Signal processing method, device, storage medium and electronic device

Abstract

The present disclosure relates to the field of signal processing, and in particular to a signal processing method, device, storage medium and electronic device, the signal processing method comprising: obtaining a constellation diagram of a signal to be processed and amplitude and phase characteristics of the signal to be processed; filtering the constellation diagram to obtain a filtered constellation diagram; inputting the filtered constellation diagram and the amplitude and phase characteristics into a signal modulation mode recognition model to obtain the modulation mode of the signal to be processed. The present disclosure can improve the accuracy of the signal modulation mode recognition result.

Description

Technical Field

[0001] The present disclosure relates to the field of signal processing technology, and in particular to a signal method, device, storage medium and electronic equipment. Background Art

[0002] With the continuous innovation and upgrading of communication technology, the modulation methods of communication signals are becoming more and more diverse. In non-cooperative communication, the receiving end of the communication system cannot know the specific modulation method of the transmitted signal, so automatic modulation recognition technology is needed to detect the modulation type of the signal. Automatic modulation recognition technology plays an important role in cognitive radio, electronic countermeasures, spectrum detection, etc.

[0003] At present, according to different discrimination methods, automatic modulation recognition methods include a likelihood function-based discrimination method and a feature-based adjustment recognition method.

[0004] Modulation recognition methods based on likelihood functions require a lot of calculations and a lot of prior information. In non-cooperative communications, a large number of signal parameters are unknown, and maximizing unknown information will lead to classification errors. Therefore, modulation recognition methods based on likelihood functions are difficult to implement in practical applications.

[0005] The feature-based adjustment recognition method is relatively easier to implement. It requires extracting features from the received signal and then using a classifier to determine the signal modulation method. However, feature extraction depends on experience, and the quality of extracted features is related to the selection of the classifier. Improper selection will seriously affect the accuracy of the recognition result.

[0006] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0007] The purpose of the present disclosure is to provide a signal processing method, device, storage medium and electronic device, thereby overcoming the problems of low data acquisition efficiency and low computing efficiency caused by the limitations and defects of related technologies at least to a certain extent.

[0008] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0009] According to a first aspect of the present disclosure, there is provided a signal processing method, comprising:

[0010] Acquire a constellation diagram of the signal to be processed and amplitude and phase characteristics of the signal to be processed;

[0011] Performing filtering on the constellation diagram to obtain a filtered constellation diagram;

[0012] The filtered constellation diagram and the amplitude and phase characteristics are input into a signal modulation mode recognition model to obtain the modulation mode of the signal to be processed.

[0013] In an exemplary embodiment of the present disclosure, acquiring the constellation diagram of the signal to be processed includes:

[0014] Determining a boundary value of the constellation diagram according to the real component of the signal to be processed and the imaginary component of the signal to be processed;

[0015] Normalizing the real component and the imaginary component according to the boundary value to obtain a normalized real component and a normalized imaginary component respectively;

[0016] Respectively converting the normalized real component and the normalized imaginary component into a preset range to obtain a converted real component and a converted imaginary component;

[0017] Rounding the converted real component and the converted imaginary component to obtain a rounded real component and a rounded imaginary component, respectively;

[0018] The constellation diagram is obtained according to the rounded real component and the rounded imaginary component.

[0019] In an exemplary embodiment of the present disclosure, determining the boundary value of the constellation diagram according to the real component of the signal to be processed and the imaginary component of the signal to be processed includes:

[0020] Obtaining a first maximum value of a real component and a second maximum value of an imaginary component of the signal to be processed;

[0021] taking the maximum value of the first maximum value and the second maximum value as the maximum value of the signal to be processed;

[0022] The boundary value of the constellation diagram is determined according to the maximum value of the signal to be processed.

[0023] In an exemplary embodiment of the present disclosure, the signal modulation mode recognition model includes a feature data processing network, a feature fusion network connected to the feature data processing network, a skip connection network connected to the feature fusion network, and a multilayer perceptron connected to the skip connection network;

[0024] The step of inputting the filtered constellation diagram and the amplitude and phase characteristics into a signal modulation mode recognition model to obtain the modulation mode of the signal to be processed includes:

[0025] Inputting the filtered constellation diagram and the amplitude-phase feature into a feature data processing network to obtain a first feature vector of the filtered constellation diagram and a second feature vector of the amplitude-phase feature;

[0026] Inputting the first vector and the second vector into the feature fusion network to obtain a feature fusion vector of the first feature vector and the second feature vector;

[0027] Inputting the feature fusion vector into the skip connection network to obtain a skip connection result of the feature fusion vector;

[0028] The skip connection result is input into the multilayer perceptron to obtain the modulation mode of the signal to be processed.

[0029] In an exemplary embodiment of the present disclosure, the feature data processing network includes a constellation embedding layer;

[0030] The step of inputting the filtered constellation diagram into a feature data processing network to obtain a first feature vector of the filtered constellation diagram comprises:

[0031] Inputting the filtered constellation diagram and the amplitude and phase features into the constellation diagram embedding layer, dividing the filtered constellation diagram to obtain a plurality of constellation sub-diagrams;

[0032] The plurality of constellation subgraphs are concatenated to generate the first eigenvector.

[0033] In an exemplary embodiment of the present disclosure, the feature data processing network includes an amplitude and phase feature embedding layer;

[0034] The step of inputting the amplitude and phase feature into a feature data processing network to obtain a second feature vector of the amplitude and phase feature comprises:

[0035] Dividing the amplitude-phase feature into equal lengths to obtain a plurality of sub-amplitude-phase features;

[0036] The multiple sub-amplitude-phase features are input into the amplitude-phase feature embedding layer, and the sub-amplitude-phase features are respectively reduced in dimension to obtain the second feature vectors of the sub-amplitude-phase features.

[0037] In an exemplary embodiment of the present disclosure, the feature fusion network includes a global average pooling layer;

[0038] The feature fusion network includes a global average pooling layer;

[0039] The step of inputting the first vector and the second vector into the feature fusion network to obtain a feature fusion vector of the first feature vector and the second feature vector includes:

[0040] Inputting the first vector and the second vector into the global average pooling layer to obtain an average value of the first vector and the second vector;

[0041] The average value is used as the feature fusion vector.

[0042] In an exemplary embodiment of the present disclosure, the feature fusion network includes a global average pooling layer;

[0043] The feature fusion network includes a global average pooling layer;

[0044] The step of inputting the first vector and the second vector into the feature fusion network to obtain a feature fusion vector of the first feature vector and the second feature vector includes:

[0045] Inputting the first vector and the second vector into the global average pooling layer to obtain an average value of the first vector and the second vector;

[0046] The average value is used as the feature fusion vector.

[0047] In an exemplary embodiment of the present disclosure, the step of inputting the first vector and the second vector into the feature fusion network to obtain a feature fusion vector of the first feature vector and the second feature vector includes:

[0048] Inputting the mark bit, the first vector and the second vector into the feature fusion network to obtain a splicing result of the mark bit, the first vector and the second vector;

[0049] The splicing result is used as the feature fusion vector.

[0050] In an exemplary embodiment of the present disclosure, inputting the feature fusion vector into the skip connection network to obtain a skip connection result of the feature fusion vector includes:

[0051] Adding a position tag to the feature fusion vector to obtain a target feature fusion vector;

[0052] Inputting the target feature fusion vector into the skip connection network to perform a layer normalization operation on the target feature fusion vector to obtain a first layer normalization result;

[0053] Performing autonomous mechanism operation on the first-layer standardized result to obtain an autonomous mechanism operation result;

[0054] The branch structure discarding method is used to perform jump connection on the autonomous mechanism operation result to obtain the jump connection result.

[0055] In an exemplary embodiment of the present disclosure, the step of inputting the skip connection result into the multilayer perceptron to obtain the modulation mode of the signal to be processed includes:

[0056] Performing a layer normalization operation on the skip connection result to obtain a second layer normalization result;

[0057] The second layer normalization result is input into the multilayer perceptron to obtain the modulation mode of the signal to be processed.

[0058] According to a second aspect of the present disclosure, there is provided a signal processing device, comprising:

[0059] A constellation diagram acquisition module, used to acquire the constellation diagram of the signal to be processed and the amplitude and phase characteristics of the signal to be processed;

[0060] A filtering processing module, used for filtering the constellation diagram to obtain a filtered constellation diagram;

[0061] The modulation mode acquisition module is used to input the filtered constellation diagram and the amplitude and phase characteristics into a signal modulation mode recognition model to obtain the modulation mode of the signal to be processed.

[0062] In an exemplary embodiment of the present disclosure, the constellation map acquisition module includes:

[0063] A boundary value determination unit, configured to determine a boundary value of the constellation diagram according to a real component of the signal to be processed and an imaginary component of the signal to be processed;

[0064] a normalization unit, configured to normalize the real component and the imaginary component according to the boundary value, to obtain a normalized real component and a normalized imaginary component respectively;

[0065] A conversion unit, used to convert the normalized real component and the normalized imaginary component into a preset range to obtain a converted real component and a converted imaginary component;

[0066] A rounding unit, used for rounding the converted real component and the converted imaginary component to obtain a rounded real component and a rounded imaginary component respectively;

[0067] A constellation diagram acquisition unit is used to obtain the constellation diagram according to the rounded real component and the rounded imaginary component.

[0068] In an exemplary embodiment of the present disclosure, the boundary value determining unit includes:

[0069] A first maximum value acquisition unit, used to acquire a first maximum value of the real component and a second maximum value of the imaginary component of the signal to be processed;

[0070] a second maximum value determining unit, configured to take the maximum value of the first maximum value and the second maximum value as the maximum value of the signal to be processed;

[0071] The boundary value determination subunit is used to determine the boundary value of the constellation diagram according to the maximum value of the signal to be processed.

[0072] In an exemplary embodiment of the present disclosure, the signal modulation mode recognition model includes a feature data processing network, a feature fusion network connected to the feature data processing network, a skip connection network connected to the feature fusion network, and a multilayer perceptron connected to the skip connection network;

[0073] The modulation mode acquisition module includes:

[0074] A feature vector acquisition unit, used for inputting the filtered constellation diagram and the amplitude and phase features into a feature data processing network to obtain a first feature vector of the filtered constellation diagram and a second feature vector of the amplitude and phase features;

[0075] A feature fusion vector acquisition unit, used for inputting the first vector and the second vector into the feature fusion network to acquire a feature fusion vector of the first feature vector and the second feature vector;

[0076] a skip connection result acquisition unit, configured to input the feature fusion vector into the skip connection network to acquire a skip connection result of the feature fusion vector;

[0077] The modulation mode acquisition unit is used to input the jump connection result into the multi-sensor to obtain the modulation mode of the signal to be processed.

[0078] In an exemplary embodiment of the present disclosure, the feature data processing network includes a constellation embedding layer;

[0079] The feature vector acquisition unit comprises:

[0080] A constellation sub-graph acquisition unit, configured to input the filtered constellation graph and the amplitude and phase features into the constellation graph embedding layer, and divide the filtered constellation graph to obtain a plurality of constellation sub-graphs;

[0081] The first feature vector generating unit is used to splice the multiple constellation sub-graphs to generate the first feature vector.

[0082] In an exemplary embodiment of the present disclosure, the feature data processing network includes an amplitude and phase feature embedding layer;

[0083] The feature vector acquisition unit comprises:

[0084] A sub-amplitude-phase feature acquisition unit, used for dividing the amplitude-phase feature into equal lengths to obtain a plurality of sub-amplitude-phase features;

[0085] The second feature vector acquisition unit is used to input the multiple sub-amplitude-phase features into the amplitude-phase feature embedding layer, and perform dimension reduction on the sub-amplitude-phase features respectively to obtain the second feature vector of the sub-amplitude-phase features.

[0086] In an exemplary embodiment of the present disclosure, the feature fusion network includes a global average pooling layer;

[0087] The feature fusion vector acquisition unit includes:

[0088] an average value obtaining unit, used for inputting the first vector and the second vector into the global average pooling layer to obtain an average value of the first vector and the second vector;

[0089] The first feature fusion vector acquisition unit is configured to use the average value as the feature fusion vector.

[0090] In an exemplary embodiment of the present disclosure, the feature fusion vector acquisition unit includes:

[0091] a splicing processing result obtaining unit, used for inputting the mark bit, the first vector and the second vector into the feature fusion network to perform splicing processing on the mark bit, the first vector and the second vector to obtain a splicing processing result;

[0092] The second feature fusion vector acquisition unit is used to use the splicing processing result as the feature fusion vector.

[0093] In an exemplary embodiment of the present disclosure, the skip connection result obtaining unit includes:

[0094] A target feature fusion vector acquisition unit, used for adding a position tag to the feature fusion vector to obtain a target feature fusion vector;

[0095] A first layer normalization result acquisition unit, used for inputting the target feature fusion vector into the skip connection network to perform a layer normalization operation on the target feature fusion vector to obtain a first layer normalization result;

[0096] An autonomous mechanism operation result acquisition unit, used for performing autonomous mechanism operation on the first layer standardized result to obtain an autonomous mechanism operation result;

[0097] The jump connection result acquisition unit is used to perform jump connection on the autonomous mechanism operation result by adopting a branch structure discarding method to obtain the jump connection result.

[0098] In an exemplary embodiment of the present disclosure, the modulation mode acquisition unit includes:

[0099] A second layer normalization result acquisition unit, configured to perform a layer normalization operation on the skip connection result to obtain a second layer normalization result;

[0100] The modulation mode acquisition subunit is used to input the second layer normalization result into the multilayer perceptron to obtain the modulation mode of the signal to be processed.

[0101] According to a third aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the steps of any one of the methods described in the first aspect are implemented.

[0102] According to a fourth aspect of the present disclosure, there is provided an electronic device, including:

[0103] Processor; and

[0104] A memory, configured to store executable instructions of the processor;

[0105] The processor is configured to perform the steps of any one of the methods described in the first aspect by executing the executable instructions.

[0106] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0107] In summary, the method provided by the present invention obtains the constellation diagram of the signal to be processed and the amplitude and phase characteristics of the signal to be processed; filters the constellation diagram to obtain a filtered constellation diagram; inputs the filtered constellation diagram and the amplitude and phase characteristics into a signal modulation mode recognition model to obtain the modulation mode of the signal to be processed. The modulation mode of the signal can be obtained without a large amount of calculation and a lot of prior information, and can be suitable for practical applications. Moreover, the modulation mode of the signal can be accurately obtained without relying on experience, thereby improving the accuracy of the recognition result.

[0108] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0109] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0110] Figure 1 A flow chart of a signal processing method in an exemplary embodiment of the present disclosure is schematically shown;

[0111] Figure 2 A schematic diagram schematically shows a signal processing process in an exemplary embodiment of the present disclosure;

[0112] Figure 3 A schematic diagram schematically shows a signal processing system architecture in an exemplary embodiment of the present disclosure;

[0113] FIG4 schematically shows a schematic diagram of a constellation diagram in an exemplary embodiment of the present disclosure;

[0114] FIG5 schematically shows a schematic diagram of a Gaussian kernel in an exemplary embodiment of the present disclosure;

[0115] FIG6 schematically shows a schematic diagram of a constellation diagram after filtering in an exemplary embodiment of the present disclosure;

[0116] Figure 7 The structure diagram of a modulation mode recognition model in an exemplary embodiment of the present disclosure is schematically shown;

[0117] Figure 8 A block diagram schematically shows a signal processing device in an exemplary embodiment of the present disclosure;

[0118] Fig. 9 A schematic diagram schematically shows a storage medium in an exemplary embodiment of the present disclosure;

[0119] Fig.10 A block diagram of an electronic device in an exemplary embodiment of the present disclosure is schematically shown.

[0120] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0121] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0122] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, device, apparatus, method or computer program product. Therefore, the present disclosure may be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0123] In view of the defects in the prior art, this exemplary embodiment first provides a signal processing method. Figure 1 As shown in , the above signal processing method may include the following steps:

[0124] S11, obtaining a constellation diagram of the signal to be processed and amplitude and phase characteristics of the signal to be processed;

[0125] S12, filtering the constellation diagram to obtain a filtered constellation diagram;

[0126] S13, inputting the filtered constellation diagram and the amplitude and phase characteristics into a signal modulation mode recognition model to obtain the modulation mode of the signal to be processed.

[0127] Figure 2 The following schematically shows a schematic diagram of a signal processing process in an exemplary embodiment of the present disclosure. Figure 2 As shown, the signal to be processed can be an in-phase quadrature (IQ) signal. The constellation diagram and amplitude-phase characteristics of the signal to be processed are obtained, and then the constellation diagram is filtered, and then the filtered constellation diagram and amplitude-phase characteristics are input into the signal modulation mode recognition model to obtain a recognition result, which is the modulation mode of the IQ signal.

[0128] To summarize, by obtaining the constellation diagram of the signal to be processed and the amplitude and phase characteristics of the signal to be processed; filtering the constellation diagram to obtain the filtered constellation diagram; inputting the filtered constellation diagram and the amplitude and phase characteristics into a signal modulation mode recognition model to obtain the modulation mode of the signal to be processed, the modulation mode of the signal can be obtained without a large amount of calculation and a lot of prior information, which can be suitable for practical applications, and the modulation mode of the signal can be accurately obtained without relying on experience, thereby improving the accuracy of the recognition result.

[0129] Below, each step of the signal modulation method in this example implementation will be described in more detail with reference to the accompanying drawings and embodiments.

[0130] In S11, a constellation diagram of the signal to be processed and amplitude and phase characteristics of the signal to be processed are obtained.

[0131] In an exemplary embodiment of the present disclosure, referring to Figure 3 The system architecture shown may include: a user-side mobile terminal device 301, a user-side smart terminal device 304, and a server 303. Data can be transmitted between the user-side mobile terminal device 301, the user-side smart terminal device 304, and the server 303 through the network 302. The network may include various connection types, such as wired communication links, wireless communication links, and the like. The above-mentioned signal processing method may be executed on the server side, the user-side terminal device, or by the user-side terminal device and the server side in collaboration. Taking the execution of the above-mentioned method on the server side as an example, the server may obtain the constellation diagram of the signal to be processed and the amplitude and phase characteristics of the signal to be processed; filter the constellation diagram to obtain the filtered constellation diagram; input the filtered constellation diagram and the amplitude and phase characteristics into the signal modulation mode recognition model to obtain the modulation mode of the signal to be processed.

[0132] In an exemplary embodiment of the present disclosure, the obtaining of the constellation diagram of the signal to be processed includes:

[0133] S111. Determine a boundary value of the constellation diagram according to the real component of the signal to be processed and the imaginary component of the signal to be processed.

[0134] In this step, the first maximum value of the real component and the second maximum value of the imaginary component of the signal to be processed are obtained; the maximum value of the first maximum value and the second maximum value is taken as the maximum value of the signal to be processed; and the boundary value of the constellation diagram is determined according to the maximum value of the signal to be processed.

[0135] For example, the signal to be processed is an IQ signal X = [x1, x2...x n ], the real component of the signal to be processed is Real(X), the imaginary component of the signal to be processed is Imag(X), the first maximum value is max[|Real(X)|], and the second maximum value is max[|Imag(X)|]. Further, the maximum value of the signal to be processed is as follows:

[0136] value max =max{max[|Real(X)|], max[|Imag(X)|]}; (1)

[0137] Among them, valuemax is the maximum value of the signal to be processed.

[0138] Further, after determining the maximum value of the signal to be processed, the boundary value of the constellation diagram is determined using the following formula:

[0139] board max =value max *(1+α); (2)

[0140] Among them, board max is the boundary value, and α is the boundary threshold factor.

[0141] S112. Normalize the real component and the imaginary component according to the boundary value to obtain a normalized real component and a normalized imaginary component respectively.

[0142] Specifically, the real part of the normalization is The imaginary part is normalized to

[0143] S113, respectively converting the normalized real component and the normalized imaginary component into a preset range to obtain a converted real component and a converted imaginary component.

[0144] In an exemplary embodiment of the present disclosure, the size of the constellation is preset to N*N. Then the normalized real component and the normalized imaginary component are respectively multiplied by half of the size of the constellation, that is, N / 2, and the normalized real component and the normalized imaginary component are converted to [-N / 2, N / 2].

[0145] Specifically, the converted real component and the converted imaginary component can be obtained by using the following formula:

[0146]

[0147] Among them, X r is the real part after transformation, X i is the imaginary component after transformation.

[0148] S114, rounding the converted real component and the converted imaginary component to obtain a rounded real component and a rounded imaginary component, respectively.

[0149] Further, after the normalized real component and the normalized imaginary component are converted to [-N / 2, N / 2], the converted real component and the converted imaginary component are rounded down to obtain rounded real component components. and the imaginary part after rounding

[0150] S115. Obtain the constellation diagram according to the rounded real component and the rounded imaginary component.

[0151] Further, we get the real part after rounding and the imaginary part after rounding Then, the real part and the imaginary part after rounding Add N / 2-1 to obtain the position information of the IQ signal in the constellation diagram. In an exemplary embodiment of the present disclosure, the constellation diagram obtained according to the rounded real component and the rounded imaginary component is as follows: Figure 4(a) to Figure 4(e) As shown. Among them, Figure 4 (a) shows the constellation diagram of the signal whose modulation mode is π / 2 binary phase shift keying (BPSK), Figure 4 (b) shows the constellation diagram of the signal whose modulation mode is quadrature phase shift keying (QPSK), Figure 4 (c) shows the constellation diagram of the signal whose modulation mode is 16 quadrature amplitude modulation (QAM), Figure 4 (d) shows the constellation diagram of the signal whose modulation mode is 64QAM, and Figure 4 (e) shows the constellation diagram of the signal whose modulation mode is 256QAM.

[0152] In an exemplary embodiment of the present disclosure, the amplitude and phase characteristics of the signal to be processed may be obtained by using the following formula:

[0153]

[0154]

[0155] Among them, amp(x i ) is the amplitude feature in the amplitude-phase feature, ang(x i )The phase characteristics in the amplitude-phase characteristics.

[0156] In step S12, the constellation diagram is filtered to obtain a filtered constellation diagram.

[0157] In an exemplary embodiment of the present disclosure, a Gaussian Blur algorithm may be used to perform smoothing filtering on the generated constellation diagram, and then fill in the blocks of empty data in the constellation point cluster, thereby increasing the number of samples.

[0158] Assume that the size of the constellation f(x, y) is P*Q, the size of the Gaussian filter G(x, y) is p*q, and the distribution of the Gaussian filter is as follows:

[0159]

[0160] The constellation diagram can be filtered using the following formula:

[0161]

[0162] in, Both p and q are odd integers.

[0163] In an exemplary embodiment of the present disclosure, two-dimensional convolution Gaussian kernels of sizes 3 and 5 can be selected respectively to perform smoothing filtering on the constellation diagram. The standard deviation of the Gaussian distribution is 1, and the shape of the Gaussian kernel is shown in Figures 5(a) and 5(b), where 5(a) is a schematic diagram of a Gaussian kernel of size 3*3, and 5(b) is a schematic diagram of a Gaussian kernel of size 5*5.

[0164] The constellation diagrams of the five modulation types after Gaussian kernel filtering are as follows Figure 6(a) to Figure 6(e) As shown, Figure 6(a) to Figure 6(e) The constellation diagram shown uses a 5*5 2D convolution Gaussian kernel for smoothing. Figure 6(a) to Figure 6(e) It can be seen that after smoothing filtering with a 5*5 two-dimensional convolution Gaussian kernel, the data distribution of each constellation point cluster is relatively consistent under each modulation mode. In each constellation point cluster, the empty data points near the cluster center are well supplemented, and the smoothness of the data distribution is also relatively good at the edge of the point cluster. Finally, the constellation diagram after Gaussian blur is normalized to a range between [0, 1] to obtain the final constellation diagram matrix. Specifically, Figure 6 (a) shows the constellation diagram of the signal with a modulation mode of π / 2BPSK after filtering, Figure 6 (b) shows the constellation diagram of the signal with a modulation mode of QPSK after filtering, Figure 6 (c) shows the constellation diagram of the signal with a modulation mode of 16QAM after filtering, Figure 6 (d) shows the constellation diagram of the signal with a modulation mode of 64QAM after filtering, and Figure 6 (e) shows the constellation diagram of the signal with a modulation mode of 256QAM after filtering.

[0165] By using the Gaussian Blur algorithm to smooth the generated constellation diagram, the ability of the low-resolution constellation diagram to represent the original data can be enhanced without expanding the number of sampling points of the data set.

[0166] In step S13, the filtered constellation diagram and the amplitude and phase characteristics are input into a signal modulation mode recognition model to obtain the modulation mode of the signal to be processed.

[0167] In an exemplary embodiment of the present disclosure, Figure 7 As shown, the signal modulation mode recognition model 70 includes a feature data processing network 71, a feature fusion network 72 connected to the feature data processing network 71, a skip connection network 73 connected to the feature fusion network 72, and a multilayer perceptron 75 connected to the skip connection network 74;

[0168] Based on the above content, in an exemplary embodiment of the present disclosure, the above-mentioned inputting the filtered constellation diagram and the amplitude and phase characteristics into the signal modulation mode recognition model to obtain the modulation mode of the signal to be processed includes:

[0169] S131, inputting the filtered constellation diagram and the amplitude-phase feature into a feature data processing network to obtain a first feature vector of the filtered constellation diagram and a second feature vector of the amplitude-phase feature;

[0170] In an exemplary embodiment of the present disclosure, Figure 7 As shown, the feature data processing network 71 includes a constellation diagram embedding layer 711;

[0171] The step of inputting the filtered constellation diagram into a feature data processing network to obtain a first feature vector of the filtered constellation diagram comprises:

[0172] The filtered constellation diagram and the amplitude and phase features are input into the constellation diagram embedding layer, the filtered constellation diagram is divided to obtain a plurality of constellation sub-diagrams; and the plurality of constellation sub-diagrams are spliced ​​to generate the first feature vector.

[0173] In an exemplary embodiment of the present disclosure, the signal modulation mode recognition model is a transformer model. Since the input of the traditional transformer model is a two-dimensional vector sequence, including the length and dimension [R, D] of the vector, and the constellation diagram is a three-dimensional matrix with dimensions [W, H, C], where W, H, and C represent the width, height, and number of channels of the constellation diagram, respectively. Therefore, a constellation diagram embedding layer is required to transform the constellation diagram in form and convert the constellation diagram into a two-dimensional vector sequence. First, divide the constellation diagram into R blocks of equal size, and then expand it into a one-dimensional vector. Assuming the size of the block is P, the size of each vector is P 2 ×C, the constellation diagram changes from a W×H×C image to R*(P 2 ×C), R is the sequence length of the vector of the constellation diagram input signal modulation mode recognition model.

[0174] In an exemplary embodiment of the present disclosure, the feature data processing network 71 includes an amplitude and phase feature embedding layer 712;

[0175] Based on the above content, in an exemplary embodiment of the present disclosure, the inputting the amplitude and phase feature into a feature data processing network to obtain a second feature vector of the amplitude and phase feature includes:

[0176] The amplitude-phase feature is divided into equal lengths to obtain a plurality of sub-amplitude-phase features; the plurality of sub-amplitude-phase features are input into the amplitude-phase feature embedding layer, and the sub-amplitude-phase features are respectively reduced in dimension to obtain a second feature vector of the sub-amplitude-phase feature.

[0177] In an exemplary embodiment of the present disclosure, the amplitude and phase characteristics are the characteristics of the signal to be processed in the time domain. The dimension of the amplitude and phase characteristics in the time domain is [E, D], where E represents the length of the amplitude and phase characteristic data frame, and D is equal to 2, which respectively represent the amplitude characteristic and phase characteristic in the amplitude and phase characteristics. However, in the communication system, the length E of the amplitude and phase characteristic data frame is usually relatively large, and the long sequence will lead to an increase in the computational complexity when the subsequent signal modulation mode recognition model is operated. Therefore, it is necessary to divide the amplitude and phase characteristics to reduce the sequence length.

[0178] In an exemplary embodiment of the present disclosure, the amplitude-phase feature can be divided into L sub-amplitude-phase features by equal length, where L is an integer greater than 1. Each sub-amplitude-phase feature is then input into an amplitude-phase feature embedding layer to perform embedding dimensionality reduction operations on the sub-amplitude-phase features to obtain the second feature vector.

[0179] By dividing the amplitude and phase features into equal lengths, the sequence length of the amplitude and phase features input into the modulation mode recognition model can be effectively reduced, thereby reducing the complexity of the calculation and improving the calculation efficiency.

[0180] S132, inputting the first vector and the second vector into the feature fusion network to obtain a feature fusion vector of the first feature vector and the second feature vector;

[0181] The feature fusion network 72 includes a global average pooling layer 721;

[0182] The step of inputting the first vector and the second vector into the feature fusion network to obtain a feature fusion vector of the first feature vector and the second feature vector includes:

[0183] The first vector and the second vector are input into the global average pooling layer to obtain an average value of the first vector and the second vector; and the average value is used as the feature fusion vector.

[0184] In an exemplary embodiment of the present disclosure, obtaining an average value of the first vector and the second vector includes:

[0185] Get each size P in the first vector 2 ×C vectors, and is the average value of each sub-amplitude phase feature after dimensionality reduction in the second eigenvector.

[0186] Through the global average pooling layer, each size is P 2 ×C vectors and the sub-amplitude and phase features after dimensionality reduction are averaged so that each size is P 2 ×C vector and each dimension-reduced sub-amplitude phase feature output a corresponding neural node, thereby making each size P 2 The connection between the vector of ×C and the sub-amplitude and phase features after dimensionality reduction is more intuitive, which is conducive to the feature fusion of the model. Secondly, it makes it easier to convert the feature fusion vector into classification probability data, which is convenient for classifying and identifying the probability distribution corresponding to different types of modulation methods from the feature fusion vector.

[0187] In another embodiment of the present disclosure, the step of inputting the first vector and the second vector into the feature fusion network to obtain a feature fusion vector of the first feature vector and the second feature vector includes:

[0188] The mark bit, the first vector and the second vector are input into the feature fusion network to perform splicing processing on the mark bit, the first vector and the second vector to obtain a splicing processing result; and the splicing processing result is used as the feature fusion vector.

[0189] In an exemplary embodiment of the present disclosure, the tag is a Class tag, which is a learnable embedding vector. In the self-attention operation mechanism, the tag will be associated with each of the first feature vector and the second feature vector, each of which has a size of P 2 ×C vector and each sub-amplitude feature are used for attention operation, so that the marker contains each size P 2 ×C vector and the characteristic information of each sub-amplitude phase feature, and the size of each marker bit information is P 2 The information content of the vector of ×C and each sub-amplitude phase feature is relatively balanced. Therefore, it is only necessary to extract the marker information from the feature fusion vector to classify and identify according to the marker, and obtain the debugging method of the signal to be processed, which greatly reduces the amount of data processing and improves the operation efficiency.

[0190] Since each sub-amplitude phase feature and each block's constellation map feature have their own corresponding outputs after passing through the skip connection network, the hidden state information after the self-attention mechanism is effectively fused and classified in the classification task, and the modulation mode of the signal to be processed can be accurately identified. Therefore, by using global average pooling and adding classification markers to perform feature fusion, the feature fusion problem can be effectively solved, and then the modulation mode of the signal to be processed can be accurately identified according to the feature fusion vector.

[0191] S133, inputting the feature fusion vector into the skip connection network to obtain a skip connection result of the feature fusion vector;

[0192] Based on the above content, in an exemplary embodiment of the present disclosure, the step of inputting the feature fusion vector into the skip connection network to obtain the skip connection result of the feature fusion vector includes:

[0193] A position label is added to the feature fusion vector to obtain a target feature fusion vector; the target feature fusion vector is input into the jump connection network to perform a layer standardization operation on the target feature fusion vector to obtain a first layer standardization result; an autonomous mechanism operation is performed on the first layer standardization result to obtain an autonomous mechanism operation result; and a branch structure discarding method is used to perform a jump connection on the autonomous mechanism operation result to obtain the jump connection result.

[0194] Since the amplitude-phase feature and the constellation diagram are divided, the sub-amplitude-phase feature obtained after the division and each size of P in the first vector 2 ×C vector loses its own temporal position and spatial position correlation. In order to increase the amplitude and phase characteristics of each sub-vector and the size of each P in the first vector 2 ×C in the original data, and adding a position label to the feature fusion vector to reflect the spatiotemporal characteristics of the data. In an exemplary embodiment of the present disclosure, the position label can be a supervised training position label, so that the position label can correct the position information of the data.

[0195] Furthermore, after adding a position tag to the feature fusion vector to obtain a target feature fusion vector, the target feature fusion vector is input into the skip connection network. In an exemplary embodiment of the present disclosure, Figure 7As shown, the skip connection network includes multiple cascaded skip connection sub-networks. Each of the skip connection sub-networks includes a layer normalization layer, a fully connected layer with 512 neurons connected to the layer normalization layer, a random inactivation Dropout layer connected to the fully connected layer, and a branch structure discarding module connected to the Dropout layer.

[0196] Specifically, the Layer Normalization layer of the first jump connection subnetwork performs layer normalization on the data of each patch to obtain the first layer normalization result. The fully connected layer and the Dropout layer connected to the fully connected layer perform self-attention operation on the first layer normalization result to obtain the autonomous mechanism operation result. In order to enhance the generalization ability of the model, the branch structure discarding module uses the branch structure discarding method to perform jump connection on the autonomous mechanism operation result to obtain the first jump connection result.

[0197] Furthermore, in order to further enhance the generalization ability of the model, the first skip connection result can be input into the remaining cascaded skip connection sub-networks to perform skip connection on the first skip connection result.

[0198] S134: Input the skip connection result into the multilayer perceptron to obtain a modulation mode of the signal to be processed.

[0199] Based on the above content, in an exemplary embodiment of the present disclosure, the above-mentioned inputting the jump connection result into the multilayer perceptron to obtain the modulation mode of the signal to be processed includes: performing a layer normalization operation on the jump connection result to obtain a second layer normalization result; inputting the second layer normalization result into the multilayer perceptron to obtain the modulation mode of the signal to be processed.

[0200] After the jump connection result is obtained through the cascaded jump connection subnetwork, the multilayer perceptron is used to extract useful classification information from the jump connection result, and the classification information is used to indicate the modulation mode of the signal to be processed. Specifically, the jump leveling result is first subjected to a layer normalization operation through a Layer Normalization layer to obtain a second layer of normalized result; the second layer of normalized result is then reduced in dimension through a small multilayer perceptron and converted into a probabilistic classification method, that is, the output of the multilayer perceptron is different categories of modulation modes and their corresponding probabilities. Among them, the modulation mode with the highest probability is the modulation mode of the signal to be processed. In an exemplary embodiment of the present disclosure, a cascade of fully connected layers containing a softmax activation function can be used in a multilayer perceptron to achieve classification of modulation modes.

[0201] The following is an explanation of how the signal processing method provided by the present disclosure obtains the modulation mode of the signal to be processed in combination with the calculation formula:

[0202] Denote the amplitude-phase characteristic as x, and x∈R (2048，2) , the matrix corresponding to the constellation diagram is represented by y∈R (2048，2) , then the first eigenvector and the second eigenvector can be calculated using the following formula:

[0203] x p =LSTM[Conv1D(x)]; (8)

[0204] y p =Conv2D(y); (9)

[0205]

[0206] Among them, x p represents the second eigenvector, y p represents the first eigenvector, z p Represents the concatenation result of the first eigenvector and the second eigenvector.

[0207] Using the global average pooling method, the feature fusion vector can be calculated using the following formula:

[0208] z0=z p +E pos , E pos ∈R (4*2)×128 ; (11)

[0209] Among them, z0 represents the feature fusion vector, E pos Represents a location tag.

[0210] Using the class marker method, the feature fusion vector can be obtained by the following formula:

[0211] z0=[x class ; z p ]+E pos , x class ∈R 1×128 , E pos ∈R (4*2+1)×128 ; (12)

[0212] Among them, x class Indicates the mark bit.

[0213] After entering the cascaded skip connection sub-network, the skip connection result is calculated using the following formula:

[0214] z' l=Patch_Drop(MSA(LN(z l-1 )))+z l-1 ; (13)

[0215] z l =Patch_Drop(Drop(FC(Drop(FC(z′ l )))))+z′ l ; (14)

[0216] Among them, z l represents the skip connection result, l∈[1,M]. Where M is an integer greater than 1, representing the number of cascaded skip connection sub-networks in the skip connection network.

[0217] After entering the multi-layer perceptron, the skip connection results corresponding to the global average pooling method and the skip connection results corresponding to the class tag method are calculated using formula (15) and formula (16) respectively:

[0218] y=FC(GAP(z l )); (15)

[0219] y=FC(LN(z l )); (16)

[0220] Among them, GAP represents weighted average operation, LN represents logarithmic operation, and .y represents the probability corresponding to different types of modulation modes.

[0221] After introducing the signal processing method according to the exemplary embodiment of the present invention, Figure 8 A signal processing device according to an exemplary embodiment of the present invention is described.

[0222] refer to Figure 8 As shown, the signal processing device 80 of the exemplary embodiment of the present invention may include: a constellation diagram acquisition module 801, a filtering processing module 802 and a modulation mode acquisition module 803; wherein:

[0223] A constellation diagram acquisition module 801 is used to acquire the constellation diagram of the signal to be processed and the amplitude and phase characteristics of the signal to be processed;

[0224] A filtering processing module 802 is used to filter the constellation diagram to obtain a filtered constellation diagram;

[0225] The modulation mode acquisition module 803 is used to input the filtered constellation diagram and the amplitude and phase characteristics into a signal modulation mode recognition model to obtain the modulation mode of the signal to be processed.

[0226] In an exemplary embodiment of the present disclosure, the constellation map acquisition module includes:

[0227] A boundary value determination unit, configured to determine a boundary value of the constellation diagram according to a real component of the signal to be processed and an imaginary component of the signal to be processed;

[0228] a normalization unit, configured to normalize the real component and the imaginary component according to the boundary value, to obtain a normalized real component and a normalized imaginary component respectively;

[0229] A conversion unit, used to convert the normalized real component and the normalized imaginary component into a preset range to obtain a converted real component and a converted imaginary component;

[0230] A rounding unit, used for rounding the converted real component and the converted imaginary component to obtain a rounded real component and a rounded imaginary component respectively;

[0231] A constellation diagram acquisition unit is used to obtain the constellation diagram according to the rounded real component and the rounded imaginary component.

[0232] In an exemplary embodiment of the present disclosure, the boundary value determining unit includes:

[0233] A first maximum value acquisition unit, used to acquire a first maximum value of the real component and a second maximum value of the imaginary component of the signal to be processed;

[0234] a second maximum value determining unit, configured to take the maximum value of the first maximum value and the second maximum value as the maximum value of the signal to be processed;

[0235] The boundary value determination subunit is used to determine the boundary value of the constellation diagram according to the maximum value of the signal to be processed.

[0236] In an exemplary embodiment of the present disclosure, the signal modulation mode recognition model includes a feature data processing network, a feature fusion network connected to the feature data processing network, a skip connection network connected to the feature fusion network, and a multilayer perceptron connected to the skip connection network;

[0237] The modulation mode acquisition module includes:

[0238] A feature vector acquisition unit, used for inputting the filtered constellation diagram and the amplitude and phase features into a feature data processing network to obtain a first feature vector of the filtered constellation diagram and a second feature vector of the amplitude and phase features;

[0239] A feature fusion vector acquisition unit, used for inputting the first vector and the second vector into the feature fusion network to acquire a feature fusion vector of the first feature vector and the second feature vector;

[0240] a skip connection result acquisition unit, configured to input the feature fusion vector into the skip connection network to acquire a skip connection result of the feature fusion vector;

[0241] The modulation mode acquisition unit is used to input the jump connection result into the multi-sensor to obtain the modulation mode of the signal to be processed.

[0242] In an exemplary embodiment of the present disclosure, the feature data processing network includes a constellation embedding layer;

[0243] The feature vector acquisition unit comprises:

[0244] A constellation sub-graph acquisition unit, configured to input the filtered constellation graph and the amplitude and phase features into the constellation graph embedding layer, and divide the filtered constellation graph to obtain a plurality of constellation sub-graphs;

[0245] The first feature vector generating unit is used to splice the multiple constellation sub-graphs to generate the first feature vector.

[0246] In an exemplary embodiment of the present disclosure, the feature data processing network includes an amplitude and phase feature embedding layer;

[0247] The feature vector acquisition unit comprises:

[0248] A sub-amplitude-phase feature acquisition unit, used for dividing the amplitude-phase feature into equal lengths to obtain a plurality of sub-amplitude-phase features;

[0249] The second feature vector acquisition unit is used to input the multiple sub-amplitude-phase features into the amplitude-phase feature embedding layer, and perform dimension reduction on the sub-amplitude-phase features respectively to obtain the second feature vector of the sub-amplitude-phase features.

[0250] In an exemplary embodiment of the present disclosure, the feature fusion network includes a global average pooling layer;

[0251] The feature fusion vector acquisition unit includes:

[0252] an average value obtaining unit, used for inputting the first vector and the second vector into the global average pooling layer to obtain an average value of the first vector and the second vector;

[0253] The first feature fusion vector acquisition unit is configured to use the average value as the feature fusion vector.

[0254] In an exemplary embodiment of the present disclosure, the feature fusion vector acquisition unit includes:

[0255] a splicing processing result obtaining unit, used for inputting the mark bit, the first vector and the second vector into the feature fusion network to perform splicing processing on the mark bit, the first vector and the second vector to obtain a splicing processing result;

[0256] The second feature fusion vector acquisition unit is used to use the splicing processing result as the feature fusion vector.

[0257] In an exemplary embodiment of the present disclosure, the skip connection result obtaining unit includes:

[0258] A target feature fusion vector acquisition unit, used for adding a position tag to the feature fusion vector to obtain a target feature fusion vector;

[0259] A first layer normalization result acquisition unit, used for inputting the target feature fusion vector into the skip connection network to perform a layer normalization operation on the target feature fusion vector to obtain a first layer normalization result;

[0260] An autonomous mechanism operation result acquisition unit, used for performing autonomous mechanism operation on the first layer standardized result to obtain an autonomous mechanism operation result;

[0261] The jump connection result acquisition unit is used to perform jump connection on the autonomous mechanism operation result by adopting a branch structure discarding method to obtain the jump connection result.

[0262] In an exemplary embodiment of the present disclosure, the modulation mode acquisition unit includes:

[0263] A second layer normalization result acquisition unit, configured to perform a layer normalization operation on the skip connection result to obtain a second layer normalization result;

[0264] The modulation mode acquisition subunit is used to input the second layer normalization result into the multilayer perceptron to obtain the modulation mode of the signal to be processed.

[0265] Since the functional modules of the signal processing device in the embodiment of the present invention are the same as those in the above-mentioned signal processing method in the embodiment of the present invention, they are not described in detail here.

[0266] After introducing the signal processing method and the signal processing device according to the exemplary embodiment of the present invention, Fig. 9 A storage medium according to an exemplary embodiment of the present invention is described.

[0267] refer to Fig. 9As shown, a program product 900 for implementing the above method according to an embodiment of the present invention is described, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a device, such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0268] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0269] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0270] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0271] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user computing device, partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0272] After introducing the storage medium of the exemplary embodiment of the present invention, next, reference is made to Fig.10 An electronic device according to an exemplary embodiment of the present invention is described.

[0273] Fig.10 The electronic device 100 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0274] like Fig.10 As shown, the electronic device 100 is in the form of a general computing device. The components of the electronic device 90 may include, but are not limited to: the at least one processing unit 1010, the at least one storage unit 1020, a bus 1030 connecting different system components (including the storage unit 1020 and the processing unit 1010), and a display unit 1040.

[0275] The storage unit stores program codes, which can be executed by the processing unit 1010, so that the processing unit 1010 performs the steps of various exemplary embodiments of the present invention described in the above "Exemplary Method" section of this specification. For example, the processing unit 1010 can perform the following steps: Figure 1 Steps S11 to S13 shown in FIG.

[0276] The storage unit 1020 may include a volatile storage unit, such as a random access storage unit (RAM) 10201 and / or a cache storage unit 10202, and may further include a read-only storage unit (ROM) 10203. The storage unit 1020 may also include a program / utility 10204 having a set (at least one) of program modules 10205, such program modules 10205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination thereof may include the implementation of a network environment.

[0277] The bus 1030 may include a data bus, an address bus, and a control bus.

[0278] The electronic device 100 may also communicate with one or more external devices 110 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), and such communication may be performed through an input / output (I / O) interface 1050. The electronic device 100 also includes a display unit 1040, which is connected to the input / output (I / O) interface 1050 for display. In addition, the electronic device 100 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) through a network adapter 960. As shown, the network adapter 1060 communicates with other modules of the electronic device 100 through a bus 1030. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 100, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0279] It should be noted that although several modules or submodules of the spatial data rendering system are mentioned in the above detailed description, such division is merely exemplary and not mandatory. In fact, according to an embodiment of the present invention, the features and functions of two or more units / modules described above may be embodied in one unit / module. Conversely, the features and functions of one unit / module described above may be further divided into multiple units / modules for embodiment.

[0280] In addition, although the operations of the method of the present invention are described in a specific order in the drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0281] Although the spirit and principle of the present invention have been described with reference to several specific embodiments, it should be understood that the present invention is not limited to the disclosed specific embodiments, and the division of various aspects does not mean that the features in these aspects cannot be combined to benefit, and such division is only for the convenience of expression. The present invention is intended to cover various modifications and equivalent arrangements included in the spirit and scope of the attached claims.

Claims

1. A signal processing method, characterized in that: include: Obtain a constellation diagram of a signal to be processed and amplitude and phase characteristics of the signal to be processed; wherein the constellation diagram is obtained in the following manner: determine a boundary value of the constellation diagram according to a real component of the signal to be processed and an imaginary component of the signal to be processed; normalize the real component and the imaginary component according to the boundary value to obtain a normalized real component and a normalized imaginary component, respectively; convert the normalized real component and the normalized imaginary component to a preset interval range to obtain a converted real component and a converted imaginary component; round the converted real component and the converted imaginary component to obtain a rounded real component and a rounded imaginary component, respectively; obtain the constellation diagram according to the rounded real component and the rounded imaginary component; Performing smoothing filtering on the constellation diagram based on a preset two-dimensional convolution Gaussian kernel, and filling blocks of empty data in constellation point clusters in the constellation diagram to obtain a filtered constellation diagram; The filtered constellation diagram and the amplitude and phase characteristics are input into a signal modulation mode recognition model to obtain the modulation mode of the signal to be processed, including: inputting the filtered constellation diagram and the amplitude and phase characteristics into a feature data processing network in the signal modulation mode recognition model to obtain a first feature vector of the filtered constellation diagram and a second feature vector of the amplitude and phase characteristics; inputting the first feature vector and the second feature vector into a feature fusion network in the signal modulation mode recognition model to obtain a feature fusion vector of the first feature vector and the second feature vector; inputting the feature fusion vector into a jump connection network in the signal modulation mode recognition model to obtain a jump connection result of the feature fusion vector; and inputting the jump connection result into a multilayer perceptron in the signal modulation mode recognition model to obtain the modulation mode of the signal to be processed.

2. The method according to claim 1, characterized in that Determining the boundary value of the constellation diagram according to the real component of the signal to be processed and the imaginary component of the signal to be processed includes: Obtaining a first maximum value of a real component and a second maximum value of an imaginary component of the signal to be processed; taking the maximum value of the first maximum value and the second maximum value as the maximum value of the signal to be processed; The boundary value of the constellation diagram is determined according to the maximum value of the signal to be processed.

3. The method according to claim 1, characterized in that The feature data processing network includes a constellation embedding layer; The step of inputting the filtered constellation diagram into a feature data processing network to obtain a first feature vector of the filtered constellation diagram comprises: Inputting the filtered constellation diagram and the amplitude and phase features into the constellation diagram embedding layer, dividing the filtered constellation diagram to obtain a plurality of constellation sub-diagrams; The plurality of constellation subgraphs are concatenated to generate the first eigenvector.

4. The method according to claim 1, characterized in that: The feature data processing network includes an amplitude and phase feature embedding layer; The step of inputting the amplitude and phase feature into a feature data processing network to obtain a second feature vector of the amplitude and phase feature comprises: Dividing the amplitude-phase feature into equal lengths to obtain a plurality of sub-amplitude-phase features; The multiple sub-amplitude-phase features are input into the amplitude-phase feature embedding layer, and the sub-amplitude-phase features are respectively reduced in dimension to obtain second feature vectors of the sub-amplitude-phase features.

5. The method according to claim 1, characterized in that: The feature fusion network includes a global average pooling layer; The step of inputting the first feature vector and the second feature vector into the feature fusion network to obtain a feature fusion vector of the first feature vector and the second feature vector includes: Inputting the first feature vector and the second feature vector into the global average pooling layer to obtain an average value of the first feature vector and the second feature vector; The average value is used as the feature fusion vector.

6. The method according to claim 5, characterized in that The step of inputting the first feature vector and the second feature vector into the feature fusion network to obtain a feature fusion vector of the first feature vector and the second feature vector includes: Inputting the mark bit, the first feature vector and the second feature vector into the feature fusion network to perform splicing processing on the mark bit, the first feature vector and the second feature vector to obtain a splicing processing result; The splicing processing result is used as the feature fusion vector.

7. The method according to claim 5, characterized in that The step of inputting the feature fusion vector into the skip connection network to obtain the skip connection result of the feature fusion vector comprises: Adding a position tag to the feature fusion vector to obtain a target feature fusion vector; Inputting the target feature fusion vector into the skip connection network to perform a layer normalization operation on the target feature fusion vector to obtain a first layer normalization result; Performing autonomous mechanism operation on the first-layer standardized result to obtain an autonomous mechanism operation result; The branch structure discarding method is used to perform jump connection on the autonomous mechanism operation result to obtain the jump connection result.

8. The method according to claim 7, characterized in that The step of inputting the skip connection result into the multilayer perceptron to obtain the modulation mode of the signal to be processed includes: Performing a layer normalization operation on the skip connection result to obtain a second layer normalization result; The second layer normalization result is input into the multilayer perceptron to obtain the modulation mode of the signal to be processed.

9. A signal processing device, characterized in that: include: A constellation diagram acquisition module, used to acquire a constellation diagram of a signal to be processed and amplitude and phase characteristics of the signal to be processed; wherein the constellation diagram is obtained in the following manner: determining a boundary value of the constellation diagram according to a real component of the signal to be processed and an imaginary component of the signal to be processed; normalizing the real component and the imaginary component according to the boundary value to obtain a normalized real component and a normalized imaginary component respectively; converting the normalized real component and the normalized imaginary component to a preset interval range to obtain a converted real component and a converted imaginary component respectively; rounding the converted real component and the converted imaginary component to obtain a rounded real component and a rounded imaginary component respectively; obtaining the constellation diagram according to the rounded real component and the rounded imaginary component; A filtering processing module, used for performing smoothing filtering on the constellation diagram based on a preset two-dimensional convolution Gaussian kernel, and filling blocks of empty data in the constellation point clusters in the constellation diagram to obtain a filtered constellation diagram; A modulation mode acquisition module is used to input the filtered constellation diagram and the amplitude and phase characteristics into a signal modulation mode recognition model to obtain the modulation mode of the signal to be processed, including: inputting the filtered constellation diagram and the amplitude and phase characteristics into a feature data processing network in the signal modulation mode recognition model to obtain a first feature vector of the filtered constellation diagram and a second feature vector of the amplitude and phase characteristics; inputting the first feature vector and the second feature vector into a feature fusion network in the signal modulation mode recognition model to obtain a feature fusion vector of the first feature vector and the second feature vector; inputting the feature fusion vector into a jump connection network in the signal modulation mode recognition model to obtain a jump connection result of the feature fusion vector; and inputting the jump connection result into a multilayer perceptron in the signal modulation mode recognition model to obtain the modulation mode of the signal to be processed.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

11. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; The processor is configured to perform the steps of the method according to any one of claims 1 to 8 by executing the executable instructions.

Images