Anti-noise communication method and device based on full-connection neural network

Abstract

The invention relates to the technical field of wireless communication, in particular to an anti-noise communication method and device based on a full-connection neural network. The method comprises the following steps: before a signal is sent, coding target information based on a coding sequence to obtain a target signal, and sending the target signal; the coding sequence comprises a plurality of different isometric sequences; the coding sequences are in one-to-one correspondence with various symbols in the target information; after the signal is sent out, receiving the target signal and decoding the target signal through a full-connection neural network to obtain the target information; the full-connection network takes the target signal as input and takes the target information as output; the hidden layer of the full-connection neural network comprises a plurality of full-connection layers for decoding; according to the method, the transmission signal after noise interference is decoded through the neural network, and the problem that the bit error rate is increased due to the fact that the transmission signal is interfered by large noise is effectively avoided.

Description

Technical Field

[0001] The present application relates to the field of wireless communication technology, and in particular to a noise-resistant communication method and device based on a fully connected neural network. Background Art

[0002] In current communication systems, filtering technology, modulation and demodulation technology, and spread spectrum technology can all resist noise interference to a certain extent. Among them, spread spectrum technology shows stronger noise resistance under conventional white noise interference, so spread spectrum technology has been more widely used.

[0003] Spread spectrum technology uses the correlation of communication signals to correlate the received noisy spread spectrum signal with a local known reference signal, thereby suppressing irrelevant noise. However, when the noise is large, the spread spectrum transmission signal will be submerged by the noise, resulting in an increase in the bit error rate. Summary of the invention

[0004] In order to solve the problems existing in the above-mentioned prior art, the purpose of this application is to provide a noise-resistant communication method and device based on a fully connected neural network.

[0005] In a first aspect, the present application provides a noise-resistant communication method based on a fully connected neural network, the method comprising:

[0006] Before the signal is sent, the target information is encoded based on the coding sequence to obtain the target signal and send it out; the coding sequence includes a number of different equal-length sequences; the coding sequence corresponds to each type of symbol in the target information one by one;

[0007] After the signal is sent, the target signal is received and decoded through a fully connected neural network to obtain the target information; the fully connected network takes the target signal as input and the target information as output; the hidden layer of the fully connected neural network includes multiple fully connected layers for decoding.

[0008] In one embodiment, the length of the encoding sequence is configured according to the number of input nodes of the fully connected neural network; and the step of encoding the target information based on the encoding sequence includes:

[0009] According to the coding sequence, each symbol in the target information is encoded in turn to obtain a symbol encoding column vector corresponding to each symbol and having the same length as the coding sequence;

[0010] The symbol coding column vectors are horizontally combined into a coding matrix according to the symbol order in the target information as the target signal.

[0011] In one of the embodiments, the encoding operation includes using the encoding sequence corresponding to each symbol as the symbol encoding column vector corresponding to the symbol; the encoding sequence includes a plurality of discrete sine wave sequences with different frequencies and initial phases.

[0012] In one embodiment, the step of receiving a target signal and decoding it through a fully connected neural network to obtain target information includes:

[0013] The target signal is passed through multiple fully connected layers to obtain a decoding matrix; the column vector of the decoding matrix is ​​a single-valued decoding vector;

[0014] Classify the column vectors of the decoding matrix according to the classification function to obtain a prediction matrix;

[0015] The prediction values ​​of each column vector of the prediction matrix are judged in turn to obtain target information; the judgment operation includes taking the symbol corresponding to the threshold range where the prediction value is located as the output symbol of the column vector.

[0016] In one embodiment, the symbol includes a first symbol and a second symbol with different values; the coding sequence includes a discrete first sine wave sequence corresponding to the first symbol, and a discrete second sine wave sequence corresponding to the second symbol; the first sine wave sequence and the second sine wave sequence have different frequencies and different initial phases.

[0017] In one embodiment, the classification function is a sigmoid function; the threshold range corresponding to the first symbol is less than 0.5, and the threshold range corresponding to the second symbol is not less than 0.5.

[0018] In one of the embodiments, a noisy coding matrix interfered by a preset noise is used as a training set for training a fully connected neural network; the noisy coding matrix includes a symbol coding column vector of a random signal.

[0019] In a second aspect, the present application provides an anti-noise communication device based on a fully connected neural network, the device comprising:

[0020] The coding module is used to encode the target information based on the coding sequence before the signal is sent, obtain the target signal and send it out; the coding sequence includes a number of different equal-length sequences; the coding sequence corresponds to each type of symbol in the target information one by one;

[0021] The decoding module is used to receive the target signal and decode it through a fully connected neural network to obtain target information after the signal is sent; the fully connected network takes the target signal as input and the target information as output; the hidden layer of the fully connected neural network includes multiple fully connected layers for decoding.

[0022] In a third aspect, the present application provides a computer device including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of any one of the methods in the first aspect of the present application are implemented.

[0023] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any one of the methods in the first aspect of the present application when the computer program is executed by a processor.

[0024] The present application describes a noise-resistant communication method based on a fully connected neural network, which encodes each symbol in the target information into a target signal with a wider spectrum and regularity through a coding sequence, thereby improving the spectrum utilization and noise-resistant performance of the target signal; after receiving the target signal, the target signal is feature extracted through a pre-trained fully connected neural network, and a signal close to the coding information in the signal can be identified under noise interference, thereby obtaining the corresponding symbol and decoding the noisy target signal into the target information; the weights of the hidden layer of the pre-trained fully connected neural network are optimized through the noisy signal of the training set, and a close coding sequence can be identified in the noisy signal, effectively avoiding the problem of increased bit error rate caused by large noise interference on the transmission signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the conventional technology, the drawings required for use in the embodiments or the conventional technology descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0026] Figure 1 is a flowchart of the steps of a noise-resistant communication method based on a fully connected neural network in one embodiment;

[0027] Figure 2 is a flowchart of steps for encoding target information into a target signal based on a coding sequence in one embodiment;

[0028] Figure 3 is a flowchart of steps for decoding a target signal into target information by a fully connected neural network in one embodiment;

[0029] Figure 4 is a schematic diagram of nodes in each layer of a hidden layer of an initial fully connected neural network in an embodiment;

[0030] Figure 5 is a comparison diagram of bit error rates after being subjected to noise between the method proposed in this application and the pseudo-random correlation identification method when the length of the coding sequence is 80 discrete points in one embodiment;

[0031] Figure 6 is a comparison diagram of bit error rates after being subjected to noise between the method proposed in this application and the pseudo-random correlation identification method when the length of the coding sequence is 40 discrete points in one embodiment;

[0032] Figure 7 The structure block diagram of a noise-resistant communication device based on a fully connected neural network in an embodiment. DETAILED DESCRIPTION

[0033] In order to facilitate understanding of the present application, the present application will be described more fully below with reference to the relevant drawings. Embodiments of the present application are provided in the drawings. However, the present application can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present application more thorough and comprehensive.

[0034] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0035] When used herein, the singular forms "a", "an", and "said / the" may also include plural forms, unless the context clearly indicates otherwise. It should also be understood that the terms "include / comprise" or "have" and the like specify the presence of stated features, wholes, steps, operations, components, parts, or combinations thereof, but do not exclude the possibility of the presence or addition of one or more other features, wholes, steps, operations, components, parts, or combinations thereof. At the same time, the term "and / or" used in this specification includes any and all combinations of the relevant listed items.

[0036] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0037] The present application provides a noise-resistant communication method based on a fully connected neural network, such as Figure 1 As shown, the following steps S202 to S204 are included:

[0038] S202, before the signal is sent, the target information is encoded based on the coding sequence to obtain the target signal and send it out; the coding sequence includes a plurality of different equal-length sequences; the coding sequence corresponds one-to-one to each type of symbol in the target information.

[0039] Among them, in the encoding of a group of target information, the encoding sequence includes multiple regular amplitude sequences with consistent lengths, which can be multiple waveform sampling sequences with differences in the time domain, or waveform sampling sequences with differences in the frequency domain; the number of possible values ​​of the symbols included in this group of target information is equal to the number of sequences included in the encoding sequence, that is, the encoding sequence corresponds one-to-one to each type of symbol in this group of target information.

[0040] Specifically, before the signal is sent, the encoding module uses the encoding sequence to encode the symbols of the target information, obtains the target signal including the encoding results of all symbols of the target information and sends it out; the encoding process may include directly using the encoding sequence as the encoding output of the symbol, or combining the encoding sequence with the symbol through a correlation function to obtain the encoding output.

[0041] S204, after the signal is sent, the target signal is received and decoded through a fully connected neural network to obtain target information; the fully connected network takes the target signal as input and the target information as output; the hidden layer of the fully connected neural network includes multiple fully connected layers for decoding.

[0042] Specifically, after the signal is sent, the receiving module receives the target signal interfered by noise during transmission, and then inputs the noisy target signal into a pre-trained fully connected neural network as a decoder to obtain decoded target information; the decoding process includes the fully connected neural network identifying the noisy waveform in the noisy target signal that is similar to a certain encoding sequence waveform through multiple layers of fully connected layers in the hidden layer, and thereby obtaining a prediction matrix of the noisy target signal, and then obtaining the output target information based on the prediction matrix.

[0043] The present application describes a noise-resistant communication method based on a fully connected neural network, which encodes each symbol in the target information into a target signal with a wider spectrum and regularity through a coding sequence, thereby improving the spectrum utilization and noise-resistant performance of the target signal; after receiving the target signal, the target signal is feature extracted through a pre-trained fully connected neural network, and a signal close to the coding information in the signal can be identified under noise interference, thereby obtaining the corresponding symbol and decoding the noisy target signal into the target information; the weights of the hidden layer of the pre-trained fully connected neural network are optimized through the noisy signal of the training set, and a close coding sequence can be identified in the noisy signal, effectively avoiding the problem of increased bit error rate caused by large noise interference on the transmission signal.

[0044] In an exemplary embodiment, the length of the encoding sequence is configured according to the number of input nodes of the fully connected neural network; Figure 2 As shown, encoding the target information based on the coding sequence includes the following steps S2022 to S2024:

[0045] S2022, performing encoding operations on each symbol in the target information in sequence according to the encoding sequence, to obtain a symbol encoding column vector corresponding to each symbol and having the same length as the encoding sequence.

[0046] Specifically, according to the order of symbols in the target information, each symbol in the target information is encoded according to the preset corresponding coding sequence, and the symbol code is saved in a column vector format; the length of the symbol code column vector is configured to be equal to the coding sequence, so that the form of the target signal can match the input form of the fully connected neural network.

[0047] Preferably, the encoding operation includes taking the encoding sequence corresponding to each symbol as the symbol encoding column vector corresponding to the symbol; for a number of sinusoidal wave signals with different frequencies and initial phases, based on a preset sampling interval and a preset sampling duration, sampling to obtain multiple different discrete sinusoidal wave sequences as encoding sequences, and the number of discrete points of the encoding sequence is the length of the encoding sequence.

[0048] S2024, horizontally combining the symbol coding column vectors into a coding matrix as the target signal according to the symbol order in the target information.

[0049] Specifically, the symbol coding column vectors are horizontally combined into a coding matrix according to the symbol order. The coding matrix is ​​also the target signal. The number of rows of the coding matrix is ​​equal to the length of the symbol coding column vector, and the number of columns is equal to the number of symbols in the target information.

[0050] In an exemplary embodiment, Figure 3 As shown, receiving the target signal and decoding it through a fully connected neural network to obtain the target information includes the following steps S2042 to S2046:

[0051] S2042, pass the target signal through multiple fully connected layers to obtain a decoding matrix; the column vector of the decoding matrix is ​​a single-valued decoding vector.

[0052] Specifically, after the signal is sent, the target signal interfered by noise during the transmission process is received, and the target signal is decoded by a fully connected neural network; in the fully connected neural network, the weight matrix of each fully connected layer in the hidden layer is multiplied by the target signal on the left to obtain a decoding matrix; the column vector of the decoding matrix is ​​a single-valued decoding vector, which represents the eigenvalue of the column in the target signal, and the number of rows of the decoding matrix is ​​one.

[0053] S2044, classify the column vectors of the decoding matrix according to the classification function to obtain a prediction matrix.

[0054] Specifically, the fully connected neural network also includes a fully connected layer including a classification function, wherein the classification function obtains a predicted value of each column according to the eigenvalue of the column of the decoding matrix; the predicted value is used to distinguish the symbol corresponding to the column.

[0055] S2046, performing judgment operations on the predicted values ​​of each column vector of the prediction matrix in turn to obtain target information; the judgment operation includes taking the symbol corresponding to the threshold range where the predicted value is located as the output symbol of the column vector.

[0056] Specifically, according to the threshold range of the predicted value of each column, the symbol corresponding to each column vector is determined in turn; finally, a row vector whose length is equal to the number of characters in the target information is obtained, and the row vector is the decoded target information; wherein the number of threshold ranges matches the number of possible values ​​of the symbol in the target information.

[0057] It should be understood that although Figure 1-Figure 3 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1-Figure 3 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0058] In order to further illustrate the solution of the present application, a specific example is given below to illustrate that the present application provides a noise-resistant communication method based on a fully connected neural network, comprising the following steps:

[0059] Step 1: Get the target information to be sent as 10011, recorded as s 1,5 The encoding sequence corresponding to symbol 1 preset in the encoding module is a sine wave sequence with 80 discrete points, with a frequency of 0.2π, an initial phase of 0π, and a sampling interval of 0.1s, that is, s 1 (n) = sin(0.2π·0.1·n), where n is the discrete point number, ranging from 0 to 79; the coding sequence corresponding to symbol 0 is a sine wave sequence with 80 discrete points, frequency 0.4π, initial phase 0.5π, sampling interval 0.1s, that is, s 0 (n) = sin(0.4π·0.1·n+0.5π), where n is the discrete point number, ranging from 0 to 79; the target information s is encoded by a preset encoding sequence. 1,5 Encode and obtain the encoded target signal as matrix I 80,5 =[s 1 T ,s 0 T ,s 0 T ,s1 T ,s 1 T ], where T represents the sequence transpose.

[0060] Step 2: The received target signal IN is interfered by noise during transmission. 80,5 , input the pre-trained fully connected neural network for calculation; the fully connected neural network hidden layer has three fully connected layers, and the weight matrix of the first fully connected layer is W 32,80 , the weight matrix of the second fully connected layer is W 8,32 , the weight matrix of the third fully connected layer is W 1,8 , then the decoding matrix O of the hidden layer output 1,5 It is obtained by the following formula:

[0061] O 1,5 =W 1,8 ReLU(W 8,32 ReLU(W 32,80 ·IN 80,5 ))

[0062] Among them, ReLU is the activation function.

[0063] Step 3: The output layer of the fully connected neural network calculates the prediction matrix based on the decoding matrix The calculation formula is as follows:

[0064]

[0065] Among them, sigmoid is the classification function.

[0066] Step 4: For the prediction matrix The predicted values ​​of each column in are judged; if the predicted value of the column is not less than 0.5, the symbol corresponding to the column is judged to be 1; if the predicted value of the column is less than 0.5, the symbol corresponding to the column is judged to be 0; so the prediction matrix The corresponding output target information is 10011, which matches the target information to be sent.

[0067] In a specific embodiment, the method for training a fully connected neural network comprises the following steps:

[0068] Step a: randomly generate 100 symbols with values ​​of 0 or 1, denoted as s 1,100 The 100 symbols are encoded in sequence to obtain the encoding matrix I 80,100 ; The amplitude of the sine wave of the coding sequence is set to 1. Add white noise to the coded sequence to obtain the noisy coding matrix IN 80,100 ; The mean of white noise is 0 and the deviation is 3.

[0069] Step b, the number of nodes in each layer of the hidden layer of the initial fully connected neural network is as follows Figure 4 As shown, where: I 1 to I 80 is the input node of the first fully connected layer; to is the output node of the first fully connected layer, which is also the input node of the second fully connected layer; to is the output node of the second fully connected layer, that is, the input node of the third fully connected layer; 1 It is the output node of the third fully connected layer;

[0070] In the process of training the initial fully connected neural network model using the noisy coding matrix, the binary cross entropy loss function BCEWithLogitsLoss is used; the formula is as follows:

[0071] L = BCEWithLogitsLoss(O 1,100 ,s 1,100 )

[0072] Among them, L is the difference between the output value and the expected value; the weight matrix W of the first fully connected layer is optimized by stochastic gradient descent 32×80 , the weight matrix W of the second fully connected layer 8×32 and the weight matrix W of the third fully connected layer 1×8 ; Repeat the training 2000 times to obtain a pre-trained fully connected neural network.

[0073] In order to verify the anti-noise performance of the above-mentioned anti-noise communication method based on a fully connected neural network, this embodiment conducted the following comparative test:

[0074] Two sine waves with frequency 0.2π, phase 0π, time interval 0.1s and frequency 0.4π, phase 0.5π, time interval 0.1s are selected to generate two sets of coding sequences with lengths of 80 and 40 respectively, and respectively encode symbol 1 and symbol 0 in the test signal. Correspondingly, the spread spectrum coding technology uses the classic m sequence as the spread spectrum pseudo-random sequence to spread spectrum encode symbol 1 and symbol 0 in the test signal, and the sequence lengths are also 80 and 40 respectively. Among them, the m sequence is obtained by the primitive polynomial y=x 6 +x+1, symbol 0 is encoded by any m-sequence, and symbol 1 is encoded by the inverse of the m-sequence. In order to make a fair comparison, it is stipulated that the average coding distance of the sine wave coding sequence used in this method is the same as the average coding distance of the m-sequence and its inverse.

[0075] When the length of the sine wave sequence and the m sequence is 80 discrete points, the bit error rate of the present method and the pseudo-random correlation identification method under white noise interference with a deviation of 1 to 5 is obtained, as shown in Figure 5 As shown; when the length of the sine wave sequence and the m sequence is 40 discrete points, the bit error rate of this method and the pseudo-random correlation identification method under the white noise interference with a deviation of 1 to 5 is obtained as follows Figure 6 As shown;

[0076] analyze Figure 5 and Figure 6 It can be seen that no matter whether the length of the coding sequence used is 80 or 40, the bit error rate of this method is significantly lower than that of the pseudo-random correlation identification method. This proves that the noise-resistant communication method based on a fully connected neural network proposed in this application has good noise-resistant performance.

[0077] Second, as Figure 7 As shown, the present application provides a noise-resistant communication device 700 based on a fully connected neural network, the device comprising:

[0078] The coding module 701 is used to encode the target information based on the coding sequence before the signal is sent, obtain the target signal and send it out; the coding sequence includes a plurality of different equal-length sequences; the coding sequence corresponds to each type of symbol in the target information one by one;

[0079] The decoding module 702 is used to receive the target signal and decode it through a fully connected neural network to obtain target information after the signal is sent; the fully connected network takes the target signal as input and the target information as output; the hidden layer of the fully connected neural network includes multiple fully connected layers for decoding.

[0080] For the specific definition of an anti-noise communication device based on a fully connected neural network, please refer to the definition of an anti-noise communication method based on a fully connected neural network above, which will not be repeated here. Each module in the above-mentioned anti-noise communication device based on a fully connected neural network can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. It should be noted that the division of modules in the embodiment of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation.

[0081] In a third aspect, the present application provides a computer device including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of any one of the noise-resistant communication methods based on a fully connected neural network provided in the present application.

[0082] In a fourth aspect, the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any one of the noise-resistant communication methods based on a fully connected neural network provided in the present application are implemented.

[0083] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0084] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0085] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A noise-resistant communication method based on a fully connected neural network, characterized in that: The method comprises: Before the signal is sent, the target information is encoded based on a coding sequence to obtain a target signal and send it out; the coding sequence includes a plurality of different equal-length sequences; the coding sequence corresponds one-to-one to various symbols in the target information; After the signal is sent, the target signal is received and decoded through a fully connected neural network to obtain the target information; the fully connected network takes the target signal as input and the target information as output; the hidden layer of the fully connected neural network includes multiple fully connected layers for decoding.

2. The method according to claim 1, characterized in that The length of the coding sequence is configured according to the number of input nodes of the fully connected neural network; the step of encoding the target information based on the coding sequence comprises: Perform encoding operations on each symbol in the target information in sequence according to the encoding sequence to obtain a symbol encoding column vector corresponding to each symbol and having the same length as the encoding sequence; The symbol coding column vectors are horizontally combined into a coding matrix according to the symbol sequence in the target information as the target signal.

3. The method according to claim 2, characterized in that The encoding operation includes taking the encoding sequence corresponding to each symbol as the symbol encoding column vector corresponding to the symbol; the encoding sequence includes a plurality of discrete sine wave sequences with different frequencies and initial phases.

4. The method according to claim 2, characterized in that: The steps of receiving the target signal and decoding it through a fully connected neural network to obtain the target information include: Passing the target signal through a plurality of the fully connected layers to obtain a decoding matrix; a column vector of the decoding matrix is ​​a single-valued decoding vector; Classifying the column vectors of the decoding matrix according to a classification function to obtain a prediction matrix; The prediction values ​​of each column vector of the prediction matrix are judged in turn to obtain the target information; the judgment operation includes taking the symbol corresponding to the threshold range where the prediction value is located as the output symbol of the column vector.

5. The method according to claim 4, characterized in that The symbol includes a first symbol and a second symbol with different values; the coding sequence includes a discrete first sine wave sequence corresponding to the first symbol, and a discrete second sine wave sequence corresponding to the second symbol; the first sine wave sequence and the second sine wave sequence have different frequencies and different initial phases.

6. The method according to claim 5, characterized in that The classification function is a sigmoid function; the threshold range corresponding to the first symbol is less than 0.5, and the threshold range corresponding to the second symbol is not less than 0.

5.

7. The method according to claim 1, characterized in that A noisy coding matrix interfered by a preset noise is used as a training set for training the fully connected neural network; the noisy coding matrix includes a symbol coding column vector of a random signal.

8. A noise-resistant communication device based on a fully connected neural network, characterized in that: The device comprises: The coding module is used to encode the target information based on the coding sequence before the signal is sent, obtain the target signal and send it out; the coding sequence includes a plurality of different equal-length sequences; the coding sequence corresponds one-to-one with various symbols in the target information; A decoding module is used to receive the target signal and decode it through a fully connected neural network after the signal is sent to obtain the target information; the fully connected network takes the target signal as input and the target information as output; the hidden layer of the fully connected neural network includes multiple fully connected layers for decoding.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images