CN111314255A - Low-complexity SISO and MIMO receiver generation method - Google Patents
Low-complexity SISO and MIMO receiver generation method Download PDFInfo
- Publication number
- CN111314255A CN111314255A CN202010090637.XA CN202010090637A CN111314255A CN 111314255 A CN111314255 A CN 111314255A CN 202010090637 A CN202010090637 A CN 202010090637A CN 111314255 A CN111314255 A CN 111314255A
- Authority
- CN
- China
- Prior art keywords
- layer
- receiver
- nodes
- siso
- mimo
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 23
- URWAJWIAIPFPJE-YFMIWBNJSA-N sisomycin Chemical compound O1C[C@@](O)(C)[C@H](NC)[C@@H](O)[C@H]1O[C@@H]1[C@@H](O)[C@H](O[C@@H]2[C@@H](CC=C(CN)O2)N)[C@@H](N)C[C@H]1N URWAJWIAIPFPJE-YFMIWBNJSA-N 0.000 title claims abstract 14
- 238000001514 detection method Methods 0.000 claims abstract description 48
- 238000005070 sampling Methods 0.000 claims abstract description 31
- 230000006835 compression Effects 0.000 claims abstract description 20
- 238000007906 compression Methods 0.000 claims abstract description 20
- 238000012549 training Methods 0.000 claims abstract description 17
- 238000005259 measurement Methods 0.000 claims abstract description 7
- 230000006870 function Effects 0.000 claims description 47
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000005540 biological transmission Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 5
- 230000004044 response Effects 0.000 claims description 4
- 238000007493 shaping process Methods 0.000 claims description 3
- 230000007547 defect Effects 0.000 claims description 2
- 230000010363 phase shift Effects 0.000 claims 1
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 5
- 238000013527 convolutional neural network Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 239000000654 additive Substances 0.000 description 2
- 230000000996 additive effect Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005562 fading Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000007476 Maximum Likelihood Methods 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 230000006735 deficit Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/0008—Modulated-carrier systems arrangements for allowing a transmitter or receiver to use more than one type of modulation
-
- 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
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
- H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
- H04B7/0413—MIMO systems
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B7/00—Radio transmission systems, i.e. using radiation field
- H04B7/02—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas
- H04B7/04—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas
- H04B7/08—Diversity systems; Multi-antenna system, i.e. transmission or reception using multiple antennas using two or more spaced independent antennas at the receiving station
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Signal Processing (AREA)
- Theoretical Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Radio Transmission System (AREA)
Abstract
The invention discloses a method for generating low-complexity SISO and MIMO receivers, which comprises the following steps: s1, sampling SISO and MIMO signals at a receiving end; s2, finding an optimal discriminant function, and converting the signal detection problem of different modulation modes in the SISO and MIMO system into the waveform identification problem; s3, training the receiver to minimize the loss function corresponding to the discriminant function of the hidden layer and the output layer, and further obtaining the optimal parameters of the hidden layer and the output layer of the compression stack type self-encoder receiver; and S4, selecting a strategy based on the number of nodes and the number of layers of the bit error rate measurement, and obtaining the low-complexity compression stack type self-encoder receiver structure based on the selection strategy of the number of nodes and the number of layers. The invention can simplify the structure of the receiver, and under the conditions of SISO and MIMO, the error rate of signal detection of different modulation modes under different channels reaches or exceeds the optimal detection theoretical value, and meanwhile, the invention has stronger robustness for CFO and phase deviation.
Description
Technical Field
The invention relates to the technical field of digital communication, in particular to a method for generating low-complexity SISO and MIMO receivers.
Background
With the application of Deep Learning (DL) in the fields of computer vision, natural language processing, automatic driving, and the like, more and more learners apply DL in the field of communication signal processing, such as radar signal processing, signal arrival angle estimation, signal modulation identification, and the like, so as to solve the complex problem that a satisfactory result cannot be obtained by using a conventional statistical method.
At present, based on the research result of DL signal detection, such as the joint optimal transmitting and receiving based on the deep self-encoder, in the structure, the channel is considered as a layer of the neural network of the self-encoder, and is described by using an accurate conditional probability density function, however, in the actual communication system, the information of the channel can not be obtained before the link is established; a BPSK demodulator based on a Convolutional Neural Network (CNN) detects a phase of a BPSK signal by using the CNN, a low-pass filter, a converter, and a synchronization module, but the CNN has a complex structure and includes many optimization parameters, and training and offline deployment of the Network are difficult. The structure of the known neural network receiver in the correlation work is complex, the neural network receiver comprises a large number of layers and nodes, the training difficulty of the network is large, and meanwhile, the known signal demodulation is performed under ideal conditions, namely Carrier Frequency Offset (CFO) and phase Offset (phase Offset) are assumed to be absent.
It is the focus of the present invention to design a low complexity compressed Stacked self-encoder (CSAE) receiver that does not rely on channel models and noise assumptions and has strong robustness to CFO and phase offsets in practical communication systems.
Disclosure of Invention
The invention aims to provide a low-complexity CSAE (continuous adaptive echo cancellation) neural network receiver which is not dependent on a channel model and a noise hypothesis and is suitable for non-ideal Single-Input Single-Output (SISO) and Multiple-Input Multiple-Output (MIMO) conditions, and the number of nodes and the number of layers of each layer of the deep neural network receiver are designed by using Bit Error Rate (BER) as a measurement, so that the structure of the receiver is simplified. The CSAE receiver has a simple structure, only comprises 2 hidden layers, and can reach or exceed the optimal detection theoretical value for the signal detection error rate of different modulation modes under an Additive White Gaussian Noise channel (AWGN) and a Rayleigh fading channel (Rayleigh) under the conditions of SISO and MIMO, and has stronger robustness for CFO and phase offset.
To achieve the above object, with reference to fig. 1, the present invention provides a method for generating a low-complexity SISO and MIMO receiver, where the method includes:
s1, sampling SISO and MIMO signals at a receiving end, wherein a sampling data set comprises signal data of multiple modulation modes of different channels under ideal conditions and non-ideal conditions, and the non-ideal conditions mean that the received signals contain signal defects including channel damage, CFO and phase offset;
s2, finding the optimal discriminant functionThe problem of detecting signals of different modulation modes in SISO and MIMO systems is converted into the problem of identifying waveforms, and the problem of detecting a compressed stack type self-encoder receiver is described as follows:
in the formula, r is a matrix formed by sampling signals of a receiving end, H is a channel response matrix, and theta is a parameter in a discriminant function;
s3, dividing the discriminant function f (r, H) in the step S2 into hidden layer discriminant functions fh(. DEG) and output layer discriminant function foTwo parts, using the sampling in step S1Training the receiver by the data set to ensure that the loss function corresponding to the discriminant function of the hidden layer and the output layer is minimum, and further obtaining the optimal parameters of the hidden layer and the output layer of the compression stack type self-encoder receiver as follows:
in the formula, thetah-optAnd thetao-optOptimal parameters for the implicit layer and the output layer of the compressed stack type self-encoder receiver respectively,andthe loss functions of the hidden layer and the output layer respectively,is the output characteristic matrix of the hidden layer,the method comprises the steps that an output matrix of a compression stack type self-encoder receiver after signal detection is carried out;
s4, according to the parameter optimization process in the step S3, based on the node number and the layer number selection strategy of the error rate measurement, selecting a few nodes with less node number and higher detection precision layer by layer as the candidate node number of the layer, and simultaneously selecting the nodes with least node number and higher precision of all layers, and minimizing the node number and the layer number of the compression stack type self-encoder receiver, so as to obtain the low-complexity compression stack type self-encoder receiver structure based on the node number and the layer number selection strategy.
In a further embodiment, in step S1, the sampling SISO and MIMO signals at the receiving end means:
s11, creating a signal sending and receiving model, wherein, for the MIMO system, the numbers of the sending antennas and the receiving antennas are respectively NtAnd NrIn which N isr≥NtFor SISO system, it is equivalent to Nr=N t1 and H-I1×1The MIMO system of (1); the transmission signal adopts four different modulation modes of BPSK, QPSK, 4PAM and 16QAM, and each transmission symbol adopts a rectangular pulse shaping functionWherein T issIs a symbol period;
s12, L samples, i.e., r, are taken of the received signal in each symbol periodi=[ri(0) ri(1) ...ri(L-1)]TMeanwhile, the number of symbols sampled is M, and the size r of the data set formed after sampling is ═ r0r1... rM-1]L×M。
In a further embodiment, in step S2, the finding of the optimal discriminant functionThe process of converting the signal detection problem of different modulation modes in SISO and MIMO systems into the waveform identification problem comprises the following steps:
s21, extracting features from the input sample data;
s22, training a classifier;
s23, finding out the optimal discriminant function under different modulation waveforms, and identifying and demodulating the input waveform through the optimal discriminant function; the optimal discriminant function comprises a hidden layer discriminant function and an output layer discriminant function, and each part contains a parameter theta related to the layer structurehAnd thetao。
In a further embodiment, in step S3, when the optimal parameter of the discriminant function is found, the loss function is lost through the hidden layerAnd output layer loss functionMinimization is performed to achieve optimization of the parameters, wherein the property requirements of the sampled data set in step S1 are determined according to the convergence speed requirements of the aforementioned loss function and the accuracy requirements of the signal demodulation.
In a further embodiment, in step S4, the strategy for selecting the number of nodes and the number of layers based on the bit error rate metric is to simplify the number of nodes and the number of layers by using a sampled data set of BPSK in an AWGN channel;
and (3) fine-tuning the number and the layer number of the simplified nodes by adopting QPSK, 4PAM and 16QAM sampling data sets under AWGN and Rayleigh channels respectively to obtain a stable low-complexity compression stack type self-encoder receiver structure capable of realizing signal detection under various modulation modes and different channels.
In a further embodiment, in step S4, the process of obtaining a low-complexity compressed stacked self-encoder receiver structure based on the node number and layer number selection policy includes the following steps:
s41, setting NiRepresenting the number of neurons on the i-th hidden layer, k being the number of sampling points per symbol, and Delta being used as a measure for determining the theoretical bit error rate Delta under the modulation modetBit error rate delta detected by a compression stack self-encoder receivercThe error between;
s42, adopting BPSK sampling data set under AWGN channel as training data of network, and counting N nodes on the first hidden layer1Iteration is respectively carried out from 1 to k, network output detection accuracy corresponding to different node numbers is recorded, the detection accuracy is selected from the detection performances of the different node numbers to meet the requirement of a preset accuracy threshold, a plurality of nodes with the minimum node number are used as the candidate node number of the first hidden layer, and delta is judgedc-δtIf yes, go to step S44, otherwise, go to step S43;
s43, setting the number of nodes in the current hidden layer as the candidate nodes obtained in the step S42Increasing the number of the next hidden layers, respectively changing the number of nodes in the next hidden layers, iteratively selecting according to the strategy in the step S42, determining the number of candidate nodes in the next layer, and repeating the operation until a given condition delta is metc-δt<Δ;
S44, selecting candidate nodes with the highest output detection accuracy and the least number of nodes in each hidden layer as the nodes of the network and the configuration of the layer number;
s45, fine adjustment of network configuration is carried out by adopting different modulation modes and sampling data sets under channels, and the optimal node number and layer number configuration of the compression stack type self-encoder receiver are determined;
and S46, under SISO and MIMO conditions, respectively, adopting four different modulation modes of BPSK, QPSK, 4PAM and 16QAM to perform signal detection on the generated low-complexity compression stacked self-encoder receiver under the AWGN and Rayleigh channels under ideal conditions and non-ideal conditions.
Compared with the prior art, the technical scheme of the invention has the following remarkable beneficial effects:
(1) the CSAE receiver provided by the invention does not depend on a channel model and a noise hypothesis, and meanwhile, the NLNS strategy designs the number of nodes and the number of layers of each layer of the deep neural network receiver by using a Bit Error Rate (BER) as a measurement, so that the structure and the complexity of the receiver can be simplified.
(2) The CSAE receiver has a simple structure, only comprises 2 hidden layers, and can reach or exceed the optimal detection theoretical value for the signal detection error rate of different modulation modes under an Additive White Gaussian Noise channel (AWGN) and a Rayleigh fading channel (Rayleigh) under the conditions of SISO and MIMO, and has stronger robustness for CFO and phase offset.
(3) The invention simulates SISO, MIMO and signals of various modulation modes under different channel models under ideal and non-ideal conditions, and carries out comparative analysis with the traditional optimal detection performance, the detection performance of the CSAE receiver under the ideal condition reaches or exceeds the optimal theoretical value, and the CSAE receiver has stronger robustness under the non-ideal condition.
It should be understood that all combinations of the foregoing concepts and additional concepts described in greater detail below can be considered as part of the inventive subject matter of this disclosure unless such concepts are mutually inconsistent. In addition, all combinations of claimed subject matter are considered a part of the presently disclosed subject matter.
The foregoing and other aspects, embodiments and features of the present teachings can be more fully understood from the following description taken in conjunction with the accompanying drawings. Additional aspects of the present invention, such as features and/or advantages of exemplary embodiments, will be apparent from the description which follows, or may be learned by practice of specific embodiments in accordance with the teachings of the present invention.
Drawings
The drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component that is illustrated in various figures may be represented by a like numeral. For purposes of clarity, not every component may be labeled in every drawing. Embodiments of various aspects of the present invention will now be described, by way of example, with reference to the accompanying drawings, in which:
fig. 1 is a flow chart of the low complexity SISO and MIMO receiver generation method based on a compression stacked self-encoder of the present invention.
Fig. 2 is a block diagram of a receiver structure of a compression-stacked self-encoder according to the present invention.
Fig. 3 is a graph illustrating the effect of SNR of the training set on detection performance of a CSAE receiver.
FIG. 4 is a graph of the error rate performance of a SISO-CSAE receiver over an AWGN channel.
Fig. 5 is a diagram illustrating the error rate performance of a SISO-CSAE receiver in a Rayleigh channel.
Fig. 6 is a schematic diagram of the error rate performance of the MIMO-CSAE receiver under the Rayleigh channel.
Detailed Description
In order to better understand the technical content of the present invention, specific embodiments are described below with reference to the accompanying drawings.
The design and signal detection of the low-complexity CSAE receiver provided by the invention adopt the following steps:
step 1: sampling SISO and MIMO signals at a receiving end, wherein for the MIMO communication system, the numbers of transmitting antennas and receiving antennas are respectively NtAnd NrIn which N isr≥NtFor SISO system, N is the equivalentr=N t1 and H-I1×1The MIMO system adopts four different modulation modes of BPSK, QPSK, 4PAM and 16QAM as the transmission signals, and each transmission symbol adopts a rectangular pulse shaping functionWherein T issFor a symbol period, the received signal typically contains channel impairments, CFO and phase offset;
step 2: the receiver finds the optimal discriminant function through a large amount of sampling data in step 1Therefore, the problem of signal detection of different modulation modes in SISO and MIMO systems is converted into the problem of waveform identification, so the detection problem of the CSAE receiver can be described as
In the formula, r is a matrix formed by sampling signals of a receiving end, H is a channel response matrix, and theta is a parameter in a discriminant function;
and step 3: the discriminant function f (r, H) in step 2 can be divided into hidden layer discriminant functions fh(. DEG) and output layer discriminant function foTraining the receiver by using the sampling data set in the step 1 to minimize the loss function corresponding to the discriminant functions of the hidden layer and the output layer, and further obtaining the optimal parameters of the hidden layer and the output layer of the CSAE receiver
In the formula, thetah-optAnd thetao-optOptimal parameters for the hidden layer and the output layer of the CSAE receiver respectively,andthe loss functions of the hidden layer and the output layer respectively,is the output characteristic matrix of the hidden layer,is an output matrix of the CSAE receiver after signal detection;
and 4, step 4: according to the parameter optimization process in the step 3, a node Number and layer Number Selection strategy (NLNS) based on BER measurement is designed, a few node numbers with small node numbers and high detection precision are selected layer by layer to serve as candidate node numbers of the layer, and simultaneously, nodes with minimum node numbers and high precision of all Layers are selected, so that the node Number and the layer Number of the CSAE receiver are minimized, and the low-complexity CSAE receiver structure based on the NLNS strategy is obtained.
The CSAE receiver provided by the invention does not depend on a channel model and a noise hypothesis, and meanwhile, the structure and the complexity of the receiver can be simplified by the provided NLNS strategy.
The low complexity compressed stacked self-encoder receiver proposed by the present invention has been validated on Matlab platform. The simulation result shows that the detection performance of the receiver under the ideal condition can reach or exceed the optimal theoretical value, and the receiver has stronger robustness to CFO and phase offset. Fig. 2 is a block diagram of a compressed stacked self-encoder receiver designed based on NLNS strategy.
The following gives specific steps implemented in one example of the present invention:
(1) firstly, the signals of different modulation modes sent by a receiving end are sampled, and the received signals under SISO and MIMO conditions can be expressed as:
in the formula, riIs the signal on the ith receiving antenna, αnIs a baseband transmission data sequence adopting a certain modulation mode, wherein TsFor the symbol period, f is the carrier frequency offset, ζ is the phase offset, hi,jIs the channel response between the jth transmit antenna and the ith receive antenna, Z ═ Z1,z2…zNr]TIs a mean of 0 and a variance of σ2Independent and equally distributed white gaussian noise. The sampling value of the formula (4) under a certain modulation mode can be expressed as
Wherein r isi=[ri(0) ri(1) ... ri(L-1)]TIs the number of sample points within each symbol and M represents the number of cycles of sampling.
(2) The CSAE receiver trains the network according to the signals sampled by the formula (5) and finds the optimal discriminant functionThereby minimizing the error probability function at the receiving endThe transmitted signal is reconstructed. CSAE reception under optimal discriminant functionThe detection problem of the machine can be further expressed as:
in the formula (I), the compound is shown in the specification,is the overall loss function of CSAE, which is used to measure the reconstructed signalAnd a transmission signal sjThe error between. Total loss functionIncluding a hidden layer loss functionAnd output layer loss function
(3) The hidden layer loss function uses Euclidean square norm, KL divergence is used as a regularization item to avoid overfitting in training, the output layer uses a Softmax classifier to estimate each input data, and the maximum likelihood criterion is used for waveform prediction of the input data after the hidden layer characteristic extraction, namely:
where Q is the size of the sampled data set and θhIs a parameter of the hidden layer, lambda is a weight factor, m is the number of neurons of the hidden layer, chi is a sparse coefficient,is the mean activation degree, theta, of neurons m in the hidden layergIs the parameter of the output layer, 1 {. is the indication function, N is the level kind of the signal under a certain modulation mode.
(4) And (3) training and iterating the CSAE receiver according to the loss function expression in the step (3), and obtaining the optimal parameters of the receiver by minimizing the loss function, namely the formula (2) and the formula (3).
(5) According to the CSAE receiver training method in (4), the number of nodes and the number of layers of each layer of the compression stack type self-encoder receiver are determined by BER measurement, thereby simplifying the structure of the receiver. The number of nodes and tier number selection strategy based on the BER metric is as follows, where N isiRepresents the number of neurons on the ith hidden layer, k is the number of sampling points of each symbol (in this example, the sampling point is set to be 8), and Delta is used for measuring and determining the theoretical bit error rate Delta under the modulation modetBit error rate delta with CSAE receiver detectioncThe error between.
① the number of nodes N on the first hidden layer is determined by using the sampled data set of BPSK in AWGN channel as the training data of the network1And respectively iterating from 1 to 8, and recording the network output detection accuracy corresponding to different node numbers. From the detection performances of the different node numbers, 4 nodes with higher detection accuracy and fewer node numbers are selected as the candidate node number of the first hidden layer. If deltac-δtA further hidden layer needs to be added > delta.
② the number of nodes in the first hidden layer is the number of candidate nodes in S1, the number of nodes in the second hidden layer is changed respectively, and iterative selection is performed according to the strategy in S1, thereby determining the number of candidate nodes in the second layer.
③ repeat the above operations until the given condition delta is satisfiedc-δtAnd delta, selecting the candidate node with the highest output detection accuracy and the least number of nodes in each hidden layer as the node of the network and the configuration of the layer number.
④, fine-tuning the network configuration by using different modulation modes and sampling data sets under channels, and further determining the optimal node number and layer number configuration of the CSAE receiver.
(6) And (3) under SISO and MIMO conditions, performing signal detection on the low-complexity CSAE receiver designed in the step (5) under ideal conditions and non-ideal conditions by adopting four different modulation modes of BPSK, QPSK, 4PAM and 16QAM under AWGN and Rayleigh channels.
Fig. 3 shows the effect of Signal to Noise Ratio (SNR) settings of the training set on detection performance of a CSAE receiver. Under an 8 × 20K data set, a network is trained by using a data set with SNR (signal to noise ratio) of 1dB, and the error code performance of a CSAE receiver reaches the theoretical performance; when the network is trained using SNR over the entire training set, the detection performance of the CSAE receiver is substantially the same as the result of the training network when SNR is 1dB, with a gain of about 0.1dB at only 7 dB. Fig. 4, 5 and 6 use CSAE receivers to demodulate different modulated signals under SISO and MIMO conditions, respectively. According to simulation results, under ideal conditions, the detection performances of BPSK, QPSK and 4PAM under AWGN and Rayleigh channels reach the optimal detection theoretical value, the detection performance of 16QAM under AWGN channel with low SNR exceeds the theoretical value by about 1dB, and under non-ideal SISO and MIMO conditions, the detection of the CSAE receiver has strong robustness.
In this disclosure, aspects of the present invention are described with reference to the accompanying drawings, in which a number of illustrative embodiments are shown. Embodiments of the present disclosure are not necessarily defined to include all aspects of the invention. It should be appreciated that the various concepts and embodiments described above, as well as those described in greater detail below, may be implemented in any of numerous ways, as the disclosed concepts and embodiments are not limited to any one implementation. In addition, some aspects of the present disclosure may be used alone, or in any suitable combination with other aspects of the present disclosure.
Although the present invention has been described with reference to the preferred embodiments, it is not intended to be limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, the protection scope of the present invention should be determined by the appended claims.
Claims (6)
1. A method for generating a low complexity SISO and MIMO receiver, the method comprising:
s1, sampling SISO and MIMO signals at a receiving end, wherein a sampling data set comprises signal data of multiple modulation modes of different channels under ideal conditions and non-ideal conditions, and the non-ideal conditions mean that the received signals contain signal defects including channel damage, CFO and phase offset;
s2, finding the optimal discriminant functionThe problem of detecting signals of different modulation modes in SISO and MIMO systems is converted into the problem of identifying waveforms, and the problem of detecting a compressed stack type self-encoder receiver is described as follows:
in the formula, r is a matrix formed by sampling signals of a receiving end, H is a channel response matrix, and theta is a parameter in a discriminant function;
s3, dividing the discriminant function f (r, H) in the step S2 into hidden layer discriminant functions fh(. DEG) and output layer discriminant function foTwo parts, training the receiver by using the sampling data set in step S1 to minimize the loss function corresponding to the discriminant function of the hidden layer and the output layer, and further obtaining the optimal parameters of the hidden layer and the output layer of the compressed stacked self-encoder receiver as follows:
in the formula, thetah-optAnd thetao-optReceiver hiding for compression stack type self-encoderIncluding the optimal parameters of the layers and the output layer,andthe loss functions of the hidden layer and the output layer respectively,is the output characteristic matrix of the hidden layer,the method comprises the steps that an output matrix of a compression stack type self-encoder receiver after signal detection is carried out;
s4, according to the parameter optimization process in the step S3, based on the node number and the layer number selection strategy of the error rate measurement, selecting a few nodes with less node number and higher detection precision layer by layer as the candidate node number of the layer, and simultaneously selecting the nodes with least node number and higher precision of all layers, and minimizing the node number and the layer number of the compression stack type self-encoder receiver, so as to obtain the low-complexity compression stack type self-encoder receiver structure based on the node number and the layer number selection strategy.
2. The method for generating a low complexity SISO and MIMO receiver of claim 1, wherein in step S1, said sampling SISO and MIMO signals at the receiving end means:
s11, creating a signal sending and receiving model, wherein, for the MIMO system, the numbers of the sending antennas and the receiving antennas are respectively NtAnd NrIn which N isr≥NtFor SISO system, it is equivalent to Nr=Nt1 and H-I1×1The MIMO system of (1); the transmission signal adopts four different modulation modes of BPSK, QPSK, 4PAM and 16QAM, and each transmission symbol adopts a rectangular pulse shaping functionWherein T issIs a symbol period;
s12, L samples, i.e., r, are taken of the received signal in each symbol periodi=[ri(0) ri(1)...ri(L-1)]TMeanwhile, the number of symbols sampled is M, and the size r of the data set formed after sampling is ═ r0r1...rM-1]L×M。
3. The method of claim 1, wherein in step S2, the finding of the optimal discriminant function is performedThe process of converting the signal detection problem of different modulation modes in SISO and MIMO systems into the waveform identification problem comprises the following steps:
s21, extracting features from the input sample data;
s22, training a classifier;
s23, finding out the optimal discriminant function under different modulation waveforms, and identifying and demodulating the input waveform through the optimal discriminant function; the optimal discriminant function comprises a hidden layer discriminant function and an output layer discriminant function, and each part contains a parameter theta related to the layer structurehAnd thetao。
4. The method of claim 1, wherein in step S3, when performing the optimal parameter search for the discriminant function, the loss function of the hidden layer is passedAnd output layer loss functionMinimization to optimize the parameters, wherein the parameters are based on the aforementioned lossesThe convergence speed requirement of the loss function and the accuracy requirement of the signal demodulation determine the property requirement of the sample data set in step S1.
5. The method for generating low complexity SISO and MIMO receiver according to claim 2, wherein in step S4, the strategy for selecting the number of nodes and the number of layers based on the bit error rate measure is,
simplifying the number of nodes and the number of layers by adopting a sampling data set of BPSK (binary phase shift keying) under an AWGN (AWGN) channel;
and (3) fine-tuning the number and the layer number of the simplified nodes by adopting QPSK, 4PAM and 16QAM sampling data sets under AWGN and Rayleigh channels respectively to obtain a stable low-complexity compression stack type self-encoder receiver structure capable of realizing signal detection under various modulation modes and different channels.
6. The method of claim 1, wherein in step S4, the step of obtaining a low complexity compressed stacked self-encoder receiver structure based on node number and layer number selection strategy comprises the steps of:
s41, setting NiRepresenting the number of neurons on the i-th hidden layer, k being the number of sampling points per symbol, and Delta being used as a measure for determining the theoretical bit error rate Delta under the modulation modetBit error rate delta detected by a compression stack self-encoder receivercThe error between;
s42, adopting BPSK sampling data set under AWGN channel as training data of network, and counting N nodes on the first hidden layer1Iteration is respectively carried out from 1 to k, network output detection accuracy corresponding to different node numbers is recorded, the detection accuracy is selected from the detection performances of the different node numbers to meet the requirement of a preset accuracy threshold, a plurality of nodes with the minimum node number are used as the candidate node number of the first hidden layer, and delta is judgedc-δtIf yes, go to step S44, otherwise, go to step S43;
s43, hiding the nodes in the current layerSetting the number as the number of candidate nodes obtained in step S42, adding the next hidden layer, respectively changing the number of nodes in the next hidden layer, performing iterative selection according to the strategy in step S42, determining the number of candidate nodes in the next layer, and repeating the above operations until a given condition delta is satisfiedc-δt<Δ;
S44, selecting candidate nodes with the highest output detection accuracy and the least number of nodes in each hidden layer as the nodes of the network and the configuration of the layer number;
s45, fine adjustment of network configuration is carried out by adopting different modulation modes and sampling data sets under channels, and the optimal node number and layer number configuration of the compression stack type self-encoder receiver are determined;
and S46, under SISO and MIMO conditions, respectively, adopting four different modulation modes of BPSK, QPSK, 4PAM and 16QAM to perform signal detection on the generated low-complexity compression stacked self-encoder receiver under the AWGN and Rayleigh channels under ideal conditions and non-ideal conditions.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010090637.XA CN111314255B (en) | 2020-02-13 | 2020-02-13 | Low-complexity SISO and MIMO receiver generation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010090637.XA CN111314255B (en) | 2020-02-13 | 2020-02-13 | Low-complexity SISO and MIMO receiver generation method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111314255A true CN111314255A (en) | 2020-06-19 |
CN111314255B CN111314255B (en) | 2021-06-08 |
Family
ID=71147034
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010090637.XA Expired - Fee Related CN111314255B (en) | 2020-02-13 | 2020-02-13 | Low-complexity SISO and MIMO receiver generation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111314255B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114124630A (en) * | 2021-11-12 | 2022-03-01 | 中国电子科技集团公司第二十八研究所 | Symbol detection and modulation identification method based on compact self-encoder |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101442388A (en) * | 2008-12-29 | 2009-05-27 | 北京邮电大学 | Precoding method and apparatus for multi-input multi-output system |
CN105142177A (en) * | 2015-08-05 | 2015-12-09 | 西安电子科技大学 | Complex neural network channel prediction method |
US20170076001A1 (en) * | 2015-09-14 | 2017-03-16 | Thomson Reuters Global Resources | Docket search and analytics engine |
CN109688076A (en) * | 2018-04-19 | 2019-04-26 | 南京邮电大学 | The blind checking method of noise chaotic neural network based on discrete more level sluggishnesses |
CN109714086A (en) * | 2019-01-23 | 2019-05-03 | 上海大学 | Optimization MIMO detection method based on deep learning |
CN109831400A (en) * | 2019-03-06 | 2019-05-31 | 深兰科技(上海)有限公司 | A kind of received method and apparatus of OFDM signal |
CN110061946A (en) * | 2019-03-28 | 2019-07-26 | 南京邮电大学 | A kind of depth signal detection method towards high-speed rail |
CN110191073A (en) * | 2019-06-28 | 2019-08-30 | 华侨大学 | A kind of Modulation Mode Recognition method based on deep learning suitable for scene change |
CN110719239A (en) * | 2019-09-29 | 2020-01-21 | 东南大学 | Data model dual-drive combined MIMO channel estimation and signal detection method |
-
2020
- 2020-02-13 CN CN202010090637.XA patent/CN111314255B/en not_active Expired - Fee Related
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101442388A (en) * | 2008-12-29 | 2009-05-27 | 北京邮电大学 | Precoding method and apparatus for multi-input multi-output system |
CN105142177A (en) * | 2015-08-05 | 2015-12-09 | 西安电子科技大学 | Complex neural network channel prediction method |
US20170076001A1 (en) * | 2015-09-14 | 2017-03-16 | Thomson Reuters Global Resources | Docket search and analytics engine |
CN109688076A (en) * | 2018-04-19 | 2019-04-26 | 南京邮电大学 | The blind checking method of noise chaotic neural network based on discrete more level sluggishnesses |
CN109714086A (en) * | 2019-01-23 | 2019-05-03 | 上海大学 | Optimization MIMO detection method based on deep learning |
CN109831400A (en) * | 2019-03-06 | 2019-05-31 | 深兰科技(上海)有限公司 | A kind of received method and apparatus of OFDM signal |
CN110061946A (en) * | 2019-03-28 | 2019-07-26 | 南京邮电大学 | A kind of depth signal detection method towards high-speed rail |
CN110191073A (en) * | 2019-06-28 | 2019-08-30 | 华侨大学 | A kind of Modulation Mode Recognition method based on deep learning suitable for scene change |
CN110719239A (en) * | 2019-09-29 | 2020-01-21 | 东南大学 | Data model dual-drive combined MIMO channel estimation and signal detection method |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114124630A (en) * | 2021-11-12 | 2022-03-01 | 中国电子科技集团公司第二十八研究所 | Symbol detection and modulation identification method based on compact self-encoder |
CN114124630B (en) * | 2021-11-12 | 2023-10-24 | 中国电子科技集团公司第二十八研究所 | Symbol detection and modulation recognition method based on compact self-encoder |
Also Published As
Publication number | Publication date |
---|---|
CN111314255B (en) | 2021-06-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7817740B2 (en) | Method and system for minimum mean squared error soft interference cancellation (MMSE-SIC) based suboptimal maximum likelihood (ML) detection for multiple input multiple output (MIMO) wireless system | |
KR20070081786A (en) | An apparatus and method for multiple input multiple output in communication | |
CN112215335B (en) | System detection method based on deep learning | |
CN1116476A (en) | Reception method and CDMA receiver | |
CN107835068B (en) | Low-complexity orthogonal space modulation spherical decoding detection algorithm with transmit diversity | |
CN111628833B (en) | MIMO antenna number estimation method based on convolutional neural network | |
CN101588335A (en) | Utilize the MIMO detection method and the system of channel relevancy | |
US8243843B2 (en) | Systems and methods for low-complexity MIMO detection using leaf-node prediction via look-up tables | |
CN101026435A (en) | Low-complexity maximum likelihood detecting method and device for communication system | |
Zheng et al. | DemodNet: Learning soft demodulation from hard information using convolutional neural network | |
CN111314255B (en) | Low-complexity SISO and MIMO receiver generation method | |
CN109286587B (en) | Multi-active generalized spatial modulation detection method | |
CN109818891B (en) | Lattice reduction assisted low-complexity greedy sphere decoding detection method | |
CN112636855A (en) | OFDM signal detection method | |
CN109818663A (en) | A kind of low complex degree difference quadrature spatial modulation detection method | |
CN113489545B (en) | Light space pulse position modulation step-by-step classification detection method based on K-means clustering | |
CN101197603A (en) | Low complexity step-by-step detecting system and method of multi-antenna system based on spherical decoding | |
US20080232491A1 (en) | Systems and methods for low-complexity mimo detection with analytical leaf-node prediction | |
CN109167648B (en) | Candidate constellation point set generation method and MIMO spatial multiplexing detection method | |
US8477880B2 (en) | Method and apparatus for signal detection based on MMSE in MIMO communication system | |
KR101348557B1 (en) | Method for detecting signal using mimo-ofdm system and apparatus thereof | |
KR101937559B1 (en) | Linear approximation signal detection apparatus using MIMO-OFDM system and method thereof | |
US8081577B2 (en) | Method of calculating soft value and method of detecting transmission signal | |
GB2439770A (en) | Decision error compensation in wireless MIMO receivers | |
KR101342626B1 (en) | Apparatus and method for low-complexity detection based on unified iterative tree searching in multiple input multiple output systems |
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 | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210608 |
|
CF01 | Termination of patent right due to non-payment of annual fee |