Digital self-interference cancellation method and device based on enhanced neural network structure

Abstract

The invention discloses a digital self-interference cancellation method and device based on an enhanced neural network structure, and the method comprises the steps: carrying out the complex value splitting operation and the non-linearity removal operation of a first stepped grid structure, obtaining a second stepped grid structure, carrying out the complex value splitting operation and the non-linearity removal operation of a first sliding window grid structure, and obtaining a second sliding window grid structure; obtaining a second sliding window grid structure; obtaining a first output of the dynamic phase layer according to the second stepped grid structure or the second sliding window grid structure; constructing a vector decomposition structure to separate amplitude information and phase information; feeding the amplitude information into the nonlinear hidden layer to obtain a second output; fusing the phase information and the second output; feeding the fused data into a full-connection output layer, and outputting a simulation result; and according to the simulation result, performing self-interference cancellation operation on the received signal to obtain a target signal. The method can improve the digital self-interference cancellation performance and reduce the network complexity, and can be widely applied to the technical field of full-duplex communication.

Description

Technical Field

[0001] The present invention relates to the technical field of full-duplex communication, and in particular to a digital self-interference cancellation method and device based on an enhanced neural network structure. Background Art

[0002] Most existing neural networks (NNs) structures have problems of high computational complexity and poor self-interference (SI) suppression ability. When modeling non-linear SI with typical NNs, although a certain degree of SI suppression can be achieved, since the process of simulating non-linearity by typical NNs does not conform to the generation law of non-linear SI, it requires a high computational complexity to ensure the cancellation performance. Most existing improved NNs reduce the network structure complexity from the perspective of optimizing the typical NNs network structure and cascading different typical NNs. While achieving certain results, they still do not combine the generation law of non-linear SI, resulting in a high computational complexity. Summary of the Invention

[0003] In view of this, the main purpose of the embodiments of the present invention is to provide a digital self-interference cancellation method and device based on an enhanced neural network structure, in order to solve at least one of the problems in the prior art. The present invention can improve the digital self-interference cancellation performance and reduce the network complexity.

[0004] To achieve the above object, on the one hand, an embodiment of the present invention provides a digital self-interference cancellation method based on an enhanced neural network structure, including the following steps:

[0005] Perform a complex-valued splitting operation and a denonlinearization operation on the first stepped grid structure to obtain a second stepped grid structure, and perform a complex-valued splitting operation and a denonlinearization operation on the first sliding window grid structure to obtain a second sliding window grid structure;

[0006] Obtain a first output of the dynamic phase layer according to the second stepped grid structure or the second sliding window grid structure;

[0007] Construct a vector decomposition structure, and separate the amplitude information and the phase information according to the first output of the dynamic phase layer and the vector decomposition structure;

[0008] Feed the amplitude information into the non-linear hidden layer to obtain a second output of the non-linear hidden layer;

[0009] Perform a fusion operation on the phase information and the second output to obtain fusion data;

[0010] Feed the fusion data into the fully connected output layer to output a simulation result;

[0011] According to the simulation results, perform self-interference cancellation operation on the received signal to obtain the target signal.

[0012] In some embodiments, a digital self-interference cancellation method based on an enhanced neural network structure further includes the following steps:

[0013] Obtain a full-duplex transceiver data set;

[0014] Split the full-duplex transceiver data set to obtain a training set and a test set;

[0015] Train a first neural network based on the second stepped grid structure according to the training set to obtain a second neural network;

[0016] Train a third neural network based on the second sliding window grid structure according to the training set to obtain a fourth neural network;

[0017] According to the test set, perform self-interference cancellation performance tests on the second neural network and the fourth neural network to obtain test results.

[0018] In some embodiments, the operation of performing complex-valued splitting and denonlinearization on the first stepped grid structure to obtain the second stepped grid structure includes the following steps:

[0019] Split the first complex value of the first stepped grid structure into a first real part and a first imaginary part;

[0020] Add a first neuron to the first hidden layer of the first stepped grid structure;

[0021] Combine the original second neuron of the first stepped grid structure with the first neuron, and replace the initial activation function of the first hidden layer with a linear activation function to obtain a second hidden layer;

[0022] Obtain the second stepped grid structure according to the first real part, the first imaginary part, and the second hidden layer.

[0023] In some embodiments, the operation of performing complex-valued splitting and denonlinearization on the first sliding window grid structure to obtain the second sliding window grid structure includes the following steps:

[0024] Split the second complex value of the first sliding window grid structure into a second real part and a second imaginary part;

[0025] Add a third neuron to the third hidden layer of the first sliding window grid structure;

[0026] Combine the original fourth neuron of the first sliding window grid structure with the third neuron, and replace the initial activation function of the third hidden layer with a linear activation function to obtain a fourth hidden layer;

[0027] Obtain the second sliding window grid structure according to the second real part, the second imaginary part, and the fourth hidden layer.

[0028] In some embodiments, the formula used to obtain the first output of the dynamic phase layer according to the second stepped grid structure or the second sliding window grid structure includes:

[0029] o ′ (n) = [o ′ 1 (n), o ′ 2 (n), …, o ′ 2G-1 (n), o ′ 2G (n)] T ;

[0030] Among them, o ′ (n) represents the first output of the dynamic phase layer; o ′ 1 (n), o ′ 2 (n), …, o ′ 2G-1 (n), o ′ 2G (n) represents the outputs of each neuron in the dynamic phase layer; G represents the number of neurons in the non-linear hidden layer; n represents the discrete time; [·] T represents the transpose operation.

[0031] In some embodiments, the formula used to construct the vector decomposition structure and separate the amplitude information and phase information according to the first output of the dynamic phase layer and the vector decomposition structure includes:

[0032]

[0033] Among them, s i (n) represents the amplitude information; sinθ i represents the sine phase; cosθ i represents the cosine phase; G represents the number of neurons in the non-linear hidden layer; n represents the discrete time; o ′ 2i (n), o ′ 2i-1 (n) represent the outputs of the 2i-th and (2i - 1)-th neurons in the dynamic phase layer.

[0034] In some embodiments, the formula used for feeding the amplitude information into the non-linear hidden layer to obtain the second output of the non-linear hidden layer includes:

[0035] x ′ NN (n) = [|s 1 (n)|, |s 1 (n)| 2 , …, |s G (n)|, |s G (n)| 2 T ;

[0036] o ″ NN (n) = [o 1 ″ (n), o 2 ″ (n), …, o G ″ (n)] T ;

[0037] where x ′ NN (n) represents the input of the non-linear hidden layer; s 1 (n), …, s G (n) represent the amplitude information; [·] T represents the transpose operation; o ″ NN (n) represents the second output of the non-linear hidden layer; o 1 ″ (n), o 2 ″ (n), …, o G ″ (n) represent the outputs of each neuron in the non-linear hidden layer.

[0038] In some embodiments, the formula used for fusing the phase information and the second output to obtain the fusion data includes:

[0039] x ″ NN (n) = [o 1 ″ (n) sinθ 1 , o 1 ′ (n) cosθ 1 , …, o G ″ (n) sinθ G , o​G ″ (n)cosθ G T ;

[0040] where x ″ NN (n) represents the fused data; o 1 ″ (n), …, o G ″ (n) represents the outputs of the neurons in the non - linear hidden layer; sinθ 1 , …, sinθ G represents the respective sine phases; cosθ 1 , …, cosθ G represents the respective cosine phases; [·] T represents the transpose operation.

[0041] In some embodiments, for the self - interference cancellation operation on the received signal according to the simulation result to obtain the target signal, the formula used includes:

[0042]

[0043] where y(n) represents the target signal; y SI (n) represents the received signal; represents the simulation result.

[0044] To achieve the above object, on the other hand, an embodiment of the present invention proposes a digital self - interference cancellation device based on an enhanced neural network structure, and the device includes:

[0045] The first module is used to perform a complex - valued splitting operation and a denormalization operation on the first stepped grid structure to obtain the second stepped grid structure, and perform a complex - valued splitting operation and a denormalization operation on the first sliding window grid structure to obtain the second sliding window grid structure;

[0046] The second module is used to obtain the first output of the dynamic phase layer according to the second stepped grid structure or the second sliding window grid structure;

[0047] The third module is used to construct a vector decomposition structure, and separate the amplitude information and the phase information according to the first output of the dynamic phase layer and the vector decomposition structure;

[0048] The fourth module is used to feed the amplitude information into the non - linear hidden layer to obtain the second output of the non - linear hidden layer;

[0049] The fifth module is used to perform a fusion operation on the phase information and the second output to obtain the fused data;​

[0050] A sixth module, configured to feed the fusion data into a fully-connected output layer and output a simulation result;

[0051] A seventh module, configured to perform self-interference cancellation operation on the received signal according to the simulation result to obtain a target signal.

[0052] To achieve the above object, on the other hand, an embodiment of the present invention provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the foregoing digital self-interference cancellation method based on an enhanced neural network structure.

[0053] To achieve the above object, on the other hand, an embodiment of the present invention provides a computer-readable storage medium, which stores a computer program, and when the computer program is executed by a processor, it implements the foregoing digital self-interference cancellation method based on an enhanced neural network structure.

[0054] To achieve the above object, on the other hand, an embodiment of the present invention provides a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device can read the computer instructions from the computer-readable storage medium, and when the processor executes the computer instructions, the computer device executes the foregoing digital self-interference cancellation method based on an enhanced neural network structure.

[0055] The embodiments of the present invention at least include the following beneficial effects: The present invention provides a digital self-interference cancellation method and device based on an enhanced neural network structure. The solution performs a complex-valued splitting operation and a denonlinearization operation on the first stepped grid structure to obtain a second stepped grid structure, and performs a complex-valued splitting operation and a denonlinearization operation on the first sliding window grid structure to obtain a second sliding window grid structure; obtains a first output of the dynamic phase layer according to the second stepped grid structure or the second sliding window grid structure; constructs a vector decomposition structure, and separates and obtains amplitude information and phase information according to the first output of the dynamic phase layer and the vector decomposition structure; feeds the amplitude information into a nonlinear hidden layer to obtain a second output of the nonlinear hidden layer; performs a fusion operation on the phase information and the second output to obtain fusion data; feeds the fusion data into a fully-connected output layer to output a simulation result; performs a self-interference cancellation operation on the received signal according to the simulation result to obtain a target signal. By applying the stepped grid structure and the sliding window grid structure to the neural network, the network complexity is reduced while ensuring the digital self-interference cancellation performance. Description of the Drawings

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0057] Figure 1 is a flowchart of a digital self-interference cancellation method based on an enhanced neural network structure provided by an embodiment of the present invention;

[0058] Figure 2 is a schematic diagram of the complex value splitting and de - non - linearization of the LWGS structure provided by an embodiment of the present invention;

[0059] Figure 3 is a schematic diagram of the complex value splitting and de - non - linearization of the MWGS structure provided by an embodiment of the present invention;

[0060] Figure 4 is a schematic diagram of the structure of a neural network canceller based on enhanced LWGS provided by an embodiment of the present invention;

[0061] Figure 5 is a schematic diagram of the structure of a neural network canceller based on enhanced MWGS provided by an embodiment of the present invention;

[0062] Figure 6 is a schematic diagram of the function in a full - duplex transceiver provided by an embodiment of the present invention;

[0063] Figure 7 is a schematic diagram of the comparison of cancellation effects provided by an embodiment of the present invention;

[0064] Figure 8 is a schematic diagram of the hardware structure of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0065] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the following further details the present invention in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with the embodiments of the present invention. They are only examples of devices and methods consistent with some aspects of the embodiments of the present invention detailed in the appended claims.

[0066] It should be noted that although the functional modules are divided in the system schematic diagram and the logical sequence is shown in the flowchart, in some cases, the steps shown or described may be executed in a different module division from that in the system or a different order from that in the flowchart. The terms "first / S100" and "second / S200" in the specification, claims and the above-mentioned drawings may be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present invention, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "when" as used herein may be interpreted as "when" or "while" or "in response to determining".

[0067] The terms "at least one", "a plurality of", "each", "any one" and the like used in the present invention, at least one includes one, two or more than two, a plurality of includes two or more than two, each refers to each of the corresponding plurality, and any one refers to any one of the plurality.

[0068] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. The terms used herein are only for the purpose of describing the embodiments of the present invention and are not intended to limit the present invention.

[0069] Before elaborating on the embodiments of the present invention in detail, some nouns and terms involved in the embodiments of the present invention are first described, and the nouns and terms involved in the embodiments of the present invention are applicable to the following explanations.

[0070] Simultaneous Transmit and Receive (STAR): A technology for transmitting and receiving electromagnetic wave signals in the same radio system at the same time and in the same frequency band.

[0071] Self-Interference Cancellation (SIC): Refers to the process of suppressing strong self-interference signals from the transmitter end by different means in different parts of the simultaneous transmit and receive system, including the propagation domain, analog domain, and digital domain. Self-interference cancellation is the key to realizing the simultaneous transmit and receive technology.

[0072] Ladderwise Grid Structure (LWGS): Refers to a neural network structure for digital self-interference cancellation of a full-duplex transceiver. The network is trained with complex data and constructs the connection between the neurons of the input layer and the only hidden layer based on a trapezoidal topology structure.

[0073] Moving window grid structure (MWGS): A neural network structure used for digital self-interference cancellation in full-duplex transceivers. This network is trained with complex data and constructs connections between neurons in the input layer and the only hidden layer based on its sliding window topology.

[0074] With the rapid development of the radio field, wireless frequency band resources have gradually shown a trend of exhaustion. To address this problem, many researchers have proposed the concept of simultaneous transceiver technology, also known as simultaneous co-frequency full-duplex technology. This technology aims to simultaneously transmit and receive electromagnetic wave signals using the same time and frequency resources in the same medium resource. Therefore, this technology can double the utilization rate of time and spectrum resources, but at the same time, it also introduces a strong self-interference problem.

[0075] In the past decade, some progress has been made in simultaneous transceiver technology. To suppress the influence of self-interference signals at the receiving end as much as possible, there are three commonly used self-interference cancellation methods, namely propagation domain, analog domain, and digital domain cancellation. Generally speaking, the linear self-interference component can be modeled by weighted combination of different time delay terms, and the nonlinear self-interference component can be characterized by a polynomial model. However, the polynomial model requires a large number of model parameters and high computational complexity, especially when the nonlinear order is high. Therefore, neural networks have been developed to model nonlinear self-interference.

[0076] Most of the existing neural network structures have problems of high computational complexity and poor self-interference suppression ability. When using typical neural networks to model nonlinear self-interference, although a certain degree of self-interference suppression can be achieved, since the process of simulating nonlinearity by typical neural networks does not conform to the generation law of nonlinear self-interference, it requires high computational complexity to ensure the cancellation performance. Most of the existing improved neural networks reduce the complexity of the network structure from the perspective of optimizing the network structure of typical neural networks or cascading different typical neural networks. While achieving certain results, they still do not combine the generation law of nonlinear self-interference, resulting in high computational complexity remaining.

[0077] In view of this, as Figure 1 shown, the embodiments of the present invention provide a digital self-interference cancellation method based on an enhanced neural network structure, and this method may include but is not limited to steps S100 to S700:

[0078] Step S100, perform a complex value splitting operation and a denonlinearization operation on the first stepped grid structure to obtain a second stepped grid structure, and perform a complex value splitting operation and a denonlinearization operation on the first moving window grid structure to obtain a second moving window grid structure;

[0079] Step S200: Obtain the first output of the dynamic phase layer according to the second stepped grid structure or the second sliding window grid structure.

[0080] Step S300: Construct a vector decomposition structure, and separate the amplitude information and the phase information according to the first output of the dynamic phase layer and the vector decomposition structure.

[0081] Step S400: Feed the amplitude information into the non-linear hidden layer to obtain the second output of the non-linear hidden layer.

[0082] Step S500: Perform a fusion operation on the phase information and the second output to obtain fusion data.

[0083] Step S600: Feed the fusion data into the fully connected output layer to output an analog result.

[0084] Step S700: Perform self-interference cancellation on the received signal according to the analog result to obtain a target signal.

[0085] In steps S100 to S700 of some embodiments, the stepped grid structure and the sliding window grid structure after complex value splitting and de-nonlinearization are used as the dynamic phase layer, then the phase information and the amplitude information are separated, only the amplitude information is sent into the non-linear hidden layer to characterize non-linear self-interference, and finally the phase information is restored and the output is obtained through the fully connected layer.

[0086] In some embodiments, step S100 may include but is not limited to steps S101 to S104:

[0087] Step S101: Split the first complex value of the first stepped grid structure into a first real part and a first imaginary part.

[0088] Step S102: Add a first neuron to the first hidden layer of the first stepped grid structure.

[0089] Step S103: Combine the original second neuron of the first stepped grid structure with the first neuron, and replace the initial activation function of the first hidden layer with a linear activation function to obtain a second hidden layer.

[0090] Step S104: Obtain the second stepped grid structure according to the first real part, the first imaginary part and the second hidden layer.

[0091] In step S101 of some embodiments, the first complex value of the first stepped grid structure is split into a first real part and a first imaginary part, and the data formats of the first real part and the first imaginary part are both real values, that is, real numbers. For example Figure 2As shown, the first complex value x 1 (n) can be split into a first real part and a first imaginary part Exemplarily, the original complex data is split into a real part and an imaginary part, and then combined into real-valued input data. Among them, the expression of the real-valued input data is:

[0092]

[0093] Among them, x NN (n) represents the input of the neural network digital canceller; represents the first real part; represents the first imaginary part; [·] T represents the transpose operation.

[0094] In steps S102 to S103 of some embodiments, a new neuron is added to the output of each original hidden layer (i.e., the first hidden layer), and the newly added hidden layer neuron and the original neuron are respectively used to represent the real part and the imaginary part of the hidden layer output data. Then, the activation function of the hidden layer in the first stepped grid structure is replaced with a linear activation function.

[0095] Before step S104 in some embodiments, it further includes deleting the output layer of the first stepped grid structure and connecting to the vector decomposition structure in the subsequent steps. Then, according to the first real part, the first imaginary part, and the second hidden layer, an enhanced stepped grid structure as shown in Figure 2 can be obtained, that is, the second stepped grid structure.

[0096] As shown in Figure 2 , the input data of the stepped grid structure is changed to a format in which the real part and the imaginary part appear alternately. Therefore, an additional connection is added for the input at a certain moment. Since the weights and biases in the neurons are all real values, an additional neuron is needed to describe the real part and the imaginary part of the output data respectively. In addition, the non-linear activation function in the hidden layer of the original stepped grid structure is replaced with a linear activation function, and the network connection rule remains unchanged. Among them, Figure 2 the h in 1 , h 2 ,..., h N represents N neurons.

[0097] In some embodiments, step S100 may further include but is not limited to steps S111 to S114:

[0098] Step S111, splitting the second complex value of the first sliding window grid structure into a second real part and a second imaginary part;

[0099] Step S112, adding a third neuron to the third hidden layer of the first sliding window grid structure;

[0100] Step S113: Combine the original fourth neuron of the first sliding window grid structure with the third neuron, and replace the initial activation function of the third hidden layer with a linear activation function to obtain a fourth hidden layer.

[0101] Step S114: Obtain the second sliding window grid structure based on the second real part, the second imaginary part, and the fourth hidden layer.

[0102] In step S111 of some embodiments, split the second complex value of the first sliding window grid structure into a second real part and a second imaginary part, and the data formats of the second real part and the second imaginary part are both real values, that is, real numbers. As Figure 3 shown, the second complex value x 2 (n) can be split into a second real part and a second imaginary part Exemplarily, split the original complex number data into a real part and an imaginary part, and then combine them into real-valued input data. Among them, the expression of the real-valued input data is:

[0103]

[0104] where x NN (n) represents the input of the neural network digital canceller; represents the second real part; represents the second imaginary part.

[0105] In steps S112 to S113 of some embodiments, a new neuron is added to the output of each original hidden layer (i.e., the third hidden layer), and the newly added hidden layer neuron and the original neuron are respectively used to represent the real part and the imaginary part of the hidden layer output data. Then, replace the activation function of the hidden layer in the first sliding window grid structure with a linear activation function.

[0106] Before step S114 of some embodiments, it further includes deleting the output layer of the first sliding window grid structure, which is used to connect the vector decomposition structure in the subsequent steps. Then, based on the second real part, the second imaginary part, and the fourth hidden layer, an enhanced sliding window grid structure as Figure 3 shown can be obtained, that is, the second sliding window grid structure.

[0107] As Figure 3As shown, the input data of the sliding window grid structure is changed to a format where the real part and the imaginary part appear alternately. Therefore, an additional connection is added for the input at a certain moment. Since the weights and biases in the neurons are both real values, an additional neuron is needed to separately describe the real part and the imaginary part of the output data. Additionally, the non-linear activation function in the hidden layer of the original sliding window grid structure is replaced with a linear activation function, and the network connection rule remains unchanged. For ease of representation, optionally, Figure 3 the sliding window size of the sliding window grid structure in

[0108] In step S200 of some embodiments, the second stepped grid structure or the second sliding window grid structure is used as the dynamic phase layer of the neural network. Through the grid design concept of the stepped grid structure and the sliding window grid structure and applying it to the neural network, the network complexity can be reduced while ensuring the digital self-interference cancellation performance. Then the output of the dynamic phase layer is:

[0109] o ′ (n)=[o ′ 1 (n),o ′ 2 (n),…,o ′ 2G-1 (n),o ′ 2G (n)] T ;

[0110] where, o ′ (n) represents the first output of the dynamic phase layer; o ′ 1 (n),o ′ 2 (n),…,o ′ 2G-1 (n),o ′ 2G (n) represents the outputs of the respective neurons in the dynamic phase layer; G represents the number of neurons in the non-linear hidden layer; n represents the discrete time; [·] T represents the transpose operation.

[0111] In step S300 of some embodiments, a vector decomposition structure is designed and constructed. According to the first output of the dynamic phase layer, the amplitude information and the phase information are separated through this vector decomposition structure. Exemplarily, the expression of the vector decomposition structure (vectordecompose structure, VDS) is:

[0112]

[0113] where, s i (n) represents the amplitude information; sinθi represents the sine phase; cosθ i represents the cosine phase; G represents the number of neurons in the non - linear hidden layer; n represents the discrete time; o ′ 2i (n), o ′ 2i-1 (n) represents the outputs of the 2i - th and (2i - 1)-th neurons in the dynamic phase layer.

[0114] In step S400 of some embodiments, feeding the amplitude information into a non - linear hidden layer with the activation function Tanh can obtain the second output of the non - linear hidden layer. This step is a key step in simulating non - linear self - interference, and the output result conforms to the non - linear working law. Among them, the amplitude information includes the modulus value and the square of the modulus value. Exemplarily, the input and output data of the non - linear hidden layer are respectively:

[0115] x ′ NN (n)=[|s 1 (n)|,|s 1 (n)| 2 ,…,|s G (n)|,|s G (n)| 2 T ;

[0116] o ″ NN (n)=[o 1 ″ (n),o 2 ″ (n),…,o G ″ (n)] T ;

[0117] Among them, x ′ NN (n) represents the input of the non - linear hidden layer; s 1 (n),…,s G (n) represents the amplitude information; [·] T represents the transpose operation; o ″ NN (n) represents the second output of the non - linear hidden layer; o 1 ″ (n),o 2 ″ (n),…,o G ″ (n) represents the outputs of each neuron in the non - linear hidden layer; NN represents the neural network.

[0118] ​In some embodiments, step S400 may include, but is not limited to, steps S401 to S402:

[0119] Step S401, solve the Taylor expansion of the Tanh function. Exemplarily, the Taylor expansion of the Tanh function is:

[0120]

[0121] where tanh(x) represents the Tanh function; e x represents the exponential function.

[0122] Step S402, solve the output expression of the non - linear hidden layer. Exemplarily, the output expression of the non - linear hidden layer is:

[0123]

[0124] where w i (n)=[w i,1 (n), w i,2 (n),…, w i,G (n)] represents the connection weights of the i - th neuron in the non - linear hidden layer; b i (n) represents the bias of the i - th neuron in the non - linear hidden layer.

[0125] In step S500 of some embodiments, a phase recovery (PR) operation is designed. By fusing the amplitude information in the second output of the non - linear hidden layer with the original phase information, phase recovery is achieved. Exemplarily, the expression of the fused data is:

[0126] x ″ NN (n)=[o 1 ″ (n)sinθ 1 , o 1 ″ (n)cosθ 1 ,…, o G ″ (n)sinθ G , o G ″ (n)cosθ G T ;

[0127] where x ″ NN (n) represents the fused data; o 1 ″ (n),…, o G ″ ​(n) represents the output of each neuron in the non - linear hidden layer; sinθ 1 ,…,sinθ G represents each sine phase; cosθ 1 ,…,cosθ G represents each cosine phase; [·] T represents the transpose operation.

[0128] In step S600 of some embodiments, feeding the fused data into the fully - connected output layer can obtain the real part and the imaginary part of the self - interference estimation, that is, the simulation result, and this simulation result can be used to subtract the self - interference from the received signal.

[0129] In step S700 of some embodiments, according to the simulation result of the neural network canceller based on the enhanced stepped grid structure or the enhanced sliding window grid structure, performing a self - interference cancellation operation on the received signal can obtain the remaining signal after cancellation, that is, the target signal. Exemplarily, the expression for obtaining the target signal is:

[0130]

[0131] where y(n) represents the target signal; y SI (n) represents the received signal; represents the simulation result.

[0132] In some embodiments, the digital self - interference cancellation method based on the enhanced neural network structure may further include: obtaining a full - duplex transceiver data set; splitting the full - duplex transceiver data set to obtain a training set and a test set; training a first neural network based on the second stepped grid structure according to the training set to obtain a second neural network; training a third neural network based on the second sliding window grid structure according to the training set to obtain a fourth neural network; and performing a self - interference cancellation performance test on the second neural network and the fourth neural network according to the test set to obtain a test result.

[0133] Exemplarily, obtaining a public data set, which is the transceiver data of a 10 - MHz orthogonal frequency - division multiplexing signal running on a full - duplex platform. Splitting this data set into a training set and a test set, and using the training set to train a neural network based on the enhanced stepped grid structure and a neural network based on the enhanced sliding window grid structure respectively. Then using the test set to test the self - interference cancellation performance of the trained enhanced neural network.

[0134] Such as Figure 4As shown, a neural network structure based on an enhanced stepped grid structure is described. Among them, the dynamic phase layer is a stepped grid structure after complex value splitting and de - nonlinearization. The number of neurons is 2G, and every two neurons form a group to constitute complex data and input it into the vector decomposition structure. The amplitude information and phase information are respectively extracted through the vector decomposition structure. The amplitude information includes the modulus value and the square of the modulus value, and the phase information is used for phase recovery of the input data of the output layer. The number of neurons in the non - linear hidden layer is G, and the activation function is the Tanh function. After performing the phase recovery operation on the output of the non - linear hidden layer, the simulation result is output through the fully - connected output layer.

[0135] Taking the neural network canceller based on the enhanced stepped grid structure as an example, the implementation process of a digital self - interference cancellation method based on the enhanced neural network structure in an embodiment of the present invention is introduced as follows:

[0136] Step 1: Perform complex value splitting and de - nonlinearization on the stepped grid structure;

[0137] Optionally, in an embodiment of the present invention, the number of hidden layer neurons in the stepped grid structure after complex value splitting and de - nonlinearization can be 8, 12, 16, and the corresponding values of G are 4, 6, 8 respectively.

[0138] Step 2: Use the stepped grid structure after complex value splitting and de - nonlinearization as the dynamic phase layer.

[0139] Step 3: Design a vector decomposition structure to separate the amplitude information and phase information;

[0140] Optionally, there are G vector decomposition structures in an embodiment of the present invention.

[0141] Step 4: Feed the amplitude information into the non - linear hidden layer with the Tanh function as the activation function, where the amplitude information includes the modulus value and the square of the modulus value.

[0142] Step 5: After the phase recovery operation, fuse the amplitude information output by the non - linear hidden layer and the phase information in Step 2.

[0143] Step 6: Feed the fused data into the fully - connected output layer to obtain the real part and imaginary part of the self - interference estimation, which are used to subtract the self - interference from the received signal.

[0144] Step 7: Split the data set into a training set and a test set, and use the training set to train the enhanced LWGS network.

[0145] Step 8: Use the test set to test the self - interference cancellation performance of the trained enhanced LWGS network.

[0146] As Figure 5As shown, a neural network structure based on a sliding window grid structure is described. The dynamic phase layer is a sliding window grid structure after complex value splitting and de - nonlinearization, with the number of neurons being 2G, the sliding window size being 2, and every two neurons forming a group to constitute complex data and input into the vector decomposition structure. The vector decomposition structure extracts amplitude information and phase information respectively. The amplitude information includes the modulus value and the square of the modulus value, and the phase information is used for phase recovery of the input data of the output layer. The number of neurons in the non - linear hidden layer is G, and the activation function is the Tanh function. After performing phase recovery operations on the output of the non - linear hidden layer, the simulation results are output through the fully - connected output layer.

[0147] Taking the neural network canceller based on the enhanced sliding window grid structure as an example, the implementation process of a digital self - interference cancellation method based on the enhanced neural network structure in an embodiment of the present invention is introduced as follows:

[0148] Step 1: Perform complex value splitting and de - nonlinearization on the sliding window grid structure;

[0149] Step 2: Use the sliding window grid structure after complex value splitting and de - nonlinearization as the dynamic phase layer.

[0150] Step 3: Design a vector decomposition structure to separate amplitude information and phase information;

[0151] Optionally, there are G vector decomposition structures in an embodiment of the present invention.

[0152] Step 4: Feed the amplitude information into a non - linear hidden layer with the Tanh function as the activation function, where the amplitude information includes the modulus value and the square of the modulus value.

[0153] Step 5: After performing phase recovery operations, fuse the amplitude information output by the non - linear hidden layer and the phase information in Step 2.

[0154] Step 6: Feed the fused data into the fully - connected output layer to obtain the real part and the imaginary part of the self - interference estimation, which are used to subtract the self - interference from the received signal.

[0155] Step 7: Split the data set into a training set and a test set, and use the training set to train the enhanced LWGS network.

[0156] Step 8: Use the test set to test the self - interference cancellation performance of the trained enhanced LWGS network.

[0157] Reference Figure 6 , describes a typical application scenario of an embodiment of the present invention, that is, a non - linear canceller. As Figure 6 shown, x(n) is a digital baseband signal, x IQ (n) is the signal after passing through the up - mixer, x PA(n) is the transmitted signal after passing through the power amplifier. After passing through the self-interference channel, low-noise amplifier (LNA), downconverter, and analog-to-digital converter (ADC), it becomes y SI (n), where is the linear self-interference time-domain response. According to the simulation results of the neural network canceller based on enhanced LWGS or enhanced MWGS the remaining signal after cancellation is y(n). Therefore, the digital self-interference cancellation ratio can be defined to measure the network performance:

[0158]

[0159] where, γ represents dB represents the digital self-interference cancellation ratio.

[0160] Reference Figure 7 , taking the performance of the 5th-order non-linear polynomial canceller as a reference, compare the results of the neural network canceller under different parameter settings. As Figure 7 shown, it describes the self-interference cancellation performance of the linear canceller, typical polynomial canceller, neural network canceller based on enhanced laddered grid structure (ALWGS), and neural network canceller based on enhanced sliding window grid structure (AMWGS). Among them, the numbers in the legend names represent the number of neurons. For example, enhanced laddered grid structure (6) means G = 6; enhanced sliding window grid structure (3,5) means G = 3 and the sliding window size is 5. According to Figure 7 the results shown, it can be seen that the digital self-interference cancellation performance can be improved by the enhanced laddered grid structure and enhanced sliding window grid structure of the present invention.

[0161] The embodiment of the present invention also provides a digital self-interference cancellation device based on an enhanced neural network structure, which can implement the above-mentioned digital self-interference cancellation method based on an enhanced neural network structure. The device includes:

[0162] The first module is used to perform a complex-valued splitting operation and a denonlinearization operation on the first laddered grid structure to obtain a second laddered grid structure, and perform a complex-valued splitting operation and a denonlinearization operation on the first sliding window grid structure to obtain a second sliding window grid structure;

[0163] The second module is used to obtain the first output of the dynamic phase layer according to the second laddered grid structure or the second sliding window grid structure;

[0164] The third module is used to construct a vector decomposition structure, and separate the amplitude information and phase information according to the first output of the dynamic phase layer and the vector decomposition structure;

[0165] A fourth module, configured to feed the amplitude information into a non-linear hidden layer to obtain a second output of the non-linear hidden layer;

[0166] A fifth module, configured to perform a fusion operation on the phase information and the second output to obtain fusion data;

[0167] A sixth module, configured to feed the fusion data into a fully-connected output layer to output an analog result;

[0168] A seventh module, configured to perform self-interference cancellation on the received signal according to the analog result to obtain a target signal.

[0169] It can be understood that the content in the above method embodiments is applicable to the device embodiments. The functions specifically implemented by the device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0170] An embodiment of the present invention further provides an electronic device, which includes a processor and a memory. The memory stores a computer program, and when the processor executes the computer program, the above-mentioned digital self-interference cancellation method based on an enhanced neural network structure is implemented. The electronic device can be any intelligent terminal including a tablet computer, an in-vehicle computer, etc.

[0171] It can be understood that the content in the above method embodiments is applicable to the device embodiments. The functions specifically implemented by the device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0172] Reference Figure 8 , Figure 8 schematically shows the hardware structure of an electronic device according to another embodiment. The electronic device includes:

[0173] A processor 801, which can be implemented by using a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, etc., and is configured to execute relevant programs to implement the technical solution provided by the embodiment of the present invention;

[0174] The memory 802 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM), etc. The memory 802 can store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 802 and are called by the processor 801 to execute a digital self-interference cancellation method based on an enhanced neural network structure according to an embodiment of the present invention;

[0175] The input / output interface 803 is used to implement information input and output;

[0176] The communication interface 804 is used to implement communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WI FI, Bluetooth, etc.);

[0177] The bus 805 transmits information between various components of the device (such as the processor 801, the memory 802, the input / output interface 803, and the communication interface 804);

[0178] Among them, the processor 801, the memory 802, the input / output interface 803, and the communication interface 804 achieve communication connections with each other inside the device through the bus 805.

[0179] An embodiment of the present invention further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the above-mentioned digital self-interference cancellation method based on an enhanced neural network structure.

[0180] It can be understood that the content in the above method embodiments is applicable to the embodiments of this storage medium. The functions specifically implemented by the embodiments of this storage medium are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0181] An embodiment of the present invention further provides a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the above-mentioned digital self-interference cancellation method based on an enhanced neural network structure.

[0182] In summary, a digital self-interference cancellation method and device based on an enhanced neural network structure according to an embodiment of the present invention have the following advantages:

[0183] 1. The embodiments of the present invention design the structure of the neural network in combination with the non-linear working law, so that the neural network canceller has excellent performance in characterizing non-linear self-interference.

[0184] 2. The embodiments of the present invention utilize the grid design concept of LWGS and MWGS and apply it to the neural network, so as to reduce the network complexity while ensuring the digital self-interference cancellation performance. The present invention can be widely applied to the fields of radar phased array and full-duplex communication technology.

[0185] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order mentioned in the operating diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously or the blocks can sometimes be executed in the reverse order. In addition, the embodiments presented and described in the flowcharts of the present invention are provided by way of example for the purpose of providing a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logical flows presented herein. Alternative embodiments are contemplated, in which the order of various operations is changed and the sub-operations described as part of a larger operation are executed independently.

[0186] In addition, although the present invention has been described in the context of functional modules, it should be understood that one or more of the functions and / or features described, unless otherwise stated to the contrary, may be integrated in a single physical device and / or software module, or one or more functions and / or features may be implemented in separate physical devices or software modules. It can also be understood that a detailed discussion of the actual implementation of each module is not necessary for understanding the present invention. Rather, given the attributes, functions, and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the modules will be understood within the ordinary skills of an engineer. Therefore, those skilled in the art can implement the present invention as set forth in the claims without undue experimentation. It can also be understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0187] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0188] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in combination with an instruction execution system, apparatus, or device.

[0189] More specific examples (non-exhaustive list) of computer-readable media include the following: electrical connection parts with one or more wirings (electronic devices), portable computer disk cartridges (magnetic devices), random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memories), fiber optic devices, and portable compact disc read-only memories (CDROM). Additionally, a computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, then editing, interpreting, or otherwise processing it as appropriate, and then storing it in a computer memory.

[0190] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits with logic gate circuits for implementing logic functions on data signals, application specific integrated circuits with suitable combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0191] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0192] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the claims and their equivalents.

[0193] The above has specifically described the preferred embodiments of the present invention, but the present invention is not limited to the described embodiments. Those skilled in the art can also make various equivalent deformations or substitutions without departing from the spirit of the present invention, and these equivalent deformations or substitutions are all included within the scope defined by the claims of the present invention.

Claims

1. A digital self-interference cancellation method based on an enhanced neural network structure, characterized in that: The following steps are involved: Performing a complex value splitting operation and a de-linearization operation on the first stepped grid structure to obtain a second stepped grid structure, and performing a complex value splitting operation and a de-linearization operation on the first sliding window grid structure to obtain a second sliding window grid structure; Acquire a first output of a dynamic phase layer according to the second stepped grid structure or the second sliding window grid structure; constructing a vector decomposition structure, and separating amplitude information and phase information according to the first output of the dynamic phase layer and the vector decomposition structure; Feeding the amplitude information into a nonlinear hidden layer to obtain a second output of the nonlinear hidden layer; Performing a fusion operation on the phase information and the second output to obtain fused data; Feeding the fused data into a fully connected output layer to output a simulation result; According to the simulation results, a self-interference cancellation operation is performed on the received signal to obtain a target signal.

2. The digital self-interference cancellation method based on enhanced neural network structure according to claim 1 is characterized in that: The following steps are also included: Get full-duplex transceiver data set; Splitting the full-duplex transceiver data set to obtain a training set and a test set; According to the training set, training the first neural network based on the second stepped grid structure to obtain a second neural network; According to the training set, training a third neural network based on the second sliding window grid structure to obtain a fourth neural network; According to the test set, a self-interference cancellation performance test is performed on the second neural network and the fourth neural network to obtain a test result.

3. The digital self-interference cancellation method based on enhanced neural network structure according to claim 1 is characterized in that: The step of performing a complex value splitting operation and a de-linearization operation on the first stepped grid structure to obtain a second stepped grid structure comprises the following steps: Splitting a first complex value of the first stepped grid structure into a first real part and a first imaginary part; Adding a first neuron to the first hidden layer of the first stepped grid structure; The original second neuron of the first stepped grid structure is combined with the first neuron, and the initial activation function of the first hidden layer is replaced with a linear activation function to obtain a second hidden layer; The second stepped grid structure is obtained according to the first real part, the first imaginary part and the second hidden layer.

4. The digital self-interference cancellation method based on enhanced neural network structure according to claim 1 is characterized in that: The step of performing a complex value splitting operation and a de-linearization operation on the first sliding window grid structure to obtain a second sliding window grid structure comprises the following steps: Splitting the second complex value of the first sliding window grid structure into a second real part and a second imaginary part; Adding a third neuron to the third hidden layer of the first sliding window grid structure; The fourth neuron originally in the first sliding window grid structure is combined with the third neuron, and the initial activation function of the third hidden layer is replaced with a linear activation function to obtain a fourth hidden layer; The second sliding window grid structure is obtained according to the second real part, the second imaginary part and the fourth hidden layer.

5. The digital self-interference cancellation method based on enhanced neural network structure according to claim 1 is characterized in that: The formula used for obtaining the first output of the dynamic phase layer according to the second stepped grid structure or the second sliding window grid structure includes: o'(n)=[o' i (n),o′2(n),…,o′ 2G-1 (n),o' 2G (n)] T ; Where o′(n) represents the first output of the dynamic phase layer; o′1(n), o′2(n),…, o′ 2G-1 (n),o′ 2G (n) represents the output of each neuron in the dynamic phase layer; G represents the number of neurons in the nonlinear hidden layer; n represents the discrete time; [·] T Represents a transpose operation.

6. The digital self-interference cancellation method based on enhanced neural network structure according to claim 1 is characterized in that: The constructing of the vector decomposition structure, according to the first output of the dynamic phase layer and the vector decomposition structure, separates the amplitude information and the phase information, and the formula used includes: Among them, s i (n) represents amplitude information; sinθ i Represents the sinusoidal phase; cosθ i represents the cosine phase; G represents the number of neurons in the nonlinear hidden layer; n represents the discrete time; o′ 2i (n), o′ 2i-1 (n) represents the output of the 2i-th and 2i-1-th neurons in the dynamic phase layer.

7. The digital self-interference cancellation method based on enhanced neural network structure according to claim 1 is characterized in that: The amplitude information is fed into a nonlinear hidden layer to obtain a second output of the nonlinear hidden layer, and the formula used includes: x′ NN (n)=[|s1(n)|,|s1(n)| 2 ,…,|s G (n)|,|s G (n)| 2 ] T ; o″ NN (n)=[o″1(n),o″2(n),…,o″ G (n)] T ; Among them, x′ NN (n) represents the input of the nonlinear hidden layer; s1(n),…,s G (n) indicates amplitude information; [·] T Indicates transpose operation; o″ NN (n) represents the second output of the nonlinear hidden layer; o″1(n), o″2(n),…,o″ G (n) represents the output of each neuron in the nonlinear hidden layer.

8. The digital self-interference cancellation method based on enhanced neural network structure according to claim 1 is characterized in that: The phase information and the second output are fused to obtain fused data, and the formula used includes: x″ NN (n)=[o″1(n)sinθ1,o″1(n)cosθ1,…,o″ G (n)sinθ G ,o″ G (n)cosθ G ] T ; Among them, x″ NN (n) represents fused data; o″1(n),…,o″ G (n) represents the output of each neuron in the nonlinear hidden layer; sinθ1,…,sinθ G Represents each sinusoidal phase; cosθ1,…,cosθ G Represents the phase of each cosine; [·] T Represents a transpose operation.

9. The digital self-interference cancellation method based on enhanced neural network structure according to claim 1 is characterized in that: According to the simulation result, the received signal is subjected to self-interference cancellation operation to obtain the target signal, and the formula used includes: Where y(n) represents the target signal; y SI (n) indicates a received signal; Represents the simulation results.

10. A digital self-interference cancellation device based on an enhanced neural network structure, characterized in that: include: The first module is used to perform a complex value splitting operation and a de-linearization operation on the first stepped grid structure to obtain a second stepped grid structure, and to perform a complex value splitting operation and a de-linearization operation on the first sliding window grid structure to obtain a second sliding window grid structure; A second module, configured to obtain a first output of a dynamic phase layer according to the second stepped grid structure or the second sliding window grid structure; A third module is used to construct a vector decomposition structure, and separate the amplitude information and the phase information according to the first output of the dynamic phase layer and the vector decomposition structure; A fourth module is used to feed the amplitude information into a nonlinear hidden layer to obtain a second output of the nonlinear hidden layer; A fifth module, configured to perform a fusion operation on the phase information and the second output to obtain fused data; The sixth module is used to feed the fused data into a fully connected output layer and output a simulation result; The seventh module is used to perform a self-interference cancellation operation on the received signal according to the simulation result to obtain a target signal.

Images