CN114745248B - DM-GSM signal detection method based on convolutional neural network - Google Patents
DM-GSM signal detection method based on convolutional neural network Download PDFInfo
- Publication number
- CN114745248B CN114745248B CN202210358224.4A CN202210358224A CN114745248B CN 114745248 B CN114745248 B CN 114745248B CN 202210358224 A CN202210358224 A CN 202210358224A CN 114745248 B CN114745248 B CN 114745248B
- Authority
- CN
- China
- Prior art keywords
- dimensional
- vector
- matrix
- sample
- cnn
- 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.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 45
- 238000013527 convolutional neural network Methods 0.000 title claims description 22
- 239000013598 vector Substances 0.000 claims abstract description 49
- 239000011159 matrix material Substances 0.000 claims abstract description 45
- 238000012549 training Methods 0.000 claims abstract description 19
- 238000013528 artificial neural network Methods 0.000 claims abstract description 18
- 238000000034 method Methods 0.000 claims abstract description 14
- 230000005540 biological transmission Effects 0.000 claims abstract description 10
- 230000006854 communication Effects 0.000 claims abstract description 8
- 230000006870 function Effects 0.000 claims description 21
- 230000004913 activation Effects 0.000 claims description 12
- 238000005562 fading Methods 0.000 claims description 12
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 9
- 238000013507 mapping Methods 0.000 claims description 6
- 238000004088 simulation Methods 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 241001417495 Serranidae Species 0.000 claims description 3
- 239000000654 additive Substances 0.000 claims description 3
- 230000000996 additive effect Effects 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000007476 Maximum Likelihood Methods 0.000 abstract description 11
- 238000004891 communication Methods 0.000 abstract description 5
- 238000013135 deep learning Methods 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/32—Carrier systems characterised by combinations of two or more of the types covered by groups H04L27/02, H04L27/10, H04L27/18 or H04L27/26
- H04L27/34—Amplitude- and phase-modulated carrier systems, e.g. quadrature-amplitude modulated carrier systems
- H04L27/36—Modulator circuits; Transmitter circuits
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/32—Carrier systems characterised by combinations of two or more of the types covered by groups H04L27/02, H04L27/10, H04L27/18 or H04L27/26
- H04L27/34—Amplitude- and phase-modulated carrier systems, e.g. quadrature-amplitude modulated carrier systems
- H04L27/38—Demodulator circuits; Receiver circuits
-
- 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)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Signal Processing (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)
- Computer Networks & Wireless Communication (AREA)
- Radio Transmission System (AREA)
- Digital Transmission Methods That Use Modulated Carrier Waves (AREA)
Abstract
A Convolution Neural Network (CNN) -based dual-mode generalized spatial modulation (DM-GSM) signal detection method belongs to the technical field of signal detection in a communication system. Firstly, generating a sample database based on a DM-GSM system, and remolding a received signal sample and a channel parameter sample into a two-dimensional input matrix; then, constructing a CNN network, and extracting characteristic information in a transmission symbol by utilizing a two-dimensional convolution layer; and finally, performing offline training on the neural network by utilizing the preprocessed data samples and the known transmission vector labels, and performing real-time online signal detection on the trained network according to the input sample data. The invention adopts the deep learning method to detect the signals of the DM-GSM system, effectively reduces the complexity of the signal detection of the traditional DM-GSM system, obtains the optimal error rate performance of the detection near the Maximum Likelihood (ML), and is obviously superior to the traditional linear detection method.
Description
Technical Field
The invention relates to a DM-GSM signal detection method based on convolutional neural network, which belongs to the technical field of signal detection in communication systems.
Background
Spatial Modulation (SM) is a multi-antenna transmission technique that utilizes both a transmit antenna index and a digital modulation constellation to transmit information bits. With the increase of the number of transmitting antennas, only one antenna is activated in each time slot, which causes the waste of space resources. To overcome the limitations of SM, generalized Spatial Modulation (GSM) techniques are proposed in which multiple transmit antennas are activated for transmission of a data stream in each time slot. The GSM scheme can achieve higher spectral efficiency compared to the conventional SM scheme. The dual-mode generalized spatial modulation (DM-GSM) scheme utilizes index bits to divide all transmitting antennas into two groups, and two resolvable constellation modulation symbols are respectively transmitted on the two groups of antennas at the same time, so that the frequency spectrum efficiency of the system is effectively improved. Although DM-GSM may achieve better performance, DM-GSM has higher detection complexity than conventional GSM. The traditional linear detection algorithm can be directly applied to a GSM system, such as a Minimum Mean Square Error (MMSE) detection algorithm and a Zero Forcing (ZF) detection algorithm, and the performance loss of the algorithm is larger although the calculation complexity of the algorithm is reduced compared with that of a maximum likelihood detection (ML) detection algorithm. In recent years, deep learning has been widely used in signal detection due to its strong learning ability.
Albinsaid et al (see H.Albinsaid, K.Singh, S.Biswas, C. -P.Li, and M. -S.Alouini, "Block Deep Neural Network-Based Signal Detector for Generalized Spatial Modulation," IEEE Communications Letters, vol.24, no.12, pp.2775-2779, dec.2020.) propose an architecture based on a packet depth neural network (B-DNN) in which the active antenna and its transmitted constellation symbols are detected by smaller DNNs, and the bit error rate performance of the proposed B-DNN detector is superior to that of conventional ZF and MMSE detection schemes, achieving near ML detection performance at low complexity. Kim et al (see J.Kim, H.Ro, and H.park, "Deep Learning-Based Detector for Dual Mode OFDM With Index Modulation," IEEE Wireless Communications Letters, vol.10, no.7, pp.1562-1566, jul.2021.) propose a Deep Learning based DM-OFDM-IM detector that employs Convolutional Neural Network (CNN) and Deep Neural Network (DNN) connections to detect index bits and carrier bits, respectively, to achieve bit error rate performance approaching that of ML detectors, while reducing system complexity. The scheme utilizes the neural network to perform joint detection, has higher complexity, is limited to a certain extent in an actual communication scene, cannot effectively improve the system performance, and is not yet applied to DM-GSM system signal detection.
Disclosure of Invention
According to the defects and shortcomings of the prior art and the solution, the invention provides a DM-GSM signal detection method based on a convolutional neural network, which effectively reduces the complexity of a system, obtains the error rate performance close to ML detection, and is superior to the traditional linear detection method.
The technical scheme of the invention is as follows:
a DM-GSM, dual-mode generalized spatial modulation signal detection method based on convolutional neural network is realized by DM-GSM system, the system mainly comprises a transmitting end module, a channel and a receiving end module, the transmitting end module comprises a bit grouper, an index selector, a constellation mapper and N t A root transmit antenna; the channel is a typical rayleigh fading channel; the receiving end module comprises N r A root receiving antenna, a preprocessing module and a CNN-DM detector; the CNN-DM detector comprises a two-dimensional convolution layer, a Flatten layer and two full connection layers, and adopts a Dropout layer to prevent overfitting; the transmitting end index selector is used for selecting the transmitting end index from N according to the mapping rule of bits to transmitting antennas t Activation of N in root transmit antenna p The constellation mapper adopts two different constellation modes to modulate the symbols; assuming that the channel obeys rayleigh fading, the receiving end knows perfect channel state information, and the communication process comprises three steps: firstly, generating a sample database based on a DM-GSM system, and remolding a received signal sample and a channel parameter sample into a two-dimensional input matrix; then constructing a convolutional neural network, namely a CNN network, and extracting characteristic information in a transmitted symbol by utilizing a two-dimensional convolutional layer; finally, the neural network is trained offline by utilizing the preprocessed data samples and the known transmission vector labels, and the trained network performs real-time online signal detection according to the input sample data, and the method detects DM-GSM signals by utilizing CNN, and comprises the following specific steps:
1) Generating a sample database based on a DM-GSM system, and reshaping received signal samples and channel parameter samples into a two-dimensional input matrix:
in a DM-GSM system, an input bit sequence b is divided into two parts by a bit packetizer, the first part comprisingIndex bits, where->Represents the maximum integer less than x +.! Representing the factorization, the index bits are fed into an antenna index selector to select two active antenna subsets, denoted as I A And I B Two subsets respectively contain p 1 And p 2 The total number of active antennas is N p ＝p 1 +p 2 The method comprises the steps of carrying out a first treatment on the surface of the The second part comprises p 1 log 2 m A +p 2 log 2 m B Constellation modulated bits, p 1 log 2 m A P 2 log 2 m B The constellation modulation bits respectively correspond to the activated antenna subset I A And I B And respectively sent to a modulation mapper adopting two constellations of A and B for modulation, +.>Representing constellation a modulation symbol set,/->Representing the set of constellation B modulation symbols,and->The number of constellation points in (a) is m respectively A And m B The modulated signal is transmitted on the active antenna through the spatial mapper; through spatial mapping, the transmitted signal vector is expressed as +.>In which there is N p Non-zero elements, "> Respectively representing the ith transmitting antenna to transmit symbols modulated with A, B constellation, [] T Representing vector transpose>Represents N t X 1-dimensional complex vector, emissionAfter the signal vector x is transmitted by the Rayleigh fading channel, the signal vector received by the receiving end is +.>And y=hx+n, where +.>Is the mean value is 0 and the variance is sigma 2 Is a complex additive white gaussian noise vector,/>Represents N r ×N t The elements of the Viril fading channel matrix obey complex Gaussian distribution with the mean value of 0 and the variance of 1;
generating a sample database based on DM-GSM system simulation, converting complex signals into real-valued signals which can be processed by a neural network, remolding a received signal sample and a channel parameter sample into a two-dimensional input matrix Z through a preprocessing module, and storing channel coefficients and received signals corresponding to a kth receiving antenna in a column vector Z k In the concrete expression is Wherein Re and Im represent the real and imaginary parts of the complex number, H, respectively k,i Is the kth row, the ith column element, y of the channel matrix H k Is the reception signal of the kth reception antenna, willN composed of y and H can be obtained as input to the neural network r ×2(N t +1) a dimensional input matrix to sufficiently extract inter-symbol feature information;
2) Constructing a CNN network, and extracting characteristic information in a transmitted symbol by utilizing a two-dimensional convolution layer:
the convolution layer of the CNN-DM detector uses the preprocessed two-dimensional matrix Z as network input, and the convolution kernel is thatWherein C represents the number of convolution kernels, ">The ith element representing the c-th convolution kernel,/->Representing 1×2 (N) t +1) real vector in dimension, convolutional layer activation function is Relu function, i.e., f Relu (x) =max (0, x), convolutional layer output matrix +.>Wherein the nth row, column c element of D is represented asz[n,i]An nth row and an ith column element representing a convolution layer input matrix z, B c Representing the offset corresponding to the c-th convolution kernel; in the flat layer, the two-dimensional matrix D of the convolutional layer output is converted into a one-dimensional arrayThe activation function of the first full connection layer is the Relu function, the output layer activation function is the Sigmoid function, which maps variables to [0,1 ]]The expression is f Sig (x)＝1/(1+e -x ) The output of the CNN-DM detector may be expressed asWherein W is 1 、b 1 Weight matrix and bias vector of the first full connection layer respectively, W 2 、b 2 Respectively a weight matrix and a bias vector of the output layer;
3) Offline training is carried out on the neural network by utilizing the preprocessed data samples and the known transmission vector labels, and the trained network carries out real-time online signal detection according to the input sample data:
in the training process, based on DM-GSM systemThe generated sample data is reshaped into a two-dimensional matrix vector by a preprocessing module, the two-dimensional matrix vector is used as an input feature vector of a signal detection network, and an actual sending bit sequence is a corresponding tag vector; CNN-DM detector learning parameter θ= { K 1 ,K 2 ,…,K C ,B 1 ,B 2 ,…,B C ,W 1 ,b 1 ,W 2 ,b 2 Training with a binary cross entropy loss function for bi-classification to optimize network parameters, expressed asWherein y is i For the information bits transmitted +.>Representing estimated information bits, N being the number of bits transmitted, the training data set and the test data set being 2.8X10 respectively in size 5 And 1.2X10 5 Setting the training round number as 100, and optimizing a network model by adopting an Adam gradient descent optimization algorithm;
after offline training, the receiver can use the trained detector with optimal parameters to detect the bit sequence in real time, and the received signal y and the channel matrix H are input into the CNN-DM detector, so as to directly give the estimated transmitted information bit sequence
The invention provides a DM-GSM signal detection method based on a convolutional neural network, which utilizes the strong learning ability of the neural network to realize the performance close to the optimal detection under the lower complexity, is superior to the traditional ZF and MMSE detection schemes, and can obtain the performance superior to ML detection under the noise distribution deviating from a standard Gaussian model.
Drawings
Fig. 1 is a schematic block diagram of the system architecture of the present invention.
Fig. 2 is a schematic block diagram of the detection network structure of the method of the present invention.
Fig. 3 is a graph of bit error rate performance versus simulation for the method of the present invention and for Maximum Likelihood (ML) detection, zero Forcing (ZF) detection, minimum Mean Square Error (MMSE) detection. It can be seen from fig. 3 that the system of the present invention is far better than conventional ZF and MMSE detection and can achieve system performance approaching ML.
Detailed Description
The invention is further illustrated, but not limited, by the following figures and examples.
Examples:
a DM-GSM, dual-mode generalized spatial modulation signal detection method based on convolutional neural network is realized by DM-GSM system, as shown in figure 1, the system mainly comprises a transmitting end module, a channel and a receiving end module, the transmitting end module comprises a bit grouper, an index selector, a constellation mapper and N t A root transmit antenna; the channel is a typical rayleigh fading channel; the receiving end module comprises N r A root receiving antenna, a preprocessing module and a CNN-DM detector; the CNN-DM detector comprises a two-dimensional convolution layer, a Flatten layer and two full connection layers, and adopts a Dropout layer to prevent overfitting; the transmitting end index selector is used for selecting the transmitting end index from N according to the mapping rule of bits to transmitting antennas t Activation of N in root transmit antenna p The constellation mapper adopts two different constellation modes to modulate the symbols; assuming that the channel obeys rayleigh fading, the receiving end knows perfect channel state information, and the communication process comprises three steps: firstly, generating a sample database based on a DM-GSM system, and remolding a received signal sample and a channel parameter sample into a two-dimensional input matrix; then constructing a convolutional neural network, namely a CNN network, and extracting characteristic information in a transmitted symbol by utilizing a two-dimensional convolutional layer; finally, the neural network is trained offline by utilizing the preprocessed data samples and the known transmission vector labels, and the trained network performs real-time online signal detection according to the input sample data, and the method detects DM-GSM signals by utilizing CNN, and comprises the following specific steps:
1) Generating a sample database based on a DM-GSM system, and reshaping received signal samples and channel parameter samples into a two-dimensional input matrix:
in DM-GSM system, transmissionThe incoming bit sequence b is divided into two parts by a bit packetizer, the first part comprisingIndex bits, where->Represents the maximum integer less than x +.! Representing the factorization, the index bits are fed into an antenna index selector to select two active antenna subsets, denoted as I A And I B Two subsets respectively contain p 1 And p 2 The total number of active antennas is N p ＝p 1 +p 2 The method comprises the steps of carrying out a first treatment on the surface of the The second part comprises p 1 log 2 m A +p 2 log 2 m B Constellation modulated bits, p 1 log 2 m A P 2 log 2 m B The constellation modulation bits respectively correspond to the activated antenna subset I A And I B And respectively sent to a modulation mapper adopting two constellations of A and B for modulation, +.>Representing constellation a modulation symbol set,/->Representing the set of constellation B modulation symbols,and->The number of constellation points in (a) is m respectively A And m B The modulated signal is transmitted on the active antenna through the spatial mapper; through spatial mapping, the transmitted signal vector is expressed as +.>In which there is N p Non-zero elements, "> Respectively representing the ith transmitting antenna to transmit symbols modulated with A, B constellation, [] T Representing vector transpose>Represents N t X 1-dimensional complex vector, after the transmitting signal vector x is transmitted by Rayleigh fading channel, the receiving end receives signal vector +.>And y=hx+n, where +.>Is the mean value is 0 and the variance is sigma 2 Is a complex additive white gaussian noise vector,/>Represents N r ×N t The elements of the Viril fading channel matrix obey complex Gaussian distribution with the mean value of 0 and the variance of 1;
generating a sample database based on DM-GSM system simulation, converting complex signals into real-valued signals which can be processed by a neural network, remolding a received signal sample and a channel parameter sample into a two-dimensional input matrix Z through a preprocessing module, and storing channel coefficients and received signals corresponding to a kth receiving antenna in a column vector Z k In the concrete expression is Wherein Re and Im represent the real and imaginary parts of the complex number, H, respectively k,i Is the kth row, the ith column element, y of the channel matrix H k Is the reception signal of the kth reception antenna, willN composed of y and H can be obtained as input to the neural network r ×2(N t +1) a dimensional input matrix to sufficiently extract inter-symbol feature information;
2) Constructing a CNN network, and extracting characteristic information in a transmitted symbol by utilizing a two-dimensional convolution layer:
the convolution layer of the CNN-DM detector uses the preprocessed two-dimensional matrix Z as network input, and the convolution kernel is thatWherein C represents the number of convolution kernels, ">The ith element representing the c-th convolution kernel,/->Representing 1×2 (N) t +1) real vector in dimension, convolutional layer activation function is Relu function, i.e., f Relu (x) =max (0, x), convolutional layer output matrix +.>Wherein the nth row, column c element of D is represented asz[n,i]An nth row and an ith column element representing a convolution layer input matrix z, B c Representing the offset corresponding to the c-th convolution kernel; in the flat layer, the two-dimensional matrix D of the convolutional layer output is converted into a one-dimensional arrayThe activation function of the first full connection layer is the Relu function, the output layer activation function is the Sigmoid function, which maps variables to [0,1 ]]The expression is f Sig (x)＝1/(1+e -x ) The output of the CNN-DM detector may be expressed asWherein W is 1 、b 1 Weight matrix and bias vector of the first full connection layer respectively, W 2 、b 2 Respectively a weight matrix and a bias vector of the output layer;
3) Offline training is carried out on the neural network by utilizing the preprocessed data samples and the known transmission vector labels, and the trained network carries out real-time online signal detection according to the input sample data:
in the training process, based on sample data generated by a DM-GSM system, a received signal sample and a channel parameter sample are remolded into two-dimensional matrix vectors through a preprocessing module, the two-dimensional matrix vectors are used as input feature vectors of a signal detection network, and an actual transmitted bit sequence is a corresponding label vector; CNN-DM detector learning parameter θ= { K 1 ,K 2 ,…,K C ,B 1 ,B 2 ,…,B C ,W 1 ,b 1 ,W 2 ,b 2 Training with a binary cross entropy loss function for bi-classification to optimize network parameters, expressed asWherein y is i For the information bits transmitted +.>Representing estimated information bits, N being the number of bits transmitted, the training data set and the test data set being 2.8X10 respectively in size 5 And 1.2X10 5 Setting the training round number as 100, and optimizing a network model by adopting an Adam gradient descent optimization algorithm;
after offline training, the receiver can use the trained detector with optimal parameters to detect the bit sequence in real time, and the received signal y and the channel matrix H are input into the CNN-DM detector, so as to directly give the estimated transmitted information bit sequence
Claims (1)
1. A DM-GSM, dual-mode generalized spatial modulation signal detection method based on convolutional neural network is realized by DM-GSM system, the system mainly comprises a transmitting end module, a channel and a receiving end module, the transmitting end module comprises a bit grouper, an index selector, a constellation mapper and N t A root transmit antenna; the channel is a typical rayleigh fading channel; the receiving end module comprises N r A root receiving antenna, a preprocessing module and a CNN-DM detector; the CNN-DM detector comprises a two-dimensional convolution layer, a Flatten layer and two full connection layers, and adopts a Dropout layer to prevent overfitting; the transmitting end index selector is used for selecting the transmitting end index from N according to the mapping rule of bits to transmitting antennas t Activation of N in root transmit antenna p The constellation mapper adopts two different constellation modes to modulate the symbols; assuming that the channel obeys rayleigh fading, the receiving end knows perfect channel state information, and the communication process comprises three steps: firstly, generating a sample database based on a DM-GSM system, and remolding a received signal sample and a channel parameter sample into a two-dimensional input matrix; then constructing a convolutional neural network, namely a CNN network, and extracting characteristic information in a transmitted symbol by utilizing a two-dimensional convolutional layer; finally, the neural network is trained offline by utilizing the preprocessed data samples and the known transmission vector labels, and the trained network performs real-time online signal detection according to the input sample data, and the method detects DM-GSM signals by utilizing CNN, and comprises the following specific steps:
1) Generating a sample database based on a DM-GSM system, and reshaping received signal samples and channel parameter samples into a two-dimensional input matrix:
in a DM-GSM system, an input bit sequence b is divided into two parts by a bit packetizer, the first part comprisingIndex bits, where->Represents the maximum integer less than x +.! Representing the factorization, the index bits are fed into an antenna index selector to select two active daysLine subset, denoted as I A And I B Two subsets respectively contain p 1 And p 2 The total number of active antennas is N p ＝p 1 +p 2 The method comprises the steps of carrying out a first treatment on the surface of the The second part comprises p 1 log 2 m A +p 2 log 2 m B Constellation modulated bits, p 1 log 2 m A P 2 log 2 m B The constellation modulation bits respectively correspond to the activated antenna subset I A And I B And respectively sent to a modulation mapper adopting two constellations of A and B for modulation, +.>Representing constellation a modulation symbol set,/->Representing constellation B modulation symbol set,>and->The number of constellation points in (a) is m respectively A And m B The modulated signal is transmitted on the active antenna through the spatial mapper; through spatial mapping, the transmitted signal vector is expressed as +.>In which there is N p Non-zero elements, "> Respectively representing the ith transmitting antenna to transmit symbols modulated with A, B constellation, [] T Representing vector transpose>Represents N t X 1-dimensional complex vector, after the transmitting signal vector x is transmitted by Rayleigh fading channel, the receiving end receives signal vector +.>And y=hx+n, where +.>Is the mean value is 0 and the variance is sigma 2 Is a complex additive white gaussian noise vector,/>Represents N r ×N t The elements of the Viril fading channel matrix obey complex Gaussian distribution with the mean value of 0 and the variance of 1;
generating a sample database based on DM-GSM system simulation, converting complex signals into real-valued signals which can be processed by a neural network, remolding a received signal sample and a channel parameter sample into a two-dimensional input matrix Z through a preprocessing module, and storing channel coefficients and received signals corresponding to a kth receiving antenna in a column vector Z k In the concrete expression isWherein Re and Im represent the real and imaginary parts of the complex number, H, respectively k,i Is the kth row, the ith column element, y of the channel matrix H k Is the reception signal of the kth reception antenna, willN composed of y and H can be obtained as input to the neural network r ×2(N t +1) a dimensional input matrix to sufficiently extract inter-symbol feature information;
2) Constructing a CNN network, and extracting characteristic information in a transmitted symbol by utilizing a two-dimensional convolution layer:
use of preprocessed two-dimensional at convolutional layer of CNN-DM detectorMatrix Z is used as network input, and convolution kernel isWherein C represents the number of convolution kernels, ">The ith element representing the c-th convolution kernel,/->Representing 1×2 (N) t +1) real vector in dimension, convolutional layer activation function is Relu function, i.e., f Relu (x) =max (0, x), convolutional layer output matrix +.>Wherein the nth row, column c element of D is represented asz[n,i]An nth row and an ith column element representing a convolution layer input matrix z, B c Representing the offset corresponding to the c-th convolution kernel; in the flat layer, the two-dimensional matrix D of the convolutional layer output is converted into a one-dimensional arrayThe activation function of the first full connection layer is the Relu function, the output layer activation function is the Sigmoid function, which maps variables to [0,1 ]]The expression is f Sig (x)＝1/(1+e -x ) The output of the CNN-DM detector may be expressed asWherein W is 1 、b 1 Weight matrix and bias vector of the first full connection layer respectively, W 2 、b 2 Respectively a weight matrix and a bias vector of the output layer;
3) Offline training is carried out on the neural network by utilizing the preprocessed data samples and the known transmission vector labels, and the trained network carries out real-time online signal detection according to the input sample data:
in the training process, based on sample data generated by a DM-GSM system, a received signal sample and a channel parameter sample are remolded into two-dimensional matrix vectors through a preprocessing module, the two-dimensional matrix vectors are used as input feature vectors of a signal detection network, and an actual transmitted bit sequence is a corresponding label vector; CNN-DM detector learning parameter θ= { K 1 ,K 2 ,…,K C ,B 1 ,B 2 ,…,B C ,W 1 ,b 1 ,W 2 ,b 2 Training with a binary cross entropy loss function for bi-classification to optimize network parameters, expressed asWherein y is i For the information bits transmitted +.>Representing estimated information bits, N being the number of bits transmitted, the training data set and the test data set being 2.8X10 respectively in size 5 And 1.2X10 5 Setting the training round number as 100, and optimizing a network model by adopting an Adam gradient descent optimization algorithm;
after offline training, the receiver can use the trained detector with optimal parameters to detect the bit sequence in real time, and the received signal y and the channel matrix H are input into the CNN-DM detector, so as to directly give the estimated transmitted information bit sequence
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210358224.4A CN114745248B (en) | 2022-04-06 | 2022-04-06 | DM-GSM signal detection method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210358224.4A CN114745248B (en) | 2022-04-06 | 2022-04-06 | DM-GSM signal detection method based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114745248A CN114745248A (en) | 2022-07-12 |
| CN114745248B true CN114745248B (en) | 2023-11-14 |
Family
ID=82279985
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210358224.4A Active CN114745248B (en) | 2022-04-06 | 2022-04-06 | DM-GSM signal detection method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114745248B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115208731B (en) * | 2022-07-15 | 2023-08-25 | 北京科技大学 | Method and device for suppressing peak-to-average power ratio (PAPR) of signal |
| CN115865593A (en) * | 2022-11-18 | 2023-03-28 | 重庆邮电大学 | OFDM-IM signal detection method based on noise reduction network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108234370A (en) * | 2017-12-22 | 2018-06-29 | 西安电子科技大学 | Modulation mode of communication signal recognition methods based on convolutional neural networks |
| CN110071750A (en) * | 2019-03-12 | 2019-07-30 | 重庆临菲电子科技有限公司 | A kind of two-dimentional idle bit index modulation method based on machine learning |
| CN110942100A (en) * | 2019-11-29 | 2020-03-31 | 山东大学 | Working method of spatial modulation system based on deep denoising neural network |
| CN111769862A (en) * | 2020-06-28 | 2020-10-13 | 南京理工大学 | Joint detection algorithm in spatial modulation network based on deep learning |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1718019B1 (en) * | 2005-04-29 | 2017-07-19 | Sony Deutschland GmbH | Equalising structure and method with maximum likelihood detection |
| KR102530000B1 (en) * | 2017-06-19 | 2023-05-08 | 버지니아 테크 인터렉추얼 프라퍼티스, 인크. | Encoding and decoding of information for wireless transmission using multi-antenna transceivers |
-
2022
- 2022-04-06 CN CN202210358224.4A patent/CN114745248B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108234370A (en) * | 2017-12-22 | 2018-06-29 | 西安电子科技大学 | Modulation mode of communication signal recognition methods based on convolutional neural networks |
| CN110071750A (en) * | 2019-03-12 | 2019-07-30 | 重庆临菲电子科技有限公司 | A kind of two-dimentional idle bit index modulation method based on machine learning |
| CN110942100A (en) * | 2019-11-29 | 2020-03-31 | 山东大学 | Working method of spatial modulation system based on deep denoising neural network |
| CN111769862A (en) * | 2020-06-28 | 2020-10-13 | 南京理工大学 | Joint detection algorithm in spatial modulation network based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114745248A (en) | 2022-07-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114745248B (en) | DM-GSM signal detection method based on convolutional neural network | |
| CN110460402B (en) | End-to-end communication system establishing method based on deep learning | |
| CN110336594B (en) | Deep learning signal detection method based on conjugate gradient descent method | |
| WO2008127035A1 (en) | Multiple-antenna space multiplexing system using enhancement signal detection | |
| CN114584448B (en) | SM-OFDM signal grouping detection method based on deep neural network | |
| CN111628833B (en) | MIMO antenna number estimation method based on convolutional neural network | |
| CN113630130B (en) | End-to-end digital communication demodulation method | |
| CN111600640B (en) | Working method of full generalized spatial modulation system based on multi-antenna state | |
| CN114745230A (en) | OTFS signal receiving and recovering method based on deep neural network structure | |
| Marseet et al. | Application of complex-valued convolutional neural network for next generation wireless networks | |
| Ji et al. | Modulation recognition in maritime multipath channels: A blind equalization-aided deep learning approach | |
| CN112242969A (en) | Novel single-bit OFDM receiver based on model-driven deep learning | |
| CN111368979A (en) | PAPR suppression method for MIMO-OFDM system | |
| CN110365414B (en) | Enhanced optical space modulation method suitable for lognormal turbulence channel | |
| KR102064301B1 (en) | Signal detection apparatus using ensemble machine learning based on MIMO system and method thereof | |
| Liu et al. | IMNet: A learning based detector for index modulation aided MIMO-OFDM systems | |
| CN114745246A (en) | Working method of orthogonal time-frequency space modulation system based on orthogonal space modulation | |
| CN113225114B (en) | Wireless communication signal sending and receiving method based on precoding joint optimization | |
| CN1909398B (en) | Power control method based on peri-zero judgment in multi-aerial system | |
| CN109462429B (en) | Beam domain modulation device and method of large-scale multiple-input multiple-output millimeter wave system | |
| CN113225117A (en) | Multi-user Massive MIMO system signal transmitting and receiving method | |
| CN109818891B (en) | Lattice reduction assisted low-complexity greedy sphere decoding detection method | |
| Li et al. | Model-driven deep learning scheme for adaptive transmission in MIMO-SCFDE system | |
| CN109818663A (en) | A kind of low complex degree difference quadrature spatial modulation detection method | |
| CN110995330B (en) | Antenna selection method based on grouping and EDAS algorithm |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |