CN116707677A - Signal detection method and device and communication equipment - Google Patents
Signal detection method and device and communication equipment Download PDFInfo
- Publication number
- CN116707677A CN116707677A CN202210171753.3A CN202210171753A CN116707677A CN 116707677 A CN116707677 A CN 116707677A CN 202210171753 A CN202210171753 A CN 202210171753A CN 116707677 A CN116707677 A CN 116707677A
- Authority
- CN
- China
- Prior art keywords
- noise
- received signal
- representing
- neural network
- signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 92
- 238000004891 communication Methods 0.000 title claims abstract description 28
- 239000011159 matrix material Substances 0.000 claims abstract description 183
- 238000013528 artificial neural network Methods 0.000 claims abstract description 112
- 238000012545 processing Methods 0.000 claims abstract description 98
- 238000000034 method Methods 0.000 claims abstract description 59
- 239000010410 layer Substances 0.000 claims description 117
- 230000006870 function Effects 0.000 claims description 90
- 238000012549 training Methods 0.000 claims description 60
- 238000013527 convolutional neural network Methods 0.000 claims description 58
- 238000004590 computer program Methods 0.000 claims description 52
- 230000005540 biological transmission Effects 0.000 claims description 29
- 230000008569 process Effects 0.000 claims description 22
- 238000003860 storage Methods 0.000 claims description 14
- 239000002356 single layer Substances 0.000 claims description 13
- 238000006243 chemical reaction Methods 0.000 claims description 7
- 238000013135 deep learning Methods 0.000 abstract description 10
- 238000010586 diagram Methods 0.000 description 15
- 238000007796 conventional method Methods 0.000 description 6
- 238000007781 pre-processing Methods 0.000 description 6
- 230000004913 activation Effects 0.000 description 5
- 238000013461 design Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 238000007476 Maximum Likelihood Methods 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 210000002569 neuron Anatomy 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/309—Measuring or estimating channel quality parameters
- H04B17/336—Signal-to-interference ratio [SIR] or carrier-to-interference ratio [CIR]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B17/00—Monitoring; Testing
- H04B17/30—Monitoring; Testing of propagation channels
- H04B17/391—Modelling the propagation channel
- H04B17/3912—Simulation models, e.g. distribution of spectral power density or received signal strength indicator [RSSI] for a given geographic region
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Electromagnetism (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Noise Elimination (AREA)
Abstract
The invention provides a signal detection method, a signal detection device and communication equipment, wherein the method comprises the following steps: acquiring a noise estimation value based on a neural network; according to the noise estimation value, carrying out noise removal processing on the received signal of the receiving end to obtain a target received signal after noise removal; and acquiring an estimated value of the transmitting signal according to the target receiving signal and a channel matrix between the receiving end and the transmitting end. The invention uses the denoising network based on the neural network as the pretreatment of the received signals, can lead the traditional detector or the existing detector based on the deep learning to obtain more excellent performance under the serious noise interference condition, and does not increase the complexity of the detector.
Description
Technical Field
The present invention relates to the field of communications technologies, and in particular, to a signal detection method, a signal detection device, and a communication device.
Background
The traditional signal detection method is mainly divided into two main types of linear detectors and nonlinear detectors. The traditional linear detector mainly comprises zero forcing detection (ZF algorithm), minimum mean square error detection (LMMSE algorithm) and the like, and has lower detection complexity but poorer detection performance; nonlinear detection includes methods such as maximum likelihood detection (ML) and Sphere Decoding (SD), and has high detection accuracy, but also has high complexity.
In the low signal-to-noise region, it is often difficult for conventional detectors to achieve good detection performance due to noise effects. Conventional detectors, such as maximum likelihood detectors, often do not have a modular design for noise removal, and thus it is difficult to eliminate the effect of noise on detection performance.
Disclosure of Invention
The invention aims to provide a signal detection method, a signal detection device and communication equipment, which solve the problem of poor detection performance of the existing signal detection mode.
The embodiment of the invention provides a signal detection method, which comprises the following steps:
acquiring a noise estimation value based on a neural network;
according to the noise estimation value, carrying out noise removal processing on the received signal of the receiving end to obtain a target received signal after noise removal;
and acquiring an estimated value of the transmitting signal according to the target receiving signal and a channel matrix between the receiving end and the transmitting end.
Optionally, the acquiring the noise estimation value based on the neural network includes:
acquiring an initial estimated value of noise;
taking the initial estimated value of the noise, the received signal and the channel matrix as input feature matrices of a convolutional neural network, and performing M-layer convolutional processing on the input feature matrices through the convolutional neural network to obtain the estimated value of the noise;
The output result of each of the M-layer convolution processes is a single-layer estimated value of noise, M is greater than or equal to 2, and M is a positive integer.
Optionally, the obtaining the initial estimated value of the noise includes:
acquiring an initial estimated value of a transmission signal according to the received signal and the channel matrix;
and acquiring the initial estimated value of the noise according to the initial estimated value of the transmitting signal, the receiving signal and the channel matrix.
Optionally, the obtaining an initial estimated value of the transmission signal according to the received signal and the channel matrix includes:
the initial estimate of the transmitted signal is calculated by the following formula:
wherein ,representing an initial estimate of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, diag (H) T H) Representation matrix H T And a diagonal matrix of diagonal elements of H, y representing the received signal.
Optionally, the obtaining the initial estimated value of the noise according to the initial estimated value of the transmission signal, the received signal and the channel matrix includes:
the initial estimate of the noise is calculated by the following formula:
wherein ,representing an initial estimate of said noise, y representing said received signal, H representing said channel matrix,/or->Representing an initial estimate of the transmitted signal.
Optionally, taking the initial estimated value of the noise, the received signal and the channel matrix as input feature matrices of a convolutional neural network, performing M-layer convolutional processing on the input feature matrices through the convolutional neural network, and obtaining the estimated value of the noise includes:
carrying out convolution processing on the input feature matrix by utilizing a one-dimensional convolution neural network to obtain a first layer estimated value of noise;
according to the first layer estimated value, noise removal processing is carried out on the received signal, and a first layer parameter is obtained;
carrying out convolution processing on the input feature matrix and the first layer parameters by using a one-dimensional convolution neural network to obtain a second layer estimated value of noise;
according to the second layer estimated value, noise removing processing is carried out on the received signal, and a second layer parameter is obtained;
and carrying out convolution processing on the input feature matrix and the second layer parameters by using a one-dimensional convolution neural network to obtain the noise estimation value.
Optionally, taking the initial estimated value of the noise, the received signal and the channel matrix as input feature matrices of a neural network, performing M-layer convolution processing on the input feature matrices through the convolution neural network, to obtain the estimated value of the noise, including:
the noise estimate is calculated by the following formula:
wherein ,representing the noise estimate,/->y represents the jointReceiving signals, H representing said channel matrix, < >>Representing an initial estimate of the noise, conv1 (·) represents a one-dimensional convolutional neural network.
Optionally, according to the noise estimation value, noise removing processing is performed on the received signal at the receiving end to obtain a target received signal after noise removal, including:
subtracting the noise estimation value from the received signal to obtain a target received signal after noise removal; wherein, the target received signal after noise removal is calculated by the following formula:
y es representing the target received signal, y representing the received signal,representing the noise estimate.
Optionally, obtaining the estimated value of the transmission signal according to the target reception signal and the channel matrix between the receiving end and the transmitting end includes:
And taking the target received signal and the channel matrix as the input of a zero forcing signal, and calculating the estimated value of the transmitted signal by using a zero forcing detection algorithm.
Optionally, the calculating, using a zero forcing detection algorithm, the target received signal and the channel matrix as inputs of the zero forcing signal, the estimated value of the transmitted signal includes:
calculating an estimated value of the transmission signal by the following formula:
wherein ,representing an estimated value of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, sgn (·) is the sign function, y es Representing the target received signal.
Optionally, the method further comprises:
acquiring a loss function of the neural network based on the training data;
and training the neural network according to the loss function.
Optionally, the acquiring the loss function of the neural network based on the training data includes:
performing cascaded linear and nonlinear conversion on training data to obtain a noise estimation value;
and comparing the noise estimated value with a noise label value to obtain the loss function.
Optionally, the acquiring the loss function of the neural network based on the training data includes:
The loss function is obtained by the following formula:
wherein ,represents L 2 A norm loss function, B representing the batch size of the training data, +.>Represents a noise estimation value, n represents a noise label value, Ω represents a parameter requiring training, II 2 Representing a 2-norm.
Optionally, training the neural network according to the loss function includes:
calculating a gradient of the loss function;
and optimizing the loss function through a back propagation algorithm, and updating parameters of the neural network until the neural network converges.
The embodiment of the invention also provides a communication device, which comprises: memory, transceiver, processor:
a memory for storing a computer program; a transceiver for transceiving data under control of the processor; a processor for reading the computer program in the memory and performing the following operations:
acquiring a noise estimation value based on a neural network;
according to the noise estimation value, carrying out noise removal processing on the received signal of the receiving end to obtain a target received signal after noise removal;
and acquiring an estimated value of the transmitting signal according to the target receiving signal and a channel matrix between the receiving end and the transmitting end.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
acquiring an initial estimated value of noise;
taking the initial estimated value of the noise, the received signal and the channel matrix as input feature matrices of a convolutional neural network, and performing M-layer convolutional processing on the input feature matrices through the convolutional neural network to obtain the estimated value of the noise;
the output result of each of the M-layer convolution processes is a single-layer estimated value of noise, M is greater than or equal to 2, and M is a positive integer.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
acquiring an initial estimated value of a transmission signal according to the received signal and the channel matrix;
and acquiring the initial estimated value of the noise according to the initial estimated value of the transmitting signal, the receiving signal and the channel matrix.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
the initial estimate of the transmitted signal is calculated by the following formula:
wherein ,representing an initial estimate of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, diag (H) T H) Representation matrix H T And a diagonal matrix of diagonal elements of H, y representing the received signal.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
the initial estimate of the noise is calculated by the following formula:
wherein ,representing an initial estimate of said noise, y representing said received signal, H representing said channel matrix,/or->Representing an initial estimate of the transmitted signal.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
carrying out convolution processing on the input feature matrix by utilizing a one-dimensional convolution neural network to obtain a first layer estimated value of noise;
according to the first layer estimated value, noise removal processing is carried out on the received signal, and a first layer parameter is obtained;
carrying out convolution processing on the input feature matrix and the first layer parameters by using a one-dimensional convolution neural network to obtain a second layer estimated value of noise;
according to the second layer estimated value, noise removing processing is carried out on the received signal, and a second layer parameter is obtained;
and carrying out convolution processing on the input feature matrix and the second layer parameters by using a one-dimensional convolution neural network to obtain the noise estimation value.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
the noise estimate is calculated by the following formula:
wherein ,representing the noise estimate,/->y represents the received signal, H represents the channel matrix,>representing an initial estimate of the noise, conv1 (·) represents a one-dimensional convolutional neural network.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
subtracting the noise estimation value from the received signal to obtain a target received signal after noise removal; wherein, the target received signal after noise removal is calculated by the following formula:
y es representing the target received signal, y representing the received signal,representing the noise estimate.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
and taking the target received signal and the channel matrix as the input of a zero forcing signal, and calculating the estimated value of the transmitted signal by using a zero forcing detection algorithm.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
Calculating an estimated value of the transmission signal by the following formula:
wherein ,representing an estimated value of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, sgn (·) is the sign function, y es Representing the target received signal.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
acquiring a loss function of the neural network based on the training data;
and training the neural network according to the loss function.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
performing cascaded linear and nonlinear conversion on training data to obtain a noise estimation value;
and comparing the noise estimated value with a noise label value to obtain the loss function.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
the loss function is obtained by the following formula:
wherein ,represents L 2 A norm loss function, B representing the batch size of the training data, +.>Represents a noise estimation value, n represents a noise label value, Ω represents a parameter requiring training, II 2 Representing a 2-norm.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
calculating a gradient of the loss function;
and optimizing the loss function through a back propagation algorithm, and updating parameters of the neural network until the neural network converges.
An embodiment of the present application provides a signal detection apparatus including:
a first acquisition unit configured to acquire a noise estimation value based on a neural network;
the first processing unit is used for removing noise from the received signal of the receiving end according to the noise estimation value to obtain a target received signal after noise removal;
and the second acquisition unit is used for acquiring the estimated value of the transmission signal according to the target received signal and the channel matrix between the receiving end and the transmitting end.
An embodiment of the application provides a processor-readable storage medium having stored thereon a computer program, characterized in that the computer program when executed by a processor implements the steps of the signal detection method described above.
The technical scheme of the application has the beneficial effects that:
according to the embodiment of the application, the noise estimation value is obtained based on the neural network, the noise is removed from the received signal, the signal detection is carried out by utilizing the denoised received signal and the channel matrix between the receiving end and the transmitting end, the estimation value of the transmitted signal is obtained, and the complexity of the signal detection is reduced in the process. The denoising network based on the neural network is used for preprocessing the received signals, so that the signal-to-noise ratio of signal detection can be improved, the detection performance of a traditional detector or a detector based on deep learning when noise interference is serious is improved, and meanwhile, the complexity of the detector is not increased.
Drawings
FIG. 1 is a flow chart of a signal detection method according to an embodiment of the invention;
FIG. 2 is a schematic diagram of a convolutional neural network according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a convolutional neural network corresponding to the module of FIG. 2 in accordance with an embodiment of the present invention;
FIG. 4 is a schematic diagram showing the performance of the denoising network and the conventional method according to the embodiment of the present invention as a function of signal-to-noise ratio;
FIG. 5 is a second diagram showing the performance of the denoising network and conventional method according to the embodiment of the present invention as a function of signal-to-noise ratio;
FIG. 6 is a third diagram showing the performance of the denoising network and conventional method according to the embodiment of the present invention as a function of signal-to-noise ratio;
fig. 7 is a schematic structural diagram of a signal detection device according to an embodiment of the present invention;
fig. 8 is a schematic structural diagram of a communication device according to an embodiment of the present invention.
Detailed Description
In order to make the technical problems, technical solutions and advantages to be solved more apparent, the following detailed description will be given with reference to the accompanying drawings and specific embodiments. In the following description, specific details such as specific configurations and components are provided merely to facilitate a thorough understanding of embodiments of the invention. It will therefore be apparent to those skilled in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. In addition, descriptions of well-known functions and constructions are omitted for clarity and conciseness.
It should be appreciated that reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present application. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
In various embodiments of the present application, it should be understood that the sequence numbers of the following processes do not mean the order of execution, and the order of execution of the processes should be determined by the functions and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.
In the embodiment of the application, the term "and/or" describes the association relation of the association objects, which means that three relations can exist, for example, a and/or B can be expressed as follows: a exists alone, A and B exist together, and B exists alone. The character "/" generally indicates that the context-dependent object is an "or" relationship.
The term "plurality" in embodiments of the present application means two or more, and other adjectives are similar.
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments, but not all embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
Specifically, the embodiment of the application provides a signal detection method, a signal detection device and communication equipment, which solve the problem of higher complexity of detection in the existing signal detection mode.
As shown in fig. 1, an embodiment of the present application provides a signal detection method, which specifically includes the following steps:
step 101, obtaining a noise estimation value based on a neural network.
In this embodiment, the neural network may be a convolutional neural network. By designing the neural network, the noise in the signal detection process can be estimated. The signal detection is to detect the transmitting signal of the transmitting end according to the array receiving signal of the receiving end and the channel matrix between the receiving end and the transmitting end. The expression of the receiving signals of the receiving end array is as follows:
y＝Hx+n
Wherein y represents a received signal of the receiving end, H represents a channel matrix between the receiving end and the transmitting end, x represents a transmitted signal of the transmitting end, and n represents a noise vector. The signal detection is based on y and H detection x. To obtain an estimate of the x, a noise estimate needs to be obtained first.
And 102, performing noise removal processing on the received signal of the receiving end according to the noise estimation value to obtain a target received signal after noise removal. After the noise estimation value is obtained, denoising is carried out on the received signal, namely noise is subtracted from the received signal, so that the influence caused by the noise is restrained.
Step 103, obtaining an estimated value of the sending signal according to the target receiving signal and a channel matrix between the receiving end and the sending end.
After denoising the received signal, performing signal detection according to the denoised received signal (i.e. the target received signal) and the signal matrix to obtain an estimated value of the transmitted signal.
In this embodiment, the transmitting end refers to a transmitting end of a signal, and the receiving end refers to a receiving end of the signal, that is, the transmitting end transmits the signal to the receiving end, and after the receiving end receives the signal, the receiving end obtains a noise estimation value based on a neural network, and performs denoising processing on the received signal to obtain the target received signal; and estimating the transmitting signal transmitted by the transmitting end according to the target receiving signal and the channel matrix obtained after denoising, so as to realize signal detection.
The transmitting end and the receiving end may be terminals or network side devices, for example: when uplink signal detection is performed, the transmitting end is a terminal, and the receiving end is network side equipment (such as a base station); when downlink signal detection is performed, the transmitting end is network side equipment (such as a base station), and the receiving end is a terminal.
According to the embodiment of the application, the noise estimation value is obtained based on the neural network, the noise is removed from the received signal, the signal detection is carried out by utilizing the denoised received signal and the channel matrix between the receiving end and the transmitting end, the estimation value of the transmitted signal is obtained, and the complexity of the signal detection is reduced in the process. The denoising network based on the neural network is used for preprocessing the received signals, so that the signal-to-noise ratio of signal detection can be improved, the detection performance of a traditional detector or a detector based on deep learning when noise interference is serious is improved, and meanwhile, the complexity of the detector is not increased.
According to the embodiment, the received signals are subjected to denoising processing, the received signals after denoising processing and the channel matrix are used for signal detection, so that the signal to noise ratio of the signal detection can be improved.
Optionally, the acquiring the noise estimation value based on the neural network includes:
step 111, an initial estimated value of noise is obtained.
In acquiring the noise estimation value based on the neural network, it is necessary to first acquire an initial estimation value of noise, which is rough. Optionally, the obtaining the initial estimated value of the noise includes: acquiring an initial estimated value of a transmission signal according to the received signal and the channel matrix; and acquiring the initial estimated value of the noise according to the initial estimated value of the transmitting signal, the receiving signal and the channel matrix.
In order to obtain the initial estimated value of the noise and reduce the computation complexity of the method, the method uses a zero forcing detection algorithm to obtain the initial estimated value of the transmission signal by inputting a reception signal and a channel matrix as zero forcing signals, and the method comprises the steps of: the initial estimate of the transmitted signal is calculated by the following formula:
wherein ,representing an initial estimate of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, diag (H) T H) Representation matrix H T And a diagonal matrix of diagonal elements of H, y representing the received signal.
Calculating the initial estimated value of the noise according to the initial estimated value of the transmission signal, optionally, obtaining the initial estimated value of the noise according to the initial estimated value of the transmission signal, the reception signal and the channel matrix includes:
the initial estimate of the noise is calculated by the following formula:
wherein ,representing an initial estimate of said noise, y representing said received signal, H representing said channel matrix,/or->Representation ofAn initial estimate of the transmitted signal.
112, taking the initial estimated value of the noise, the received signal and the channel matrix as input feature matrices of a convolutional neural network, and performing M-layer convolutional processing on the input feature matrices through the convolutional neural network to obtain the estimated value of the noise;
the output result of each of the M-layer convolution processes is a single-layer estimated value of noise, M is greater than or equal to 2, and M is a positive integer.
After obtaining the initial estimate of the noise, a noise estimate of the signal is calculated based on the initial estimate of the noise, the noise estimate being more accurate relative to the initial estimate of the noise.
Wherein a convolutional neural network may be utilized for noise estimation. And taking the initial estimated value of the noise, the received signal and the channel matrix as input characteristic matrices of the convolutional neural network. The convolutional neural network can adopt a multi-head processing mechanism, namely, the convolutional neural network with a plurality of identical modules can realize multi-layer convolutional processing of the input feature matrix. As shown in fig. 2, taking M as 3 as an example, that is, 3 processing modules may be included in the convolutional neural network, each processing module performs the convolutional processing as shown in fig. 3. The output result of each layer of convolution processing is a single-layer estimated value of noise, and the final noise estimated value is obtained by continuously optimizing the single-layer estimated value.
Optionally, the single-layer estimated value output by each layer of convolution processing is utilized to perform denoising processing on the received signal, the obtained denoising signal and the input feature matrix are used as input features of the next layer of convolution processing, so that a new noise estimated value is obtained, and the noise estimated value is output more accurately through continuous optimization of the noise estimated value.
It should be noted that fig. 2 is a schematic diagram of a convolutional neural network with M equal to 3 as an example, and further layers of convolutional processing modules may be provided according to the processing requirement for noise estimation, which is not limited herein.
Optionally, taking the initial estimated value of the noise, the received signal and the channel matrix as input feature matrices of a convolutional neural network, performing M-layer convolutional processing on the input feature matrices through the convolutional neural network, and obtaining the estimated value of the noise includes:
carrying out convolution processing on the input feature matrix by utilizing a one-dimensional convolution neural network to obtain a first layer estimated value of noise; according to the first layer estimated value, noise removal processing is carried out on the received signal, and a first layer parameter is obtained; carrying out convolution processing on the input feature matrix and the first layer parameters by using a one-dimensional convolution neural network to obtain a second layer estimated value of noise; according to the second layer estimated value, noise removing processing is carried out on the received signal, and a second layer parameter is obtained; and carrying out convolution processing on the input feature matrix and the second layer parameters by using a one-dimensional convolution neural network to obtain the noise estimation value.
In this embodiment, the first layer estimated value and the second layer estimated value are single layer estimated values of the noise. The convolutional neural network comprises three layers of convolutional processing structures, noise estimated values (namely, single-layer estimated values) can be output after each layer of convolutional processing, the noise estimated values are utilized to carry out denoising processing on the received signals, denoised parameters (namely, denoised received signals) are obtained, the denoised parameters and the input feature matrix are utilized to carry out convolutional processing of the next layer, and final noise estimated values are obtained after three layers of convolutional processing, and the noise estimated values are optimized for multiple times, so that the convolutional neural network is more accurate.
Optionally, taking the initial estimated value of the noise, the received signal and the channel matrix as input feature matrices of a convolutional neural network, performing M-layer convolutional processing on the input feature matrices through the convolutional neural network, and obtaining the estimated value of the noise includes:
the noise estimate is calculated by the following formula:
wherein ,representing the noise estimate, inputs representing an input feature matrix of the convolutional neural network,y represents the received signal, H represents the channel matrix,>representing an initial estimate of the noise, conv1 (·) represents a one-dimensional convolutional neural network.
Alternatively, the process may be carried out in a single-stage,
where NET (-) can represent a network function and Ω is a parameter that the neural network needs to train.
The convolutional neural network structure adopted by the embodiment of the application is shown in fig. 2, wherein the neural network structure comprises three layers: the total number of layers of the neural network is l=3. The initial estimated value of the noise, the received signal and the channel matrix are used as the input of the convolutional neural network, and are input in an input layer as shown in fig. 2, and the noise estimated value is output in an output layer through the processing of an implicit layer.
The neural network may include three processing modules, where each processing module corresponds to a convolutional neural network shown in fig. 3, and after the initial estimated value of the noise, the received signal and the channel matrix are input into the convolutional neural network, each module may perform feature processing as shown in fig. 3, that is: firstly, processing inputs by using the one-dimensional convolutional neural network (comprising convolution and activation function operation), taking an obtained output result as a first layer estimated value of noise, and denoising the received signal by using the first layer estimated value to obtain a first layer parameter; the first layer parameters and inputs are used as the input of the next layer, the one-dimensional convolutional neural network is used for processing the first layer parameters and the inputs, the obtained output result is used as a second layer estimated value of noise, and the second layer estimated value is used for denoising the received signal to obtain a second layer parameter; and using the second layer parameters and the inputs as the input of the next layer, and processing the second layer parameters and the inputs by using a one-dimensional convolutional neural network to obtain a final noise estimation value.
In the embodiment, each one-dimensional convolutional neural network is used as a single-layer estimated value of noise, and after denoising the received signal, the received signal and the input of the upper layer are used as the input of the lower layer at the same time. For the input layer and the hidden layer, its activation function is leak ReLu, as follows:
wherein ,ai Is represented by the formula (1), + -infinity), a fixed parameter between. It should be noted that the above activation functions are only exemplary, and other activation functions may be selected according to the requirement.
As shown in fig. 3, for the first layer convolutional neural network, in order to reduce the randomness of regression and reduce the training cost, a multi-head mechanism is adopted, that is, the convolutional neural network with the same module is adopted, and the sum of the outputs of each module is taken as the output. Each module has the same input, the same number of neurons, and the same dimension of output. The input matrix is first passed through a one-dimensional convolutional neural network, the primary feature matrix is extracted, the filter of the convolutional neural network can be set to 512, and the size of the convolutional kernel is 1. In order to further extract the characteristics of the input matrix, increase the diversity of the matrix, convert the input characteristics from low dimension to higher dimension characteristics, input the primary characteristic matrix into two identical convolutional neural networks, the filter and the convolutional kernel can be set to 1, extract the characteristic matrix with the same dimension, and superimpose the two matrices as a new characteristic matrix by using the python broadcasting mechanism. Finally, in order to meet the dimension requirement of the output matrix, namely, the dimension of the output matrix is the same as that of the label of the regression noise, a new feature matrix is taken as input, a one-dimensional convolutional neural network is adopted, a filter and a convolutional kernel of the one-dimensional convolutional neural network can be set to be 1, and output after the filter ReLu is taken as an activation function is taken as an output result of each module.
It should be noted that, in the convolutional neural network, the input layer, the hidden layer, and the output layer are different from the M-layer convolutional processing, as shown in fig. 2, the input feature matrix is input in the input layer, the hidden layer includes 3 convolutional processing modules, and the convolutional processing shown in fig. 3 is performed respectively, and the noise estimation value is output in the output layer.
Optionally, according to the noise estimation value, noise removing processing is performed on the received signal at the receiving end to obtain a target received signal after noise removal, including: and subtracting the noise estimation value from the received signal to obtain a target received signal after noise removal.
In this embodiment, the input of the neural network is the noise estimation value, and after the noise estimation value is obtained by using the neural network, the noise of the received signal is removed by using the estimation value, where the target received signal after removing the noise is calculated by the following formula:
wherein ,yes I.e., the noise-removed received signal, i.e., the target received signal, y represents the received signal,representing the noise estimate.
After denoising the received signal, the embodiment of the application obtains the target received signal, and signal detection can be performed by using the target received signal and a channel matrix, and signal estimation can be performed based on the existing signal detection algorithm, for example: conventional detection algorithms (e.g., zero-forcing detection algorithm, maximum likelihood detection algorithm), deep learning-based detection algorithms (e.g., detNet algorithm), and the like. The embodiment of the application carries out denoising treatment on the received signal, and carries out signal detection by using the denoised received signal, thereby improving the signal-to-noise ratio of detection and the detection performance.
The denoising network in this embodiment is portable (i.e., can be applied to any signal detection algorithm), and the designed denoising network is used as preprocessing for the received signal, so that a traditional detector (such as a zero forcing detection algorithm and a maximum likelihood detection algorithm) or a detector based on deep learning (such as a DetNet) can obtain more excellent performance under the serious condition of noise interference without substantially increasing the complexity of the detector.
Optionally, obtaining the estimated value of the transmission signal according to the target reception signal and the channel matrix between the receiving end and the transmitting end includes: and taking the target received signal and the channel matrix as the input of the zero forcing signal, and calculating the estimated value of the transmitted signal by using a zero forcing detection algorithm, namely calculating the estimated value of the transmitted signal according to the channel matrix and the denoised received signal.
The calculating, using a zero forcing detection algorithm, an estimated value of the transmission signal with the target reception signal and the channel matrix as inputs of the zero forcing signal includes:
calculating an estimated value of the transmission signal by the following formula:
wherein ,representing an estimated value of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, sgn (·) is the sign function, y es Representing the target received signal.
In this embodiment, the denoised received signal and the channel matrix are used to obtain an estimate of the signal by a zero forcing detection algorithm, wherein,
m represents the input of the sign function sgn (.).
As an alternative embodiment, the method further comprises:
and step 01, acquiring a loss function of the neural network based on the training data.
Optionally, training the neural network with training data is required before noise estimation based on the nerve, wherein the obtaining the loss function of the neural network based on the training data includes: performing cascaded linear and nonlinear conversion on training data to obtain a noise estimation value; and comparing the noise estimated value with a noise label value to obtain the loss function.
Optionally, the acquiring the loss function of the neural network based on the training data includes:
the loss function is obtained by the following formula:
wherein ,represents L 2 A norm loss function, B representing the batch size of the training data, +.>Represents a noise estimation value, n represents a noise label value, Ω represents a parameter requiring training, II 2 Representing a 2-norm.
In this embodiment, in order to train the neural network, training parameters of the network need to be defined. For the regression network to be of the type,the number of neurons of the output layer is 1, the dimension of the output matrix is Nx1, and the loss function is L 2 And a norm function.
And step 02, training the neural network according to the loss function.
Optionally, training the neural network according to the loss function includes: calculating a gradient of the loss function; and optimizing the loss function through a back propagation algorithm, and updating parameters of the neural network until the neural network converges.
In this embodiment, after the loss function is determined, the loss function is continuously optimized and updated through a training process, so as to obtain a converged neural network, and noise estimation can be performed based on the converged neural network. In order to prevent the neural network from being overfitted in the training process, and make the neural network excessively depend on partial characteristics in the forward transmission process, and increase generalization of the model, a Dropout algorithm is adopted, namely, the number of partial neurons is reserved with a certain probability in the training process, and the algorithm is used after the superposition of two characteristic matrixes as shown in fig. 3. Meanwhile, the Dropout algorithm is not used in the test stage.
Optionally, the embodiment deep neural network based algorithm may include a training phase, a verification phase, and a testing phase. During the training phase, the network derives a noisy predicted output by cascading linear and nonlinear transformations of the input data (i.e., the) The predicted output is compared to the tag value (i.e., n) to obtain a loss function. The gradient of the loss function can be calculated through an ADAM algorithm, and the loss function of the network is gradually optimized through back propagation, so that the parameters of the network are continuously optimized and updated until the neural network converges. Alternatively, the neural network verification, i.e., the verification phase, may be performed once per training a predetermined number of times (e.g., 10 times). In the verification and test stage, the parameters of the network are fixed, the neural network estimates noise according to the input data by inputting the test data set, and the noise part in the received signal is removed, thereby obtaining the estimation of the signalValues.
Compared with the traditional detector, the embodiment of the application does not need to design a complex optimization function to obtain a derived model, but designs a denoising network based on deep learning, and can obtain excellent performance through the design and training of a neural network; meanwhile, in the prior art, a model is deduced based on a formula, when a scene changes, the model is deduced, and the denoising model can be obtained only by defining and changing a label of the neural network for training. Therefore, compared with the conventional detector, the signal detection method of the embodiment is simple to operate, good in generalization and excellent in performance. The following description will be made by comparing with the existing signal detection method.
As shown in fig. 4, a schematic diagram of the performance of the denoising network and the conventional method according to the embodiment of the present application along with the change of the signal to noise ratio is shown, wherein the simulated channel scene is a rayleigh channel, the change range of the signal to noise ratio is designed to be [ -20, -2] db, and the simulation is performed once for every two signal to noise ratios. As can be seen from fig. 4, when the transceiver antenna array is (128, 8), the performance based on the denoising neural network is slightly better than that of the existing ZF algorithm, and since the linear detector has a large loading coefficient (N > > K) in a large-scale multiple-input multiple-output (Multiple In Multiple Out, MIMO) scenario, the performance of the linear detector is only inferior to that of the ML algorithm, N is the number of receiving end array antennas, and K is the number of transmitting end array antennas. Therefore, the method provided by the embodiment of the application has obvious performance advantages.
As shown in fig. 5, the performance of the denoising network and the conventional method provided by the embodiment of the present application changes with the signal-to-noise ratio when the transceiver antenna array is (128, 64), and the rest of simulation conditions are the same as those of fig. 4. As can be seen from fig. 5, the method based on the denoising neural network provided by the embodiment of the present application is significantly better than the performance of the conventional ZF method, and the improvement of the performance of the method is most obvious under the condition that the signal-to-noise ratio is [ -10, -2] db.
As shown in fig. 6, the performance of the denoising network and the conventional method provided by the embodiment of the present application changes with the signal-to-noise ratio when the transceiver antenna array is (128, 74), and the rest of simulation conditions are the same as those of fig. 4. As can be obtained from fig. 6, the method based on the denoising neural network provided by the embodiment of the application is significantly better than the performance of the traditional ZF method, and compared with fig. 5, the method has more obvious performance advantage when the loading coefficient ρ=n/K of the antenna is small.
In the above embodiment, compared with the existing deep learning-based detector, the neural network based on the denoising design of the embodiment of the application has more excellent performance; meanwhile, the neural network provided by the embodiment of the application does not need a complex training process, and is short in training time, small in parameter quantity and low in complexity.
According to the embodiment of the application, the noise estimation value is obtained based on the neural network, the noise is removed from the received signal, the signal detection is carried out by utilizing the denoised received signal and the channel matrix between the receiving end and the transmitting end, the estimation value of the transmitted signal is obtained, and the complexity of the signal detection is reduced in the process. The denoising network based on the neural network is used for preprocessing the received signals, so that the signal-to-noise ratio of signal detection can be improved, the detection performance of a traditional detector or a detector based on deep learning when noise interference is serious is improved, and meanwhile, the complexity of the detector is not increased.
The above embodiments are described with respect to the signal detection method of the present invention, and the following embodiments will further describe the corresponding apparatus with reference to the accompanying drawings.
Specifically, as shown in fig. 7, an embodiment of the present invention provides a signal detection apparatus 700, including:
a first acquisition unit 710 for acquiring a noise estimation value based on a neural network;
a first processing unit 720, configured to perform noise removal processing on the received signal at the receiving end according to the noise estimation value, so as to obtain a target received signal after noise removal;
a second obtaining unit 730, configured to obtain an estimated value of the transmission signal according to the target reception signal and a channel matrix between the receiving end and the transmitting end.
Optionally, the first acquisition unit includes:
a first acquisition subunit, configured to acquire an initial estimated value of noise;
the second obtaining subunit is configured to take the initial estimated value of the noise, the received signal and the channel matrix as an input feature matrix of a convolutional neural network, and perform M-layer convolution processing on the input feature matrix through the convolutional neural network to obtain the estimated value of the noise;
the output result of each of the M-layer convolution processes is a single-layer estimated value of noise, M is greater than or equal to 2, and M is a positive integer.
Optionally, the first obtaining subunit is specifically configured to:
acquiring an initial estimated value of a transmission signal according to the received signal and the channel matrix;
and acquiring the initial estimated value of the noise according to the initial estimated value of the transmitting signal, the receiving signal and the channel matrix.
Optionally, the first obtaining subunit obtains, according to the received signal and the channel matrix, an initial estimated value of a transmission signal, including:
the initial estimate of the transmitted signal is calculated by the following formula:
wherein ,representing an initial estimate of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, diag (H) T H) Representation matrix H T And a diagonal matrix of diagonal elements of H, y representing the received signal.
Optionally, the first obtaining subunit obtains the initial estimated value of the noise according to the initial estimated value of the transmission signal, the received signal and the channel matrix, including:
the initial estimate of the noise is calculated by the following formula:
wherein ,representing an initial estimate of said noise, y representing said received signal, H representing said channel matrix,/or- >Representing an initial estimate of the transmitted signal.
Optionally, the second acquisition subunit is specifically configured to:
carrying out convolution processing on the input feature matrix by utilizing a one-dimensional convolution neural network to obtain a first layer estimated value of noise;
according to the first layer estimated value, noise removal processing is carried out on the received signal, and a first layer parameter is obtained;
carrying out convolution processing on the input feature matrix and the first layer parameters by using a one-dimensional convolution neural network to obtain a second layer estimated value of noise;
according to the second layer estimated value, noise removing processing is carried out on the received signal, and a second layer parameter is obtained;
and carrying out convolution processing on the input feature matrix and the second layer parameters by using a one-dimensional convolution neural network to obtain the noise estimation value.
Optionally, the second acquisition subunit is specifically configured to:
the noise estimate is calculated by the following formula:
wherein ,representing the noise estimate,/->y represents the received signal, H represents the channel matrix,>representing an initial estimate of the noise, conv1 (·) represents a one-dimensional convolutional neural network.
Optionally, the first processing unit is specifically configured to:
Subtracting the noise estimation value from the received signal to obtain a target received signal after noise removal; wherein, the target received signal after noise removal is calculated by the following formula:
y es representing the target received signal, y representing the received signal,representing the noise estimate.
Optionally, the second acquisition unit includes:
and the third acquisition subunit is used for taking the target received signal and the channel matrix as inputs of zero forcing signals and calculating the estimated value of the transmitted signals by using a zero forcing detection algorithm.
Optionally, the third obtaining subunit is specifically configured to:
calculating an estimated value of the transmission signal by the following formula:
wherein ,representing an estimated value of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, sgn (·) is the sign function, y es Representing the target received signal.
Optionally, the apparatus further comprises:
a third acquisition unit configured to acquire a loss function of the neural network based on the training data;
and the training unit is used for training the neural network according to the loss function.
Optionally, the third obtaining unit is specifically configured to:
performing cascaded linear and nonlinear conversion on training data to obtain a noise estimation value;
And comparing the noise estimated value with a noise label value to obtain the loss function.
Optionally, the third obtaining unit is specifically configured to:
the loss function is obtained by the following formula:
wherein ,represents L 2 A norm loss function, B representing the batch size of the training data, +.>Represents a noise estimation value, n represents a noise label value, Ω represents a parameter requiring training, II 2 Representing a 2-norm.
Optionally, the training unit is specifically configured to:
calculating a gradient of the loss function;
and optimizing the loss function through a back propagation algorithm, and updating parameters of the neural network until the neural network converges.
According to the embodiment of the application, the noise estimation value is obtained based on the neural network, the noise is removed from the received signal, the signal detection is carried out by utilizing the denoised received signal and the channel matrix between the receiving end and the transmitting end, the estimation value of the transmitted signal is obtained, and the complexity of the signal detection is reduced in the process. The denoising network based on the neural network is used for preprocessing the received signals, so that the signal-to-noise ratio of signal detection can be improved, the detection performance of a traditional detector or a detector based on deep learning when noise interference is serious is improved, and meanwhile, the complexity of the detector is not increased.
It should be noted that, the above device provided in the embodiment of the present application can implement all the method steps implemented in the method embodiment and achieve the same technical effects, and detailed descriptions of the same parts and beneficial effects as those in the method embodiment in this embodiment are omitted.
It should be noted that, in the embodiment of the present application, the division of the units is schematic, which is merely a logic function division, and other division manners may be implemented in actual practice. In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a processor-readable storage medium. Based on such understanding, the technical solution of the present application may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) or a processor (processor) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
As shown in fig. 8, an embodiment of the present invention further provides a communication device, where the communication device is a receiving end of a signal, and the communication device may be a terminal or a network side device, and the communication device includes: memory 820, transceiver 800, processor 810;
a memory 820 for storing a computer program; a transceiver 800 for transceiving data under the control of the processor; a processor 810 for reading the computer program in the memory and performing the following operations:
acquiring a noise estimation value based on a neural network;
according to the noise estimation value, carrying out noise removal processing on the received signal of the receiving end to obtain a target received signal after noise removal;
and acquiring an estimated value of the transmitting signal according to the target receiving signal and a channel matrix between the receiving end and the transmitting end.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
acquiring an initial estimated value of noise;
taking the initial estimated value of the noise, the received signal and the channel matrix as input feature matrices of a convolutional neural network, and performing M-layer convolutional processing on the input feature matrices through the convolutional neural network to obtain the estimated value of the noise;
The output result of each of the M-layer convolution processes is a single-layer estimated value of noise, M is greater than or equal to 2, and M is a positive integer.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
acquiring an initial estimated value of a transmission signal according to the received signal and the channel matrix;
and acquiring the initial estimated value of the noise according to the initial estimated value of the transmitting signal, the receiving signal and the channel matrix.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
the initial estimate of the transmitted signal is calculated by the following formula:
wherein ,representing an initial estimate of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, diag (H) T H) Representation matrix H T And a diagonal matrix of diagonal elements of H, y representing the received signal.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
the initial estimate of the noise is calculated by the following formula:
wherein ,representing an initial estimate of said noise, y representing said received signal, H representing said channel matrix,/or- >Representing an initial estimate of the transmitted signal.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
carrying out convolution processing on the input feature matrix by utilizing a one-dimensional convolution neural network to obtain a first layer estimated value of noise;
according to the first layer estimated value, noise removal processing is carried out on the received signal, and a first layer parameter is obtained;
carrying out convolution processing on the input feature matrix and the first layer parameters by using a one-dimensional convolution neural network to obtain a second layer estimated value of noise;
according to the second layer estimated value, noise removing processing is carried out on the received signal, and a second layer parameter is obtained;
and carrying out convolution processing on the input feature matrix and the second layer parameters by using a one-dimensional convolution neural network to obtain the noise estimation value.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
the noise estimate is calculated by the following formula:
wherein ,representing the noise estimate,/->y represents the received signal, H represents the channel matrix,>representing an initial estimate of the noise, conv1 (·) represents a one-dimensional convolutional neural network.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
subtracting the noise estimation value from the received signal to obtain a target received signal after noise removal; wherein, the target received signal after noise removal is calculated by the following formula:
y es representing the target received signal, y representing the received signal,representing the noise estimate.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
and taking the target received signal and the channel matrix as the input of a zero forcing signal, and calculating the estimated value of the transmitted signal by using a zero forcing detection algorithm.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
calculating an estimated value of the transmission signal by the following formula:
/>
wherein ,representing an estimated value of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, sgn (·) is the sign function, y es Representing the target received signal.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
Acquiring a loss function of the neural network based on the training data;
and training the neural network according to the loss function.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
performing cascaded linear and nonlinear conversion on training data to obtain a noise estimation value;
and comparing the noise estimated value with a noise label value to obtain the loss function.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
the loss function is obtained by the following formula:
wherein ,represents L 2 A norm loss function, B representing the batch size of the training data, +.>Represents a noise estimation value, n represents a noise label value, Ω represents a parameter requiring training, II 2 Representing a 2-norm.
Optionally, the processor is configured to read the computer program in the memory and perform the following operations:
calculating a gradient of the loss function;
and optimizing the loss function through a back propagation algorithm, and updating parameters of the neural network until the neural network converges.
According to the embodiment of the application, the noise estimation value is obtained based on the neural network, the noise is removed from the received signal, the signal detection is carried out by utilizing the denoised received signal and the channel matrix between the receiving end and the transmitting end, the estimation value of the transmitted signal is obtained, and the complexity of the signal detection is reduced in the process. The denoising network based on the neural network is used for preprocessing the received signals, so that the signal-to-noise ratio of signal detection can be improved, the detection performance of a traditional detector or a detector based on deep learning when noise interference is serious is improved, and meanwhile, the complexity of the detector is not increased.
Wherein in fig. 8, a bus architecture may comprise any number of interconnected buses and bridges, and in particular one or more processors represented by processor 810 and various circuits of memory represented by memory 820, linked together. The bus architecture may also link together various other circuits such as peripheral devices, voltage regulators, power management circuits, etc., which are well known in the art and, therefore, will not be described further herein. The bus interface provides an interface. Transceiver 800 may be a number of elements, including a transmitter and a transceiver, providing a means for communicating with various other apparatus over a transmission medium. The processor 810 is responsible for managing the bus architecture and general processing, and the memory 820 may store data used by the processor 810 in performing operations.
The processor 810 may be a Central Processing Unit (CPU), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a Field programmable gate array (Field-Programmable Gate Array, FPGA) or a complex programmable logic device (Complex Programmable Logic Device, CPLD), or it may employ a multi-core architecture.
It should be noted that, the above communication device provided by the embodiment of the present invention can implement all the method steps implemented by the embodiment of the method and achieve the same technical effects, and the same parts and beneficial effects as those of the embodiment of the method in the embodiment are not described in detail herein.
In addition, the specific embodiment of the application also provides a processor readable storage medium, on which a computer program is stored, wherein the program is executed by a processor to implement the steps of the signal detection method. And the same technical effects can be achieved, and in order to avoid repetition, the description is omitted here. The readable storage medium may be any available medium or data storage device that can be accessed by a processor, including, but not limited to, magnetic storage (e.g., floppy disks, hard disks, magnetic tape, magneto-optical disks (MOs), etc.), optical storage (e.g., CD, DVD, BD, HVD, etc.), and semiconductor storage (e.g., ROM, EPROM, EEPROM, nonvolatile storage (NAND FLASH), solid State Disk (SSD)), etc.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, magnetic disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flowchart and/or block of the flowchart illustrations and/or block diagrams, and combinations of flowcharts and/or block diagrams, can be implemented by computer-executable instructions. These computer-executable instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These processor-executable instructions may also be stored in a processor-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the processor-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These processor-executable instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present application without departing from the spirit or scope of the application. Thus, it is intended that the present application also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
Claims (30)
1. A signal detection method, comprising:
acquiring a noise estimation value based on a neural network;
according to the noise estimation value, carrying out noise removal processing on the received signal of the receiving end to obtain a target received signal after noise removal;
and acquiring an estimated value of the transmitting signal according to the target receiving signal and a channel matrix between the receiving end and the transmitting end.
2. The method of claim 1, wherein the obtaining a noise estimate based on a neural network comprises:
acquiring an initial estimated value of noise;
taking the initial estimated value of the noise, the received signal and the channel matrix as input feature matrices of a convolutional neural network, and performing M-layer convolutional processing on the input feature matrices through the convolutional neural network to obtain the estimated value of the noise;
the output result of each of the M-layer convolution processes is a single-layer estimated value of noise, M is greater than or equal to 2, and M is a positive integer.
3. The method of claim 2, wherein the obtaining an initial estimate of noise comprises:
acquiring an initial estimated value of a transmission signal according to the received signal and the channel matrix;
and acquiring the initial estimated value of the noise according to the initial estimated value of the transmitting signal, the receiving signal and the channel matrix.
4. A method according to claim 3, wherein obtaining an initial estimate of the transmitted signal based on the received signal and the channel matrix comprises:
the initial estimate of the transmitted signal is calculated by the following formula:
wherein ,representing an initial estimate of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, diag (H) T H) Representation matrix H T And a diagonal matrix of diagonal elements of H, y representing the received signal.
5. The method of claim 3, wherein said obtaining an initial estimate of said noise based on said initial estimate of said transmitted signal, said received signal, and said channel matrix comprises:
the initial estimate of the noise is calculated by the following formula:
wherein ,representing an initial estimate of said noise, y representing said received signal, H representing said channel matrix,/or->Representing an initial estimate of the transmitted signal.
6. The method of claim 2, wherein taking the initial estimate of the noise, the received signal, and the channel matrix as input feature matrices of a convolutional neural network, performing M-layer convolutional processing on the input feature matrices by the convolutional neural network, and obtaining the noise estimate comprises:
carrying out convolution processing on the input feature matrix by utilizing a one-dimensional convolution neural network to obtain a first layer estimated value of noise;
According to the first layer estimated value, noise removal processing is carried out on the received signal, and a first layer parameter is obtained;
carrying out convolution processing on the input feature matrix and the first layer parameters by using a one-dimensional convolution neural network to obtain a second layer estimated value of noise;
according to the second layer estimated value, noise removing processing is carried out on the received signal, and a second layer parameter is obtained;
and carrying out convolution processing on the input feature matrix and the second layer parameters by using a one-dimensional convolution neural network to obtain the noise estimation value.
7. The method of claim 6, wherein taking the initial estimate of the noise, the received signal, and the channel matrix as input feature matrices of a neural network, performing M-layer convolution processing on the input feature matrices by the convolutional neural network, and obtaining the noise estimate comprises:
the noise estimate is calculated by the following formula:
wherein ,representing the noise estimate,/->y represents the received signal, H represents the channel matrix,>representing an initial estimate of the noise, conv1 (·) represents a one-dimensional convolutional neural network.
8. The method according to claim 1, wherein the step of removing noise from the received signal at the receiving end according to the noise estimation value to obtain the target received signal after removing noise comprises:
subtracting the noise estimation value from the received signal to obtain a target received signal after noise removal;
wherein, the target received signal after noise removal is calculated by the following formula:
y es representing the target received signal, y representing the received signal,representing the noise estimate.
9. The method of claim 1, wherein obtaining an estimate of the transmitted signal based on the target received signal and a channel matrix between the receiving end and the transmitting end comprises:
and taking the target received signal and the channel matrix as the input of a zero forcing signal, and calculating the estimated value of the transmitted signal by using a zero forcing detection algorithm.
10. The method of claim 9, wherein said calculating an estimate of the transmitted signal using a zero forcing detection algorithm using the target received signal and the channel matrix as inputs for the zero forcing signal comprises:
Calculating an estimated value of the transmission signal by the following formula:
wherein ,representing an estimated value of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, sgn (·) is the sign function, y es Representing the target received signal.
11. The method according to claim 1, wherein the method further comprises:
acquiring a loss function of the neural network based on the training data;
and training the neural network according to the loss function.
12. The method of claim 11, wherein the obtaining a loss function of the neural network based on the training data comprises:
performing cascaded linear and nonlinear conversion on training data to obtain a noise estimation value;
and comparing the noise estimated value with a noise label value to obtain the loss function.
13. The method according to claim 11 or 12, wherein the obtaining a loss function of the neural network based on training data comprises:
the loss function is obtained by the following formula:
wherein ,represents L 2 A norm loss function, B representing the batch size of the training data, +.>Represents a noise estimation value, n represents a noise label value, Ω represents a parameter requiring training, II 2 Representing a 2-norm.
14. The method of claim 11, wherein training a neural network according to the loss function comprises:
calculating a gradient of the loss function;
and optimizing the loss function through a back propagation algorithm, and updating parameters of the neural network until the neural network converges.
15. A communication device, comprising: memory, transceiver, processor:
a memory for storing a computer program; a transceiver for transceiving data under control of the processor; a processor for reading the computer program in the memory and performing the following operations:
acquiring a noise estimation value based on a neural network;
according to the noise estimation value, carrying out noise removal processing on the received signal of the receiving end to obtain a target received signal after noise removal;
and acquiring an estimated value of the transmitting signal according to the target receiving signal and a channel matrix between the receiving end and the transmitting end.
16. The communication device of claim 15, wherein the processor is configured to read the computer program in the memory and perform the following:
Acquiring an initial estimated value of noise;
taking the initial estimated value of the noise, the received signal and the channel matrix as input feature matrices of a convolutional neural network, and performing M-layer convolutional processing on the input feature matrices through the convolutional neural network to obtain the estimated value of the noise;
the output result of each of the M-layer convolution processes is a single-layer estimated value of noise, M is greater than or equal to 2, and M is a positive integer.
17. The communication device of claim 16, wherein the processor is configured to read the computer program in the memory and perform the following:
acquiring an initial estimated value of a transmission signal according to the received signal and the channel matrix;
and acquiring the initial estimated value of the noise according to the initial estimated value of the transmitting signal, the receiving signal and the channel matrix.
18. The communication device of claim 17, wherein the processor is configured to read the computer program in the memory and perform the following:
the initial estimate of the transmitted signal is calculated by the following formula:
wherein ,representing an initial estimate of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, diag (H) T H) Representation matrix H T And a diagonal matrix of diagonal elements of H, y representing the received signal.
19. The communication device of claim 17, wherein the processor is configured to read the computer program in the memory and perform the following:
the initial estimate of the noise is calculated by the following formula:
wherein ,representing an initial estimate of said noise, y representing said received signal, H representing said channel matrix,/or->Representing an initial estimate of the transmitted signal.
20. The communication device of claim 16, wherein the processor is configured to read the computer program in the memory and perform the following:
carrying out convolution processing on the input feature matrix by utilizing a one-dimensional convolution neural network to obtain a first layer estimated value of noise;
according to the first layer estimated value, noise removal processing is carried out on the received signal, and a first layer parameter is obtained;
carrying out convolution processing on the input feature matrix and the first layer parameters by using a one-dimensional convolution neural network to obtain a second layer estimated value of noise;
according to the second layer estimated value, noise removing processing is carried out on the received signal, and a second layer parameter is obtained;
And carrying out convolution processing on the input feature matrix and the second layer parameters by using a one-dimensional convolution neural network to obtain the noise estimation value.
21. The communication device of claim 20, wherein the processor is configured to read the computer program in the memory and perform the following:
the noise estimate is calculated by the following formula:
wherein ,representing the noise estimate,/->y represents the received signal, H represents the channel matrix,>representing an initial estimate of the noise, conv1 (·) represents a one-dimensional convolutional neural network.
22. The communication device of claim 15, wherein the processor is configured to read the computer program in the memory and perform the following:
subtracting the noise estimation value from the received signal to obtain a target received signal after noise removal;
wherein, the target received signal after noise removal is calculated by the following formula:
y es representing the target received signal, y representing the received signal,representing the noise estimate.
23. The communication device of claim 15, wherein the processor is configured to read the computer program in the memory and perform the following:
And taking the target received signal and the channel matrix as the input of a zero forcing signal, and calculating the estimated value of the transmitted signal by using a zero forcing detection algorithm.
24. The communication device of claim 23, wherein the processor is configured to read the computer program in the memory and perform the following:
calculating an estimated value of the transmission signal by the following formula:
wherein ,representing an estimated value of the transmitted signal, H representing the channel matrix, H T Representing the transposed matrix of the channel matrix, sgn (·) is the sign function, y es Representing the target received signal.
25. The communication device of claim 15, wherein the processor is configured to read the computer program in the memory and perform the following:
acquiring a loss function of the neural network based on the training data;
and training the neural network according to the loss function.
26. The communication device of claim 25, wherein the processor is configured to read the computer program in the memory and perform the following:
performing cascaded linear and nonlinear conversion on training data to obtain a noise estimation value;
And comparing the noise estimated value with a noise label value to obtain the loss function.
27. The communication device according to claim 25 or 26, wherein the processor is configured to read the computer program in the memory and perform the following operations:
the loss function is obtained by the following formula:
wherein ,represents L 2 A norm loss function, B representing the batch size of the training data, +.>Represents a noise estimation value, n represents a noise label value, Ω represents a parameter requiring training, II 2 Representing a 2-norm.
28. The communication device of claim 25, wherein the processor is configured to read the computer program in the memory and perform the following:
calculating a gradient of the loss function;
and optimizing the loss function through a back propagation algorithm, and updating parameters of the neural network until the neural network converges.
29. A signal detection apparatus, comprising:
a first acquisition unit configured to acquire a noise estimation value based on a neural network;
the first processing unit is used for removing noise from the received signal of the receiving end according to the noise estimation value to obtain a target received signal after noise removal;
And the second acquisition unit is used for acquiring the estimated value of the transmission signal according to the target received signal and the channel matrix between the receiving end and the transmitting end.
30. A processor readable storage medium having stored thereon a computer program, which when executed by a processor performs the steps of the signal detection method according to any of claims 1 to 14.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210171753.3A CN116707677A (en) | 2022-02-24 | 2022-02-24 | Signal detection method and device and communication equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210171753.3A CN116707677A (en) | 2022-02-24 | 2022-02-24 | Signal detection method and device and communication equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116707677A true CN116707677A (en) | 2023-09-05 |
Family
ID=87826240
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210171753.3A Pending CN116707677A (en) | 2022-02-24 | 2022-02-24 | Signal detection method and device and communication equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116707677A (en) |
-
2022
- 2022-02-24 CN CN202210171753.3A patent/CN116707677A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109995449B (en) | Millimeter wave signal detection method based on deep learning | |
| CN110336594B (en) | Deep learning signal detection method based on conjugate gradient descent method | |
| US11222264B1 (en) | Method and device for recognizing space-frequency block code | |
| CN109951214B (en) | Signal detection method suitable for large-scale MIMO system | |
| CN109450601B (en) | Signal detection method and device | |
| US20210119681A1 (en) | Apparatus and method for estimating interference based on machine learning | |
| CN113269077A (en) | Underwater acoustic communication signal modulation mode identification method based on improved gating network and residual error network | |
| CN113162665B (en) | Pre-coding method based on deep learning channel prediction | |
| CN114268388A (en) | Channel estimation method based on improved GAN network in large-scale MIMO | |
| CN114362855A (en) | Channel state prediction method and system based on LSTM | |
| Kim et al. | Massive data generation for deep learning-aided wireless systems using meta learning and generative adversarial network | |
| US20230353426A1 (en) | Channel Estimation for an Antenna Array | |
| CN116707677A (en) | Signal detection method and device and communication equipment | |
| CN115037339B (en) | Signal detection method and terminal equipment | |
| CN113037409B (en) | Large-scale MIMO system signal detection method based on deep learning | |
| CN112803976B (en) | Large-scale MIMO precoding method and system and electronic equipment | |
| CN117060952A (en) | Signal detection method and device in MIMO system | |
| WO2023237182A1 (en) | Radio receiver with multi-stage equalization | |
| CN116888899A (en) | Neural network for MU-MIMO user selection | |
| CN107248876B (en) | Generalized spatial modulation symbol detection method based on sparse Bayesian learning | |
| CN113037411B (en) | Multi-user signal detection method and device based on deep learning | |
| CN110912643B (en) | AMP detection improvement method based on machine learning | |
| CN112202510B (en) | Signal detection method based on regularization stream neural network | |
| CN115694571A (en) | Signal detection method and device based on deep learning in large-scale MIMO system | |
| CN114915321A (en) | Method and device for dynamically detecting signals in MIMO (multiple input multiple output) system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |