing the AlexNet network that performs well in the field of image recognition is more conducive to the development of modulation recognition.

[0059] Furthermore, the constellation diagram features and AlexNet network are still unable to accurately distinguish the two signals PI4QPSK and 8PSK, so the DNN network is required for secondary recognition. The use of two deep learning models has greatly improved the signal recognition accuracy.

[0060] Furthermore, the feature of signal differential high-order cumulant was selected as input into DNN network recognition, which solved the problem that the constellation diagrams of PI4QPSK and 8PSK signals are extremely similar and cannot be used for image recognition methods.

[0061] Furthermore, the constellation diagram features and differential high-order cumulant features of the five modulation signals in the test set were extracted and input into two network models for recognition to verify the effectiveness of the method.

[0062] It can be understood that the beneficial effects of the second aspect mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.

[0063] In summary, the present invention can detect the start and end positions of signals and compensate for frequency deviation of burst signals of five PSK modulation forms without any prior information, and on this basis, use the two signal features of constellation diagram and differential high-order cumulant to jointly realize accurate identification of modulation mode, which can help capture and identify signals in non-cooperative communications in military or civil fields.

[0064] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 is a flow chart of the present invention;

[0066] Figure 2 This is a graph showing the probability of successful identification of the start and end positions of a signal as a function of the signal-to-noise ratio;

[0067] Figure 3 It is the graph showing the variation of signal missed detection probability with signal-to-noise ratio;

[0068] Figure 4BPSK signal constellation diagram before and after frequency offset compensation, where (a) is the BPSK signal constellation diagram with frequency offset, and (b) is the BPSK signal constellation diagram after frequency offset compensation;

[0069] Figure 5 : The constellation diagrams of the QPSK signal before and after frequency offset compensation, where (a) is the constellation diagram of the QPSK signal with frequency offset influence, and (b) is the constellation diagram of the QPSK signal after frequency offset compensation;

[0070] Figure 6 : The constellation diagrams of the OQPSK signal before and after frequency offset compensation, where (a) is the constellation diagram of the OQPSK signal with frequency offset influence, and (b) is the constellation diagram of the OQPSK signal after frequency offset compensation;

[0071] Figure 7 PI4QPSK signal constellation diagram before and after frequency offset compensation, where (a) is the PI4PSK signal constellation diagram with frequency offset, and (b) is the PI4PSK signal constellation diagram after frequency offset compensation;

[0072] Figure 8 8PSK signal constellation diagram before and after frequency offset compensation, where (a) is the 8PSK signal constellation diagram with frequency offset, and (b) is the 8PSK signal constellation diagram after frequency offset compensation;

[0073] Fig. 9 The diagram of the recognition success rate of the compensated BPSK signal versus the carrier-to-noise ratio;

[0074] Fig.10 The diagram of the recognition success rate of the compensated QPSK signal versus the carrier-to-noise ratio;

[0075] Fig.11 The diagram of the recognition success rate of the compensated OQPSK signal versus the carrier-to-noise ratio;

[0076] Fig.12 The diagram of the recognition success rate of the compensated PI4QPSK signal versus the carrier-to-noise ratio;

[0077] Fig.13 This is a graph showing the variation of recognition success rate of the compensated 8PSK signal with the carrier-to-noise ratio. DETAILED DESCRIPTION

[0078] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0079] In the description of the present invention, it should be understood that the terms “include” and “comprising” indicate the presence of the described features, integers, S, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, S, operations, elements, components and / or collections thereof.

[0080] It should also be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0081] It should be further understood that the term "and / or" used in the present specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects are in an "or" relationship.

[0082] It should be understood that, although the terms first, second, third, etc. may be used to describe preset ranges, etc. in the embodiments of the present invention, these preset ranges should not be limited to these terms. These terms are only used to distinguish preset ranges from each other. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0083] The word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)", depending on the context.

[0084] Various structural schematic diagrams of the embodiments disclosed in the present invention are shown in the accompanying drawings. These figures are not drawn to scale, and some details are magnified and some details may be omitted for the purpose of clear expression. The shapes of various regions and layers shown in the figures and the relative sizes and positional relationships therebetween are only exemplary, and may deviate in practice due to manufacturing tolerances or technical limitations, and those skilled in the art may additionally design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0085] The present invention provides a modulation recognition method for MPSK-type burst signals based on an AlexNet network. The signal start and end positions are detected based on a delayed addition double sliding window energy detection method, and then five complex baseband signals, BPSK, QPSK, OQPSK, PI4QPSK, and 8PSK, are generated after down-conversion and matched filtering. The complex baseband signals are blindly compensated for frequency deviation based on a fast Fourier transform and spectrum analysis method to generate a training data set and a test data set for modulation recognition. Secondly, the signal features of the five signals in the training data set are extracted to generate a constellation diagram, which is input into the AlexNet network training recognition model; then the differential high-order cumulants of the signals are generated separately and input into the DNN network training recognition model. Finally, the signal features of the five signals in the test data set are extracted to generate a constellation diagram and sent to the AlexNet network for coarse classification and the DNN network for fine classification. Compared with the existing modulation recognition algorithm, the design scheme of the present invention can realize the start and end position detection and modulation mode recognition tasks of low signal-to-noise ratio burst signals under non-cooperative communication without any prior information, and ensure that the recognition accuracy of the final modulation mode is above 90%.

[0086] See also Figure 1 The present invention discloses an MPSK burst signal modulation recognition method based on AlexNet network. The specific signal recognition scheme is to identify the starting position of the burst signal according to the delayed addition double sliding window energy detection method and the delayed autocorrelation detection method, perform parameter estimation and frequency offset compensation on the signal after segmentation processing, and finally convert the processed signal into the form of a constellation diagram and input it into the AlexNet network for coarse classification, and then extract the differential high-order cumulant features of the signal and input it into the DNN network for fine classification; the specific steps are as follows:

[0087] S1. The receiver obtains a predetermined signal and determines the start and end positions of the burst signal based on a delayed-addition double sliding window energy detection method;

[0088] The specific process of signal start and end position detection is:

[0089] S101, convert the received signal sampling into a discrete sequence r(n), delay r(n) by one sampling point to obtain r(n-1), calculate the instantaneous energy of the data after adding r(n) and r(n-1) to obtain Z n ;

[0090] r(n)=s(n)+w(n)

[0091] Z n =P s +P w

[0092]

[0093] P s=[s(n)+s(n-1)]·[s(n)+s(n-1)] * ≈4|s(n)| 2

[0094] in, are the noise power before and after the delay respectively.

[0095] S102, using the principle of double sliding windows to calculate the energy of the sampling points of windows A and B and A n and B n , calculate A n and B n The ratio is M n ;

[0096]

[0097]

[0098] Where L is the window length.

[0099] S103, Find M n Maximum value M nmax ; Set the reference threshold β, when M nmax >β indicates there is a signal, otherwise there is no signal;

[0100]

[0101] S104, after performing energy detection on the signal to obtain the signal starting point, similarly use the reverse window energy ratio M nr =B n / A n Find M nr The minimum value gives the end position of the signal.

[0102]

[0103] S2, segmenting the burst signal according to the start and end positions, performing carrier frequency estimation and bandwidth estimation, and obtaining a complex baseband signal after bandpass filtering and down-conversion;

[0104] The generation process of complex baseband signal is:

[0105] S201, estimating the carrier frequency f of the signal r(n) according to the signal spectrum diagram c and bandwidth B w ;

[0106] S202, processing the signal r(n) to obtain a complex baseband signal;

[0107] The signal r(n) is processed as follows:

[0108]

[0109] Where h[n] is the unit impulse response of the bandpass filter and T is the sampling period.

[0110] S3, calculating the characteristic parameters of the complex baseband signal for matched filtering, and performing frequency offset estimation and frequency offset compensation on the signal based on fast Fourier transform, and generating a training data set and a test data set in a ratio of 7:3;

[0111] Signal parameter estimation and frequency offset compensation are specifically as follows:

[0112] S301, fixing parameters and performing matched filtering on the complex baseband signal;

[0113] S302, performing nonlinear transformation on the complex baseband signal to remove initial phase information, and performing fast Fourier transform transformation;

[0114] S303, searching for the maximum peak based on the spectrum diagram transformed by fast Fourier transform, where the frequency corresponding to the maximum peak is the frequency deviation of the complex baseband signal;

[0115] S304: Perform frequency offset compensation on the complex baseband signal.

[0116] S3041, determining a burst signal;

[0117] The burst signal is expressed as:

[0118]

[0119] Where θ is the phase shift, Δf is the frequency shift, is the initial phase, i=1,2,...,M / 2, M is the modulation order, w n represents additive white Gaussian noise.

[0120] S3042, performing N-point Fourier transform on the signal to obtain a signal spectrum;

[0121]

[0122] S3043, Search for the highest peak in the signal spectrum | F k | max Get the corresponding k max , and then according to the maximum likelihood estimation theory, the estimated expression of the frequency offset is obtained;

[0123] The estimated expression of frequency offset is as follows:

[0124]

[0125] S3044, the estimated frequency deviation is Then the compensated signal r'(n) is determined.

[0126] The compensated signal r'(n) is

[0127]

[0128] in, is the frequency compensation error.

[0129] S4. Extract the in-phase component and the orthogonal component of the training data set signal to generate a constellation diagram, design a modulation mode recognition method based on the deep learning AlexNet network idea, and train the AlexNet recognition model;

[0130] Based on the compensated signal, the modulation mode recognition method based on the deep learning AlexNet idea is as follows:

[0131] S401. Establish the AlexNet network and determine the network's loss function, learning rate, and update method;

[0132] Loss function L log (Y,P) is:

[0133]

[0134] Among them, y ik Indicates the actual modulation method number of the signal, p ik Indicates the signal modulation method number predicted by the network model.

[0135] The network learning rate is set to 0.001; the network weight parameter w t The update method uses the Adam (Adaptive MomentEstimation) optimization algorithm. The update method is as follows:

[0136] m t =β1*m t-1 +(1-β1)*g t

[0137]

[0138]

[0139]

[0140]

[0141] Among them, m t 、v t denote the exponentially weighted average, g tis the current gradient, β1 and β2 are hyperparameters, η is the learning rate, and ε is a very small positive number. is the adjusted exponentially weighted average, w t is the weight parameter of the current neural network.

[0142] S402, input the constellation diagrams of five frequency offset compensated signals, namely BPSK, QPSK, OQPSK, PI4QPSK and 8PSK, into the AlexNet network, give classification labels and train the AlexNet recognition model.

[0143] S5. Calculate the differential high-order cumulants of the signals in the training data set, design a modulation mode recognition method based on the deep learning DNN network concept, and train the recognition model;

[0144] The modulation mode recognition method based on deep learning DNN idea is as follows:

[0145] S501, establish a DNN network, the network's loss function, learning rate, and update method;

[0146] S502, input the differential high-order cumulants of PI4QPSK and 8PSK into the DNN network, give classification labels and train the DNN recognition model.

[0147] The calculation process of the eighth-order difference high-order cumulant of the signal is:

[0148] S5021. Calculate the differential characteristics of the signal

[0149]

[0150] S5022, calculate the eighth-order difference high-order cumulant of the signal

[0151]

[0152] in

[0153]

[0154]

[0155]

[0156]

[0157] Where X(k) is K=2,3,…,N, N is the length of the signal, For C 20 is the second-order difference cumulant, C 40 is the fourth-order difference cumulant, C60 is the sixth-order difference cumulant, is the differential feature quantity.

[0158] S6. Extract the in-phase components and orthogonal components of the five signals in the test data set to generate a constellation diagram, input it into the AlexNet model for rough identification and classification, then calculate the differential high-order cumulant of the signal in the classification result, and input it into the DNN model for fine identification and classification.

[0159] The classification method based on the recognition model is as follows:

[0160] S601, extracting the in-phase components and orthogonal components of the five signals in the test set to generate a constellation diagram, inputting the constellation diagram into the AlexNet recognition model for rough classification, and classifying the signals into four types: BPSK, QPSK, OQPSK, and 8PSK;

[0161] S602, calculate the eighth-order differential high-order cumulants of all 8PSK signals in the coarse classification results, input them into the DNN recognition model for fine classification, and classify them into two types of signals: PI4QPSK and 8PSK.

[0162] In another embodiment of the present invention, an MPSK-type burst signal modulation identification system based on an AlexNet network is provided. The system can be used to implement the above-mentioned MPSK-type burst signal modulation identification method based on an AlexNet network. Specifically, the MPSK-type burst signal modulation identification system based on an AlexNet network includes a receiving module, a processing module, a data module, an AlexNet module, a DNN module and a recognition module.

[0163] Among them, in the receiving module, the receiver obtains the predetermined signal and determines the start and end positions of the burst signal based on the delayed addition double sliding window energy detection method;

[0164] The processing module segments the burst signal according to the start and end positions obtained by the receiving module, performs carrier frequency estimation and bandwidth estimation, and obtains a complex baseband signal after bandpass filtering and down-conversion;

[0165] The data module calculates and processes the characteristic parameters of the complex baseband signal obtained by the processing module for matched filtering, and performs frequency offset estimation and frequency offset compensation on the complex baseband signal based on fast Fourier transform, and generates a training data set and a test data set in proportion;

[0166] AlexNet module, extracts the in-phase component and orthogonal component of the training data set signal obtained by the data module to generate a constellation diagram, and designs a modulation mode recognition method based on the deep learning AlexNet network idea to train the AlexNet recognition model;

[0167] The DNN module calculates the differential high-order cumulants of the signals in the training data set obtained by the data module, and designs a modulation mode recognition method based on the deep learning DNN network concept to train the DNN recognition model;

[0168] The recognition module extracts the in-phase components and orthogonal components of the five signals in the test data set obtained by the data module to generate a constellation diagram, inputs the AlexNet model obtained by the AlexNet module for rough recognition and classification, and then calculates the differential high-order cumulant of the signal in the classification result, and inputs the DNN model obtained by the DNN module for fine recognition and classification.

[0169] In another embodiment of the present invention, a terminal device is provided, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the MPSK-type burst signal modulation recognition method based on the AlexNet network, including:

[0170] The receiver obtains a predetermined signal, and determines the start and end positions of the burst signal based on the delayed addition double sliding window energy detection method; the burst signal is segmented according to the start and end positions, and the carrier frequency and bandwidth estimation are performed, and the complex baseband signal is obtained after bandpass filtering and down-conversion; the characteristic parameters of the complex baseband signal are calculated for matched filtering, and the frequency offset estimation and frequency offset compensation of the complex baseband signal are performed based on the fast Fourier transform, and a training data set and a test data set are generated in proportion; the in-phase component and the orthogonal component of the training data set signal are extracted to generate a constellation diagram, and a modulation mode recognition method based on the deep learning AlexNet network idea is designed to train the AlexNet recognition model; the differential high-order cumulants of the signals in the training data set are calculated, and a modulation mode recognition method based on the deep learning DNN network idea is designed to train the DNN recognition model; the in-phase component and the orthogonal component of the five signals in the test data set are extracted to generate a constellation diagram, which is input into the AlexNet model for rough identification classification, and then the differential high-order cumulants of the signals in the classification results are calculated, and input into the DNN model for fine identification classification.

[0171] In another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It is understandable that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory, or a non-volatile memory (Non0Volatile Memory), such as at least one disk memory.

[0172] The processor may load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the MPSK-type burst signal modulation identification method based on the AlexNet network in the above embodiment; the processor may load and execute the following steps:

[0173] The receiver obtains a predetermined signal, and determines the start and end positions of the burst signal based on the delayed addition double sliding window energy detection method; the burst signal is segmented according to the start and end positions, and the carrier frequency and bandwidth estimation are performed, and the complex baseband signal is obtained after bandpass filtering and down-conversion; the characteristic parameters of the complex baseband signal are calculated for matched filtering, and the frequency offset estimation and frequency offset compensation of the complex baseband signal are performed based on the fast Fourier transform, and a training data set and a test data set are generated in proportion; the in-phase component and the orthogonal component of the training data set signal are extracted to generate a constellation diagram, and a modulation mode recognition method based on the deep learning AlexNet network idea is designed to train the AlexNet recognition model; the differential high-order cumulants of the signals in the training data set are calculated, and a modulation mode recognition method based on the deep learning DNN network idea is designed to train the DNN recognition model; the in-phase component and the orthogonal component of the five signals in the test data set are extracted to generate a constellation diagram, which is input into the AlexNet model for rough identification classification, and then the differential high-order cumulants of the signals in the classification results are calculated, and input into the DNN model for fine identification classification.

[0174] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention described and shown in the drawings here can usually be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0175] The present invention adopts Matlab and Python software to implement the method, and the identification signal categories are BPSK, QPSK, OQPSK, PI4QPSK, and 8PSK.

[0176] See also Figure 2 and Figure 3 , are respectively the graphs of the probability of successful detection of the signal start and end positions as a function of the signal-to-noise ratio and the graph of the probability of missed detection as a function of the signal-to-noise ratio. Figure 2 It can be seen that the success probability of the start and end position detection increases with the increase of the signal-to-noise ratio, and the success probability can almost reach 1 at 0dB. Figure 3 It can be seen that the probability of missed detection of the signal decreases with the increase of the signal-to-noise ratio, and the probability of missed detection can almost reach 0 at 0dB. The effectiveness of the signal start and end position detection algorithm is verified.

[0177] See also Figure 4 , Figure 5, Figure 6 , Figure 7 and Figure 8 , respectively, are the constellation diagram changes before and after the frequency offset compensation of BPSK, QPSK, OQPSK, PI4QPSK, and 8PSK signals. Figure 4 , Figure 5 It can be seen that the signal constellation diagram after frequency offset compensation shows obvious BPSK and QPSK signal characteristics; Figure 6 , Figure 7 It can be seen that the signal constellation diagram after frequency offset compensation shows obvious OQPSK and PI4QPSK signal characteristics; Figure 8 Although the constellation diagrams before and after IF offset compensation are almost unchanged to the naked eye, they can be better identified using deep learning methods. Figure 8 Constellation diagram characteristics after compensation. It can be clearly seen that after frequency offset compensation, the five signals are well restored and all show easy-to-distinguish features, which verifies the effectiveness of the frequency offset compensation algorithm.

[0178] See also Fig. 9 , Fig.10 , Fig.11 , Fig.12 and Fig.13 , respectively, are the relationship between the recognition accuracy and the carrier-to-noise ratio of the BPSK, QPSK, OQPSK, PI4QPSK, and 8PSK signals after frequency offset compensation and the input of the neural network recognition model. Fig. 9 , Fig.10 , Fig.12 It can be seen from the figure that the recognition success rate of BPSK signal, QPSK and PI4QPSK after frequency offset compensation always remains 100% in the range of 4dB-12dB of carrier-to-noise ratio; Fig.11 It can be seen that the OQPSK signal recognition success rate fluctuates within the range of 4dB-12dB, but the average recognition success rate is still greater than 90%. Fig.13 It can be seen that the recognition success rate of 8PSK signal also fluctuates within the range of 4dB-12dB, but the recognition success rate under different carrier-to-noise ratios is greater than 90%, and the average recognition success rate is greater than 95%. It can be concluded that the neural network recognition model successfully identified five modulation signals, verifying the effectiveness of the modulation recognition algorithm based on deep learning.

[0179] In summary, the present invention provides a method and system for MPSK burst signal modulation recognition based on AlexNet network to complete the modulation mode recognition task of MPSK burst signal under non-cooperative communication, and finally realizes the demodulation bit error rate <10 for five burst signals including BPSK, QPSK, OQPSK, PI4QPSK and 8PSK. -2Under the condition of (corresponding BPSK signal carrier-to-noise ratio of 4dB, QPSK, OQPSK, PI4QPSK signal carrier-to-noise ratio of 7dB, 8PSK signal carrier-to-noise ratio of 12dB) the modulation type recognition accuracy reaches more than 90%.

[0180] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0181] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0182] Those skilled in the art will appreciate that the units and algorithms S of the various examples described in conjunction with the embodiments disclosed herein can be implemented in 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. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0183] In the embodiments provided by the present invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0184] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0185] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0186] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement S of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0187] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0188] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0189] These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operations S are performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 S of functions specified in one or more boxes.

[0190] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A method for identifying MPSK-type burst signal modulation based on AlexNet network, characterized in that: The following steps are involved: S1. The receiver obtains a predetermined signal and determines the start and end positions of the burst signal based on a delayed-addition double sliding window energy detection method; S2, segmenting the burst signal according to the start and end positions obtained in step S1, performing carrier frequency estimation and bandwidth estimation, and obtaining a complex baseband signal after bandpass filtering and down-conversion; S3, performing matched filtering on the characteristic parameters of the complex baseband signal obtained in step S2, and performing frequency offset estimation and frequency offset compensation on the complex baseband signal based on fast Fourier transform, and generating a training data set and a test data set in proportion; S4, extracting the in-phase component and the orthogonal component of the training data set signal obtained in step S3 to generate a constellation diagram, and designing a modulation mode recognition method based on the deep learning AlexNet network idea to train the AlexNet recognition model; S5, calculating the differential high-order cumulants of the signals in the training data set obtained in step S3, and designing a modulation mode recognition method based on the deep learning DNN network concept to train a DNN recognition model; S6. Extract the in-phase components and orthogonal components of the five signals in the test data set obtained in step S3 to generate a constellation diagram, input the AlexNet model obtained in step S4 for rough identification and classification, calculate the differential high-order cumulants of the signals in the classification results, and input the DNN model obtained in step S5 for fine identification and classification.

2. The MPSK-like burst signal modulation recognition method based on the AlexNet network according to claim 1, characterized in that: In step S1, the start and end positions of the burst signal are determined as follows: S101, convert the received signal sampling into a discrete sequence r(n), delay the discrete sequence r(n) by one sampling point to obtain r(n-1), perform instantaneous energy calculation on the data after adding the discrete sequences r(n) and r(n-1) to obtain the power and Z of the effective signal and noise in the received signal n ; S102, using the principle of double sliding windows to calculate the energy of the sampling points of windows A and B and A n and B n , calculate A n and B n The ratio is M n ; S103, calculate the M obtained in step S102 n The maximum value M nmax ; Set the reference threshold β, when M nmax >β indicates there is a signal; S104: After performing energy detection on the signal obtained in step S103 to obtain the signal starting point, use the reverse window energy ratio M nr =B n / A n Find M nr The minimum value gives the end position of the signal.

3. The MPSK-like burst signal modulation recognition method based on AlexNet network according to claim 1 is characterized in that: In step S2, the complex baseband signal is obtained as follows: S201, estimating the carrier frequency f of the signal r(n) according to the signal spectrum diagram c and bandwidth B w ; S202, processing the signal r(n) to obtain a complex baseband signal, as follows: Where h[n] is the unit impulse response of the bandpass filter, r(n) is the discrete sequence, and T is the sampling period.

4. The MPSK-type burst signal modulation recognition method based on AlexNet network according to claim 1 is characterized in that: In step S3, frequency offset estimation and frequency offset compensation are performed on the complex baseband signal based on fast Fourier transform as follows: S301, fixing parameters and performing matched filtering on the complex baseband signal; S302, performing nonlinear transformation on the complex baseband signal to remove initial phase information, and performing fast Fourier transform transformation; S303, searching for the maximum peak based on the spectrum diagram transformed by fast Fourier transform, where the frequency corresponding to the maximum peak is the frequency deviation of the complex baseband signal; S304: Perform frequency offset compensation on the complex baseband signal.

5. The MPSK-type burst signal modulation recognition method based on AlexNet network according to claim 4 is characterized in that: Step S304 is specifically as follows: S3041, determine the burst signal w n is additive Gaussian white noise, φ n is the initial phase, θ is the phase offset, and T is the symbol period; S3042, performing N-point Fourier transform on the signal to obtain a signal spectrum; S3043, Search for the highest peak in the signal spectrum | F k | max Get the corresponding k max , and then according to the maximum likelihood estimation theory, the estimated expression of the frequency offset is obtained as follows: ； S3044, the estimated frequency deviation is The compensated signal r'(n) is determined as follows: 。 6. The MPSK-type burst signal modulation recognition method based on AlexNet network according to claim 1 is characterized in that: In step S4, the modulation mode recognition method based on the deep learning AlexNet idea is specifically as follows: S401, establish an AlexNet network, and determine the loss function, learning rate, and update method of the AlexNet network; S402, input the constellation diagrams of five frequency offset compensated signals, namely BPSK, QPSK, OQPSK, PI4QPSK and 8PSK, into the AlexNet network, give classification labels and train the AlexNet recognition model.

7. The MPSK-type burst signal modulation recognition method based on AlexNet network according to claim 1 is characterized in that: In step S5, the modulation mode recognition method based on the deep learning DNN concept is specifically as follows: S501, establish a DNN network, determine the network's loss function, learning rate, and update method; S502, input the differential high-order cumulants of PI4QPSK and 8PSK into the DNN network, give classification labels and train the recognition model.

8. The MPSK-type burst signal modulation recognition method based on AlexNet network according to claim 7 is characterized in that: In step S502, the difference high-order cumulant is calculated as follows: S5021. Calculate the differential characteristics of the signal S5022, calculate the eighth-order difference high-order cumulant of the signal The specific calculation is as follows: Where N is the length of the signal, is the differential feature quantity, C 20 is the second-order difference cumulant, C 40 is the fourth-order difference cumulant, C 60 is the sixth-order difference cumulant.

9. The MPSK-type burst signal modulation recognition method based on AlexNet network according to claim 1 is characterized in that: In step S6, the classification based on the recognition model is specifically: S601, extracting the in-phase components and orthogonal components of the five signals in the test set to generate a constellation diagram, inputting the constellation diagram into the AlexNet recognition model for rough classification, and classifying the signals into four types: BPSK, QPSK, OQPSK, and 8PSK; S602, calculate the eighth-order differential high-order cumulants of all 8PSK signals in the coarse classification results, input them into the DNN recognition model for fine classification, and classify them into two types of signals: PI4QPSK and 8PSK.

10. An MPSK-type burst signal modulation recognition system based on AlexNet network, characterized in that: include: The receiving module, the receiver obtains the predetermined signal and determines the start and end positions of the burst signal based on the delayed addition double sliding window energy detection method; The processing module segments the burst signal according to the start and end positions obtained by the receiving module, performs carrier frequency estimation and bandwidth estimation, and obtains a complex baseband signal after bandpass filtering and down-conversion; The data module calculates and processes the characteristic parameters of the complex baseband signal obtained by the processing module for matched filtering, and performs frequency offset estimation and frequency offset compensation on the complex baseband signal based on fast Fourier transform, and generates a training data set and a test data set in proportion; AlexNet module, extracts the in-phase component and orthogonal component of the training data set signal obtained by the data module to generate a constellation diagram, and designs a modulation mode recognition method based on the deep learning AlexNet network idea to train the AlexNet recognition model; The DNN module calculates the differential high-order cumulants of the signals in the training data set obtained by the data module, and designs a modulation mode recognition method based on the deep learning DNN network concept to train the DNN recognition model; The recognition module extracts the in-phase components and orthogonal components of the five signals in the test data set obtained by the data module to generate a constellation diagram, inputs the AlexNet model obtained by the AlexNet module for rough recognition and classification, and then calculates the differential high-order cumulant of the signal in the classification result, and inputs the DNN model obtained by the DNN module for fine recognition and classification.

Images