Signal identification and demodulation method and system

Abstract

The invention provides a signal identification and demodulation method and system. The method comprises the following steps: receiving a target signal; dividing the target signal into signal segments with fixed lengths in a time domain, and inserting modulation classification tags into signal tokens of the signal segments; inputting the generated signal token sequence into a signal identification demodulation model, processing the signal token sequence by a complex value encoder to obtain modulation type features and extract signal modulation features, obtaining probability distribution of modulation types through a full connection layer, and determining the modulation type corresponding to the maximum probability value as the modulation type of a target signal; and the complex value decoder receives the signal modulation characteristics and demodulates the signal modulation characteristics in a time domain to obtain identified and demodulated information data. According to the method, the modulation type identification and demodulation functions are integrated into the unified complex domain Transform architecture model, so that low efficiency and performance limitation caused by separated design are overcome, complex parameter conditions such as different sampling rates can be flexibly coped with, and the universality and robustness of the system are enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication and deep learning, and in particular relates to a signal recognition and demodulation method and system. Background Art

[0002] At present, in wireless communication and signal processing applications, due to the continuous increase in data transmission rate and signal complexity, traditional signal recognition and demodulation technologies are facing severe challenges. In particular, for non-ideal propagation environments or high dynamic conditions, how to accurately identify signal types and perform efficient demodulation has become a key technical issue.

[0003] In the prior art, signal recognition and demodulation are usually divided into two independent processes, that is, signal recognition and demodulation are separated. Specifically, traditional solutions often recognize signal types based on feature extraction and classification algorithms, and then select the corresponding demodulator according to the recognition results in the demodulation stage. This undoubtedly increases the complexity and delay of the system. In addition, traditional demodulators must know the signal modulation parameters in advance and require fixed sampling rate and symbol period ratio. SPS (Samples per Symbol) refers to the number of samples contained in each symbol in a digital communication system.

[0004] In summary, although the existing technology can identify the signal type and demodulate, the separation design will result in low efficiency of signal identification and demodulation. Moreover, for signal modulation parameters with different sampling rates such as wireless communication, radar monitoring and electronic warfare, it is difficult to achieve a fixed sampling rate and symbol period ratio, that is, the existing technology is only applicable to situations where the signal modulation parameters are fixed, such as cooperative communication scenarios, and its performance is limited when applied in non-cooperative communication scenarios. Summary of the invention

[0005] The embodiments of the present invention provide a signal identification and demodulation method and system, aiming to solve the problems of low efficiency and limited performance caused by independent processing of signal identification and demodulation in the prior art.

[0006] The present invention provides a signal identification and demodulation method, the method comprising:

[0007] Receiving a target signal transmitted by a wireless channel, wherein the target signal is a wireless signal to be identified and demodulated;

[0008] Dividing the target signal into signal segments of fixed length in the time domain, and inserting a modulation classification label into the signal token header of each signal segment to generate a signal token sequence;

[0009] The signal token sequence is input into a pre-built signal recognition and demodulation model, wherein the signal recognition and demodulation model is a model built based on a complex domain Transformer architecture, and a complex-valued encoder in the complex domain Transformer architecture processes the input signal token sequence to obtain the modulation type feature and extract the signal modulation feature of the target signal in the complex domain;

[0010] The modulation type feature is used to calculate the probability distribution of the modulation type through the fully connected layer in the complex domain Transformer architecture, and the modulation type corresponding to the maximum probability value is determined as the modulation type of the target signal;

[0011] The complex-valued decoder in the complex-domain Transformer architecture receives the signal modulation features output by the complex-valued encoder, and demodulates the signal modulation features in the time domain to obtain the identified and demodulated information data.

[0012] Preferably, before the step of inputting the signal token sequence into a pre-built signal recognition and demodulation model, the method further comprises:

[0013] Determine the modulation type and modulation parameter information of the training signal to be identified and demodulated in the time domain, wherein the modulation parameter information includes at least one of a modulation parameter and a modulation parameter range;

[0014] Based on the modulation parameter range, generating a complex training signal corresponding to the training signal through signal simulation;

[0015] Generate signal data and signal parameter information corresponding to the complex training signal, and determine the optimal sampling point position of each signal parameter in the signal parameter information to obtain an optimal sampling point position sequence;

[0016] Based on the signal data, the modulation type, the optimal sampling point position sequence and the preset decoding symbol sequence, a training sample is constructed, the training sample is divided into a plurality of signal segments in the time domain, and the divided signal segments are transformed into a signal token sequence; wherein the preset decoding symbol sequence includes a modulation symbol corresponding to bi-phase shift modulation and a modulation symbol corresponding to quadrature phase shift keying, and corresponds to the modulation type of the target signal;

[0017] Generate an optimal sampling point supervision signal according to the optimal sampling point position sequence, and train the signal token sequence based on the optimal sampling point supervision signal to obtain a signal recognition and demodulation basic model based on a complex domain Transformer architecture;

[0018] After the training is completed, the signal recognition and demodulation basic model is evaluated and predicted to determine the optimal training model as the signal recognition and demodulation model.

[0019] Preferably, the step of determining the modulation type and modulation parameter information of the training signal to be identified and demodulated in the time domain, wherein the modulation parameter information includes at least one of a modulation parameter and a modulation parameter range, comprises:

[0020] Determining the number of modulation types of the training signal;

[0021] Determining a shaping filter of the training signal and filtering parameters of the filter;

[0022] determining at least one of a number of samples per symbol parameter or a range of a number of samples per symbol parameter of the signal; and,

[0023] At least one of a signal length parameter or a signal length parameter range is determined.

[0024] Preferably, the step of generating a complex training signal corresponding to the training signal through signal simulation based on the modulation parameter range includes:

[0025] For each of the N modulation types, M independent samples are generated to obtain a training sample set including N×M training samples;

[0026] Generate a simulation signal of the training sample set according to a preset signal simulation rule;

[0027] Amplitude normalization is performed on the simulation signal to obtain a complex training signal corresponding to the training signal.

[0028] Preferably, the step of generating the optimal sampling point supervision signal according to the optimal sampling point position sequence comprises:

[0029] Initialize the optimal sampling point supervision signal to a full "0" signal sequence of a preset length;

[0030] Generate an optimal sampling point supervision signal according to each value in the optimal sampling point position sequence.

[0031] Preferably, the step of training the training samples based on the optimal sampling point supervision signal to obtain a basic model of signal recognition and demodulation based on a complex domain Transformer architecture includes:

[0032] Inputting a complex training signal into the complex domain Transformer architecture preprocessor to generate a token sequence;

[0033] Inputting the token sequence output by the preprocessor into the complex-valued encoder of the complex-domain Transformer architecture to obtain modulation category identification, optimal sampling points and signal modulation characteristics;

[0034] The loss functions of the modulation category recognition, the optimal sampling point and the signal modulation characteristics are weighted and summed, and a basic model of signal recognition and demodulation of a complex domain Transformer architecture is obtained by training with a back-propagation algorithm.

[0035] Preferably, before the step of generating the best sampling point supervision signal according to the best sampling point position sequence, the method further comprises:

[0036] Using random delay length to simulate decimal point sample delay, and performing random delay processing on the training samples;

[0037] Add several "0"s to the preset positions of the training samples after random delay processing to simulate integer sample point delays;

[0038] Random noise is injected into the training samples after integer sample point delay processing to simulate the situation where the signal is affected by noise.

[0039] The present invention provides a signal identification and demodulation system, the system comprising:

[0040] A signal receiving module, used to receive a target signal transmitted by a wireless channel, wherein the target signal is a wireless signal to be identified and demodulated;

[0041] A token sequence generation module, used for dividing the target signal into signal segments of fixed length in the time domain, and inserting a modulation classification label into the signal token header of each signal segment to generate a signal token sequence;

[0042] A feature extraction module, used for inputting the signal token sequence into a pre-built signal recognition and demodulation model, wherein the signal recognition and demodulation model is a model built based on a complex domain Transformer architecture, and a complex-valued encoder in the complex domain Transformer architecture processes the input signal token sequence to obtain the modulation type feature and extract the signal modulation feature of the target signal in the complex domain;

[0043] A signal identification and demodulation module is used to calculate the probability distribution of the modulation type through the fully connected layer in the complex domain Transformer architecture, and determine the modulation type corresponding to the maximum probability value as the modulation type of the target signal; and a complex-valued decoder in the complex domain Transformer architecture receives the signal modulation characteristics output by the complex-valued encoder, and demodulates the signal modulation characteristics in the time domain to obtain the identified and demodulated information data.

[0044] Preferably, the system further comprises: a model building module, for:

[0045] Determine the modulation type and modulation parameter information of the training signal to be identified and demodulated in the time domain, wherein the modulation parameter information includes at least one of a modulation parameter and a modulation parameter range;

[0046] Based on the modulation parameter range, generating a complex training signal corresponding to the training signal through signal simulation;

[0047] Generate signal data and signal parameter information corresponding to the complex training signal, and determine the optimal sampling point position of each signal parameter in the signal parameter information to obtain an optimal sampling point position sequence;

[0048] Based on the signal data, the modulation type, the optimal sampling point position sequence and the preset decoding symbol sequence, a training sample is constructed, the training sample is divided into a plurality of signal segments in the time domain, and the divided signal segments are transformed into a signal token sequence; wherein the preset decoding symbol sequence includes a modulation symbol corresponding to bi-phase shift modulation and a modulation symbol corresponding to quadrature phase shift keying, and corresponds to the modulation type of the target signal;

[0049] Generate an optimal sampling point supervision signal according to the optimal sampling point position sequence, and train the signal token sequence based on the optimal sampling point supervision signal to obtain a signal recognition and demodulation basic model based on a complex domain Transformer architecture;

[0050] After the training is completed, the signal recognition and demodulation basic model is evaluated and predicted to determine the optimal training model as the signal recognition and demodulation model.

[0051] Preferably, the model building module is specifically used for:

[0052] Determining the number of modulation types of the training signal;

[0053] Determining a shaping filter of the training signal and filtering parameters of the filter;

[0054] determining at least one of a number of samples per symbol parameter or a range of a number of samples per symbol parameter of the signal; and,

[0055] At least one of a signal length parameter or a signal length parameter range is determined.

[0056] Preferably, the model building module is specifically used for:

[0057] For each of the N modulation types, M independent samples are generated to obtain a training sample set including N×M training samples;

[0058] Generate a simulation signal of the training sample set according to a preset signal simulation rule;

[0059] Amplitude normalization is performed on the simulation signal to obtain a complex training signal corresponding to the training signal.

[0060] Preferably, the model building module is specifically used for:

[0061] Initialize the optimal sampling point supervision signal to a full "0" signal sequence of a preset length;

[0062] Generate an optimal sampling point supervision signal according to each value in the optimal sampling point position sequence.

[0063] Preferably, the model building module is specifically used for:

[0064] Inputting a complex training signal into the complex domain Transformer architecture preprocessor to generate a token sequence;

[0065] Inputting the token sequence output by the preprocessor into the complex-valued encoder of the complex-domain Transformer architecture to obtain modulation category identification, optimal sampling points and signal modulation characteristics;

[0066] The loss functions of the modulation category recognition, the optimal sampling point and the signal modulation characteristics are weighted and summed, and a basic model of signal recognition and demodulation of a complex domain Transformer architecture is obtained by training with a back-propagation algorithm.

[0067] Preferably, the model building module is further used for:

[0068] Using random delay length to simulate decimal point sample delay, and performing random delay processing on the training samples;

[0069] Add several "0"s to the preset positions of the training samples after random delay processing to simulate integer sample point delays;

[0070] Random noise is injected into the training samples after integer sample point delay processing to simulate the situation where the signal is affected by noise.

[0071] The present invention provides an electronic device, comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the above method when executed by the processor.

[0072] The present invention provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented.

[0073] It can be seen from the above scheme that the present invention deeply integrates deep learning and signal processing technology, and provides an integrated deep learning model, that is, by integrating the recognition and demodulation functions of signal modulation types into a unified complex Transformer model, it can solve the signal recognition and demodulation problems at the same time, and effectively overcome the disadvantages of low efficiency and limited performance caused by the separate design; it also has high adaptability, does not require predicting the signal type or precise synchronization information, and can flexibly respond to complex parameter conditions such as different sampling rates. This feature enhances the versatility and robustness of the system, and can still have stable performance in a dynamically changing communication environment, which provides an efficient, flexible and adaptive solution for application scenarios such as wireless communications, radar monitoring, and electronic warfare. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] 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 only 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.

[0075] Figure 1 A flow chart of the signal identification and demodulation method provided by the present invention;

[0076] Figure 2 A structural diagram of a signal recognition and demodulation model based on a complex Transformer architecture provided by the present invention;

[0077] Figure 3 A schematic diagram of the optimal sampling point supervision signal provided by the present invention;

[0078] Figure 4 A schematic diagram of constructing a signal recognition and demodulation model provided by the present invention;

[0079] Figure 5 The use of the present invention Figure 4 Schematic diagram of the constructed signal recognition model for signal recognition and demodulation;

[0080] Figure 6 An architecture diagram of the signal recognition and demodulation system provided by the present invention;

[0081] Figure 7 Another architecture diagram of the signal recognition and demodulation system provided by the present invention;

[0082] Figure 8 A structural diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0083] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all 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.

[0084] In order to solve the problems of low efficiency and limited performance caused by independent processing of signal recognition and demodulation in the prior art, the present invention provides a signal recognition and demodulation method and system. The signal recognition and demodulation method and system provided by the present invention are described in detail below.

[0085] Embodiment 1

[0086] See also Figure 1 , is a flow chart of the signal identification and demodulation method provided by the present invention. Specifically, the method may include the following steps:

[0087] Step S101: receiving a target signal transmitted by a wireless channel, wherein the target signal is a wireless signal to be identified and demodulated.

[0088] Step S102: Figure 2 As shown, the target signal is divided into signal segments of fixed length in the time domain, and a modulation classification label is inserted into the signal token header of each signal segment to generate a signal token sequence.

[0089] It should be noted that the present application divides the target signal in the time domain, preferably, the division can be based on a fixed length to obtain several signal segments, and for the last signal segment, the length of the signal segment is most likely less than the fixed length. At this time, in order to facilitate subsequent processing, the insufficient length of the signal segment needs to be filled with zeros to make up the length, thereby forming a signal token. Of course, zero padding is only a preferred method provided by the present invention, and there may be other implementation methods.

[0090] For the signal token corresponding to each signal segment, preferably, a modulation classification label can be inserted into the signal token header, for example, a special <start>The tokens serve as modulation classification labels.

[0091] Step S103: Input the signal token sequence into a pre-constructed signal recognition and demodulation model, wherein the signal recognition and demodulation model is a model constructed based on a complex domain Transformer architecture, and the complex-valued encoder in the complex domain Transformer architecture processes the input signal token sequence to obtain the modulation type characteristics and extract the signal modulation characteristics of the target signal in the complex domain.

[0092] The Transformer model architecture is a model proposed by Google in the paper "Attentions is All you need" in 2017. It uses the Self-Attention mechanism to replace the RNN (Recurrent Neural Network) network structure commonly used in NLP (Natural Language Processing) tasks. Compared with the RNN network structure, the biggest advantage of the Transformer model architecture is that it can be calculated in parallel. See Figure 2 The Transformer model architecture is essentially an encoder-decoder architecture. Transformer can be divided into two parts: the encoding component and the decoding component. The encoding component, namely the "complex-valued encoder" below, is used to extract the signal modulation type and the sampling point supervision signal, while the decoding component, namely the "complex-valued decoder" below, is used to predict the real-time symbol probability distribution based on the historical demodulated symbols.

[0093] The complex Transformer architecture used in the present invention can be obtained by extending the real Transformer architecture in a variety of ways, and the extended Transformer architecture can directly process complex signals without converting them into real numbers for processing. For example, it can be obtained by extending the real Transformer architecture in the following ways:

[0094] (1) Complex-valued linear layer expansion

[0095] The complex weight matrix is ​​used to perform conjugate transpose multiplication with the input vector, that is, y = x W H , where W is the weight matrix in complex form.

[0096] (2) Complex-valued multi-head self-attention extension

[0097] The complex-valued attention matrix is ​​converted to a real-valued value through a specific function f(·), followed by probability normalization:

[0098]

[0099] Among them, Q, K, V are query vector, key-value vector and value vector respectively, and d is the dimension of the key-value vector.

[0100] (3) Complex-valued activation function expansion: The real and imaginary parts of the signal are activated independently, which can be expressed as

[0101]

[0102] in and denote the real part and the imaginary function respectively, and g(·) is the activation function in the real number domain.

[0103] (4) Complex-valued layer normalization expansion, by calculating the mean and variance of complex data and performing standardization:

[0104]

[0105] and Var[·] represent expectation and variance operations, and γ and β are trainable complex parameters.

[0106] It should be noted that the above-mentioned methods (1) to (4) are preferred methods provided by the present invention and should not be understood as limitations of the present invention. Other methods may also be used in practical applications and the present invention does not limit them.

[0107] For ease of reading, the following will focus on the construction process of the signal identification and demodulation model.

[0108] Step S104: The modulation type feature is used through a fully connected layer in the complex domain Transformer architecture to calculate the probability distribution of the modulation type, and the modulation type corresponding to the maximum probability value is determined as the modulation type of the target signal.

[0109] Specifically, <start>The modulation type features corresponding to the tokens are passed through the fully connected layer (see Figure 2 ) calculates the probability distribution of the modulation type, and determines the modulation type corresponding to the maximum probability value as the modulation type of the target signal.

[0110] Step S105: The complex-valued decoder in the complex-domain Transformer architecture receives the signal modulation features output by the complex-valued encoder, and demodulates the signal modulation features in the time domain to obtain identified and demodulated information data.

[0111] See also Figure 5 As shown in the figure, during the symbol decoding process, the complex-valued decoder of the complex-domain Transformer architecture takes the output features of the complex-valued encoder as the initial input. Specifically, for the first decoding step, the special <sos>The token is input into the model, marking the start of sequence demodulation; at each decoding time step, the complex-valued decoder uses the symbol recognized in the previous order to predict the symbol probability distribution of the current time step, that is, the complex-valued decoder uses the previous k-1 symbols to predict the symbol probability distribution of the current k-th time step. And the symbol with the highest probability is selected as the current decoding output and as the input of the next time step. This process is repeated until the model predicts <eos>The demodulation ends when the token is reached or the preset maximum sequence length is reached.

[0112] Furthermore, after obtaining the information data after identification and demodulation, the result can also be output. Specifically, the decoded symbol sequence obtained in the above steps can be sorted and output to form the final information data after identification and demodulation.

[0113] It can be seen that the present invention deeply integrates deep learning and signal processing technology, and provides an integrated deep learning model, that is, by integrating the recognition and demodulation functions of signal modulation types into a unified complex Transformer model, it can solve the signal recognition and demodulation problems at the same time, and effectively overcome the disadvantages of low efficiency and limited performance caused by the separate design; it also has high adaptability, does not require predictable signal type or precise synchronization information, and can flexibly respond to complex parameter conditions such as different sampling rates. This feature enhances the versatility and robustness of the system, and can still have stable performance in a dynamically changing communication environment. This provides an efficient, flexible and adaptive solution for application scenarios such as wireless communications, radar monitoring, and electronic warfare.

[0114] As mentioned above, before processing the signal token sequence, the present invention pre-builds a model based on the complex domain Transformer architecture, so that the signal modulation type and demodulation function can be integrated into the model, which is one of the key points of the present invention. Below, we will explain the construction process of the "signal recognition and demodulation model" in detail, please refer to Figures 2 to 5 .

[0115] Specifically, the signal recognition and demodulation model can be constructed according to the following steps:

[0116] Step 1: Determine the modulation type and modulation parameter information of the training signal to be identified and demodulated in the time domain, wherein the modulation parameter information includes at least one of a modulation parameter and a modulation parameter range.

[0117] In one implementation, the modulation type and modulation parameter information of the training signal to be identified and demodulated in the time domain can be determined in the following manner, specifically including: determining the number of modulation types of the training signal; determining the shaping filter of the training signal and the filtering parameters of the filter (for example, if a root raised cosine filter is used, it is necessary to determine the roll-off factor parameter or the roll-off factor parameter range of the filter); determining at least one of the signal's sampling point per symbol parameter or the sampling point per symbol parameter range; and determining the signal length parameter or at least one of the signal length parameter range.

[0118] Step 2: Based on the modulation parameter range, generate a complex training signal corresponding to the training signal through signal simulation.

[0119] In one case, for each of N modulation types, M independent samples are generated to obtain a training sample set containing N×M training samples; a simulation signal of the training sample set is generated according to a preset signal simulation rule; and the simulation signal is amplitude normalized to obtain a complex training signal corresponding to the training signal.

[0120] For example, for a modulated signal within a parameter range, a value may be randomly and uniformly selected within the specified range for signal generation.

[0121] It should be noted that, based on the customized parameter range, the simulation software can be used to generate complex training signals of multiple types and conditions to construct a simulation data set. In addition, the prior art discloses a specific method for generating training signals based on simulation signals, which will not be repeated here.

[0122] Step 3: Generate signal data and signal parameter information corresponding to the complex training signal, and determine the optimal sampling point position of each signal parameter in the signal parameter information to obtain the optimal sampling point position sequence. Figure 3 .

[0123] In one case, the best sampling point supervision signal can be generated in the following manner: the best sampling point supervision signal is initialized to a signal sequence of all "0"s of a preset length; and the best sampling point supervision signal is generated according to each value in the best sampling point position sequence.

[0124] For example, the optimal sampling point position sequence can be generated in the following manner. Consider a signal data without delay (the signal index starting number is "0"), the sampling point number parameter per symbol is s, and the signal data contains k symbols. Then the optimal sampling point position sequence is:

[0125] In a specific implementation, Figure 3 As shown, the best sampling point supervision signal can be generated according to the best sampling point position sequence. Let the best sampling point position sequence of the signal be S, and each value S[i], i=0,…,k-1 in the best sampling point position sequence S is processed according to the following two cases:

[0126] The first one is that if S[i] is an integer value, the assignment expression of the optimal sampling point supervision signal T is: T[S[i]] = 1;

[0127] Second, if S[i] is a non-integer value, the expression for assigning the optimal sampling point supervision signal T is:

[0128]

[0129] Among them, T represents the optimal sampling point supervision signal, and the operator Indicates rounding down, the operator Indicates rounding up.

[0130] Step 4, constructing a training sample based on the signal data, the modulation type, the optimal sampling point position sequence and the preset decoding symbol sequence, wherein the training sample is divided into multiple signal segments in the time domain, and the divided signal segments are transformed into a signal token sequence; wherein the preset decoding symbol sequence includes modulation symbols corresponding to bi-phase shift modulation and modulation symbols corresponding to quadrature phase shift keying, and corresponds to the modulation type of the target signal.

[0131] In one implementation, the training sample symbol sequence includes: two modulation symbols "0" and "1" corresponding to BPSK (Binary Phase Shift Keying); and four modulation symbols "00", "01", "10", and "11" corresponding to QPSK (Quadrature Phase Shift Keying). Here, symbols of different modulation types are regarded as different symbols. It should be noted that the BPSK and QPSK mentioned here are only two preferred modulation methods and are not intended to limit the present invention; of course, there may be other modulation types, and those skilled in the art may make reasonable settings according to actual needs.

[0132] In step 5, an optimal sampling point supervision signal is generated according to the optimal sampling point position sequence, and the signal token sequence is trained based on the optimal sampling point supervision signal to obtain a basic model of signal recognition and demodulation based on a complex domain Transformer architecture.

[0133] Step 6: After the training is completed, the signal recognition and demodulation basic model is evaluated and predicted to determine the optimal training model as the signal recognition and demodulation model.

[0134] In a specific implementation, a complex training signal can be input into the complex domain Transformer architecture preprocessor to generate a token sequence; the token sequence output by the preprocessor is input into the complex value encoder of the complex domain Transformer architecture to obtain modulation category identification, optimal sampling points and signal modulation characteristics; the loss functions of the modulation category identification, the optimal sampling points and the signal modulation characteristics are weighted and summed, and the signal recognition and demodulation basic model of the complex domain Transformer architecture is obtained by training using a back propagation algorithm.

[0135] Specifically, the complex domain signal is first evenly divided into a series of fixed-length segments, where the segments that are less than the length of the segments are padded with zeros; each segment passes through the fully connected layer to form a signal token. Then, a special <start>Tokens, and finally add the above token sequence to the corresponding position embedding vector.

[0136] It should be noted that <start>The token corresponds to the modulation type feature output of the signal for modulation category recognition. The probability distribution of different modulation types (such as BPSK, QPSK, etc.) is calculated through the fully connected layer, and the cross entropy loss function is used to minimize the prediction error between the real modulation category label. The signal token corresponds to the signal feature encoding output for the prediction of the best sampling point, which is extracted by another fully connected layer. The best sampling point prediction is compared with the best sampling point supervision signal T, and the binary cross entropy loss is used to optimize the estimation accuracy.

[0137] See also Figure 4 , <start>The token corresponds to the modulation feature output of the signal and the signal token corresponds to the signal feature encoding output as the decoder input. At the same time, the decoder inputs the previous symbol to predict the probability distribution of the next symbol. When predicting the first symbol, the decoder inputs a special <sos>Token, indicating the start of demodulation, its corresponding features are passed through the encoder and the fully connected layer to output the probability distribution of the first symbol type. For any subsequent decoding step, the decoder inputs the true label of the previously decoded symbol, predicts the symbol probability distribution of the current time step, and uses the cross entropy loss function to minimize the prediction error with the true symbol list label. When the decoding step reaches the end of the sequence, its supervision label is <eos>, indicating that the demodulation is completed.

[0138] Furthermore, in a preferred embodiment of the present invention, data enhancement can be performed on the training samples before the third step, i.e., generating the optimal sampling point supervision signal according to the optimal sampling point position sequence, so as to improve the performance of the constructed model. The specific method may include: using a random delay length to simulate the decimal point sample delay, and performing random delay processing on the training samples; adding a number of "0"s to the preset positions of the training samples after the random delay processing to simulate the integer sample point delay; injecting random noise into the training samples after the integer sample point delay processing to simulate the situation where the signal is affected by noise.

[0139] In one implementation, data enhancement may be performed in the following manner, specifically including:

[0140] (1) Random delay: The signal is randomly delayed, with the delay length being a random value between [0,1], simulating the delay of decimal sample points, and increasing the delay length of each value in the best sampling point position sequence. For example, consider delaying the original signal by p sampling points to form a new signal, then the value in the best sampling point position sequence of the new signal increases by p, that is, S+p, where S is the best sampling point position sequence of the original signal.

[0141] (2) Random padding: Randomly add a number of "0"s before and after the signal to simulate the integer sample point delay, where the number of added 0s is a random value between [0, n], n is the preset zero padding parameter, and each value in the optimal sampling point position sequence is increased by the number of zeros before the signal. For example, consider adding m and n 0s before and after the signal to form a new signal, then each value in the optimal sampling point position sequence of the new signal increases by m, that is, S+m, where S is the optimal sampling point position sequence of the original signal.

[0142] (3) Noise injection: Inject random noise into the signal to simulate the situation where the signal is affected by noise, where the noise size makes the signal-to-noise ratio satisfy [n l , n h ]dB interval range.

[0143] Of course, the specific methods listed here are examples and should not be understood as limiting the present invention. In practical applications, they can be reasonably set according to specific circumstances. It should also be noted that after data enhancement is performed on the training samples, the training samples in subsequent steps such as steps 3 to 6 are all data enhanced training samples.

[0144] It should be noted that the above steps 1 to 6 are the complete construction process of the identification and demodulation model provided by the present invention. It can be seen that the present invention deeply integrates deep learning and signal processing technology, and constructs an integrated deep learning model by integrating the identification and demodulation functions of the signal modulation type into a unified complex Transformer model. It can solve the signal recognition and demodulation problems at the same time, effectively overcoming the drawbacks of low efficiency and limited performance caused by the separate design; it also has a high degree of adaptability, does not require the prediction of signal type or precise synchronization information, and can flexibly respond to complex parameter conditions such as different sampling rates. This feature enhances the versatility and robustness of the system.

[0145] It should be noted that, for the method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the order of the actions described, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions involved are not necessarily required by the present application.

[0146] After introducing the signal identification and demodulation method provided by the present invention, the signal identification and demodulation system provided by the present invention will be described.

[0147] Embodiment 2

[0148] Please refer to Figure 6 , which is an architecture diagram of the signal recognition and demodulation system provided by the present invention, the system includes the following modules: a signal receiving module 210, a token sequence generation module 220, a feature extraction module 230 and a signal recognition and demodulation module 240.

[0149] The signal receiving module 210 is used to receive a target signal transmitted by a wireless channel, wherein the target signal is a wireless signal to be identified and demodulated;

[0150] A token sequence generation module 220, configured to divide the target signal into signal segments of fixed length in the time domain, and insert a modulation classification label into the signal token header of each signal segment to generate a signal token sequence;

[0151] A feature extraction module 230 is used to input the signal token sequence into a pre-built signal recognition and demodulation model, wherein the signal recognition and demodulation model is a model built based on a complex domain Transformer architecture, and a complex-valued encoder in the complex domain Transformer architecture processes the input signal token sequence to obtain the modulation type feature and extract the signal modulation feature of the target signal in the complex domain;

[0152] The signal identification and demodulation module 240 is used to calculate the probability distribution of the modulation type through the fully connected layer in the complex domain Transformer architecture, and determine the modulation type corresponding to the maximum probability value as the modulation type of the target signal; and the complex-valued decoder in the complex domain Transformer architecture receives the signal modulation characteristics output by the complex-valued encoder, and demodulates the signal modulation characteristics in the time domain to obtain the identified and demodulated information data.

[0153] Further, such as Figure 7 As shown, the system may also include a model building module 200, which is used to determine the modulation type and modulation parameter information of the training signal to be identified and demodulated in the time domain, wherein the modulation parameter information includes at least one of a modulation parameter and a modulation parameter range; based on the modulation parameter range, a complex training signal corresponding to the training signal is generated by signal simulation; signal data and signal parameter information corresponding to the complex training signal are generated, and the optimal sampling point position of each signal parameter in the signal parameter information is determined to obtain an optimal sampling point position sequence; based on the signal data, the modulation type, the optimal sampling point position sequence and the preset decoding symbol sequence, a training sample is constructed, and the training The sample is divided into multiple signal segments in the time domain, and the divided signal segments are transformed into a signal token sequence; wherein the preset decoding symbol sequence includes a modulation symbol corresponding to bi-phase phase shift modulation and a modulation symbol corresponding to quadrature phase shift keying, and corresponds to the modulation type of the target signal; an optimal sampling point supervision signal is generated according to the optimal sampling point position sequence, and the signal token sequence is trained based on the optimal sampling point supervision signal to obtain a signal recognition and demodulation basic model based on a complex domain Transformer architecture; after the training is completed, the signal recognition and demodulation basic model is evaluated and predicted, and the optimal training model is determined to be the signal recognition and demodulation model.

[0154] It should be noted that, since the process of constructing the signal recognition and demodulation model has been described in detail in the method embodiment, the relevant details can be found in the description of the method embodiment part, which will not be repeated here.

[0155] In one implementation, the model building module 200 is specifically used to determine the number of modulation types of the training signal; determine the shaping filter of the training signal and the filtering parameters of the filter; determine at least one of the sampling point number per symbol parameter or the sampling point number per symbol parameter range of the signal; and determine at least one of the signal length parameter or the signal length parameter range.

[0156] In one scenario, the model building module 200 is specifically used to generate M independent samples for each of N modulation types to obtain a training sample set including N×M training samples; generate a simulation signal of the training sample set according to a preset signal simulation rule; and perform amplitude normalization processing on the simulation signal to obtain a complex training signal corresponding to the training signal.

[0157] In another case, the model building module 200 is specifically used to initialize the best sampling point supervision signal to a signal sequence of all "0"s of a preset length; and generate the best sampling point supervision signal according to each value in the best sampling point position sequence.

[0158] In another case, the model building module 200 is specifically used to input the complex training signal into the complex domain Transformer architecture preprocessor to generate a token sequence; input the token sequence output by the preprocessor into the complex value encoder of the complex domain Transformer architecture to obtain modulation category identification, optimal sampling point and signal modulation characteristics; perform weighted summation on the loss functions of the modulation category identification, the optimal sampling point and the signal modulation characteristics, and use the back propagation algorithm to train to obtain the signal recognition and demodulation basic model of the complex domain Transformer architecture.

[0159] Furthermore, the model building module 200 is also used to simulate the decimal point sample delay using a random delay scale, and perform random delay processing on the training samples; add a number of "0"s at preset positions to the training samples after random delay processing to simulate integer sample point delays; and inject random noise into the training samples after integer sample point delay processing to simulate the situation where the signal is affected by noise.

[0160] It should be noted that, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method embodiment part.

[0161] Embodiment 3

[0162] To solve the above problems, the present invention provides an electronic device, such as Figure 8 As shown, it includes a memory 310, a processor 320, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above method applied to the electronic device is implemented.

[0163] It should be noted that the electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The electronic device may include, but is not limited to, a processor 320 and a memory 310. Those skilled in the art will understand that Figure 8 These are merely examples of electronic devices and do not constitute a limitation of the electronic device. The electronic device may include more or fewer components than those shown in the figure, or a combination of certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0164] The processor 320 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 gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0165] The memory 310 may be an internal storage unit of the electronic device, such as a hard disk or memory of the electronic device. The memory 310 may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped on the electronic device. Further, the memory 310 may also include both an internal storage unit of the electronic device and an external storage device. The memory 310 is used to store the computer program and other programs and data required by the electronic device. The memory 310 may also be used to temporarily store data that has been output or is to be output.

[0166] Embodiment 4

[0167] The embodiment of the present invention further provides a computer-readable storage medium, which may be the computer-readable storage medium contained in the memory in the above embodiment; or may be a computer-readable storage medium that exists independently and is not assembled into the commuting announcement control device. The computer-readable storage medium stores one or more computer programs, and when the programs are executed by the processor, the above method is implemented.

[0168] 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 the steps 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, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. 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.

[0169] Although the embodiments disclosed in the present invention are as above, the contents described are only embodiments adopted for facilitating the understanding of the present invention and are not intended to limit the present invention. Any technician in the technical field to which the present invention belongs can make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in the present invention, but the protection scope of the present invention shall still be subject to the scope defined in the attached claims.< / eos> < / sos> < / start> < / start> < / start> < / eos> < / sos> < / start> < / start>

Claims

1. A signal recognition and demodulation method, characterized in that: The method comprises: Receiving a target signal transmitted by a wireless channel, wherein the target signal is a wireless signal to be identified and demodulated; Dividing the target signal into signal segments of fixed length in the time domain, and inserting a modulation classification label into the signal token header of each signal segment to generate a signal token sequence; The signal token sequence is input into a pre-built signal recognition and demodulation model, wherein the signal recognition and demodulation model is a model built based on a complex domain Transformer architecture, and a complex-valued encoder in the complex domain Transformer architecture processes the input signal token sequence to obtain the modulation type feature and extract the signal modulation feature of the target signal in the complex domain; The modulation type feature is used to calculate the probability distribution of the modulation type through the fully connected layer in the complex domain Transformer architecture, and the modulation type corresponding to the maximum probability value is determined as the modulation type of the target signal; The complex-valued decoder in the complex-domain Transformer architecture receives the signal modulation features output by the complex-valued encoder, and demodulates the signal modulation features in the time domain to obtain the identified and demodulated information data.

2. The signal identification and demodulation method according to claim 1, characterized in that: Before the step of inputting the signal token sequence into a pre-built signal recognition and demodulation model, the method further includes: Determine the modulation type and modulation parameter information of the training signal to be identified and demodulated in the time domain, wherein the modulation parameter information includes at least one of a modulation parameter and a modulation parameter range; Based on the modulation parameter range, generating a complex training signal corresponding to the training signal through signal simulation; Generate signal data and signal parameter information corresponding to the complex training signal, and determine the optimal sampling point position of each signal parameter in the signal parameter information to obtain an optimal sampling point position sequence; Based on the signal data, the modulation type, the optimal sampling point position sequence and the preset decoding symbol sequence, a training sample is constructed, the training sample is divided into a plurality of signal segments in the time domain, and the divided signal segments are transformed into a signal token sequence; wherein the preset decoding symbol sequence includes a modulation symbol corresponding to bi-phase shift modulation and a modulation symbol corresponding to quadrature phase shift keying, and corresponds to the modulation type of the target signal; Generate an optimal sampling point supervision signal according to the optimal sampling point position sequence, and train the signal token sequence based on the optimal sampling point supervision signal to obtain a signal recognition and demodulation basic model based on a complex domain Transformer architecture; After the training is completed, the signal recognition and demodulation basic model is evaluated and predicted to determine the optimal training model as the signal recognition and demodulation model.

3. The signal recognition and demodulation method according to claim 2, characterized in that: The step of determining the modulation type and modulation parameter information of the training signal to be identified and demodulated in the time domain, wherein the modulation parameter information includes at least one of a modulation parameter and a modulation parameter range, comprises: Determining the number of modulation types of the training signal; Determining a shaping filter of the training signal and filtering parameters of the filter; determining at least one of a number of samples per symbol parameter or a range of a number of samples per symbol parameter of the signal; and, At least one of a signal length parameter or a signal length parameter range is determined.

4. The signal identification and demodulation method according to claim 2, characterized in that: The step of generating a complex training signal corresponding to the training signal through signal simulation based on the modulation parameter range includes: For each of the N modulation types, M independent samples are generated to obtain a training sample set including N×M training samples; Generate a simulation signal of the training sample set according to a preset signal simulation rule; Amplitude normalization is performed on the simulation signal to obtain a complex training signal corresponding to the training signal.

5. The signal identification and demodulation method according to claim 2, characterized in that: The step of generating an optimal sampling point supervision signal according to the optimal sampling point position sequence comprises: Initialize the optimal sampling point supervision signal to a full "0" signal sequence of a preset length; Generate an optimal sampling point supervision signal according to each value in the optimal sampling point position sequence.

6. The signal identification and demodulation method according to claim 2, characterized in that: The step of training the training samples based on the optimal sampling point supervision signal to obtain a signal recognition and demodulation basic model based on a complex domain Transformer architecture includes: Inputting a complex training signal into the complex domain Transformer architecture preprocessor to generate a token sequence; Inputting the token sequence output by the preprocessor into the complex-valued encoder of the complex-domain Transformer architecture to obtain modulation category identification, optimal sampling points and signal modulation characteristics; The loss functions of the modulation category recognition, the optimal sampling point and the signal modulation characteristics are weighted and summed, and a basic model of signal recognition and demodulation of a complex domain Transformer architecture is obtained by training with a back-propagation algorithm.

7. The signal identification and demodulation method according to any one of claims 2 to 6, characterized in that: Before the step of generating the best sampling point supervision signal according to the best sampling point position sequence, the method further includes: Using random delay length to simulate decimal point sample delay, and performing random delay processing on the training samples; Add several "0"s to the preset positions of the training samples after random delay processing to simulate integer sample point delays; Random noise is injected into the training samples after integer sample point delay processing to simulate the situation where the signal is affected by noise.

8. A signal recognition and demodulation system, characterized in that: The system comprises: A signal receiving module, used to receive a target signal transmitted by a wireless channel, wherein the target signal is a wireless signal to be identified and demodulated; A token sequence generation module, used for dividing the target signal into signal segments of fixed length in the time domain, and inserting a modulation classification label into the signal token header of each signal segment to generate a signal token sequence; A feature extraction module, used for inputting the signal token sequence into a pre-built signal recognition and demodulation model, wherein the signal recognition and demodulation model is a model built based on a complex domain Transformer architecture, and a complex-valued encoder in the complex domain Transformer architecture processes the input signal token sequence to obtain the modulation type feature and extract the signal modulation feature of the target signal in the complex domain; A signal identification and demodulation module is used to calculate the probability distribution of the modulation type through the fully connected layer in the complex domain Transformer architecture, and determine the modulation type corresponding to the maximum probability value as the modulation type of the target signal; and a complex-valued decoder in the complex domain Transformer architecture receives the signal modulation characteristics output by the complex-valued encoder, and demodulates the signal modulation characteristics in the time domain to obtain the identified and demodulated information data.

9. An electronic device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the method according to any one of claims 1 to 7 when executed by the processor.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images