CN115694571A - Signal detection method and device based on deep learning in large-scale MIMO system - Google Patents
Signal detection method and device based on deep learning in large-scale MIMO system Download PDFInfo
- Publication number
- CN115694571A CN115694571A CN202211348083.4A CN202211348083A CN115694571A CN 115694571 A CN115694571 A CN 115694571A CN 202211348083 A CN202211348083 A CN 202211348083A CN 115694571 A CN115694571 A CN 115694571A
- Authority
- CN
- China
- Prior art keywords
- signal
- received signal
- objective function
- deep learning
- interference
- 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 51
- 238000013135 deep learning Methods 0.000 title claims abstract description 26
- 238000000034 method Methods 0.000 claims abstract description 23
- 230000006870 function Effects 0.000 claims description 25
- 239000011159 matrix material Substances 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 7
- 238000013507 mapping Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 abstract description 35
- 230000008030 elimination Effects 0.000 abstract description 9
- 238000003379 elimination reaction Methods 0.000 abstract description 9
- 238000013528 artificial neural network Methods 0.000 abstract description 7
- 238000012549 training Methods 0.000 description 19
- 238000010586 diagram Methods 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 8
- 238000007476 Maximum Likelihood Methods 0.000 description 7
- 230000005540 biological transmission Effects 0.000 description 6
- 241001168730 Simo Species 0.000 description 5
- 238000004891 communication Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 239000000126 substance Substances 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000006978 adaptation Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000005562 fading Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- 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
- Radio Transmission System (AREA)
Abstract
The invention discloses a signal detection method and a device based on deep learning in a large-scale MIMO system, wherein the method obtains a preliminary estimation value of a transmitting signal through interference elimination in the first step; secondly, based on the obtained preliminary estimation value, carrying out interference elimination on the received signal of each receiving antenna; and thirdly, detecting the received signal after the interference elimination through a deep neural network to obtain an accurate estimation value of the transmitted signal. The method solves the technical problems of high algorithm complexity and low detection precision in the prior art. The method combining the interference elimination and the deep neural network greatly improves the detection precision of the signal under the condition of reducing the complexity of the algorithm.
Description
Technical Field
The invention belongs to the technical field of wireless signal detection, and particularly relates to a signal detection method and device based on deep learning in a large-scale MIMO system.
Background
The wireless communication is communication between a transmitting end and a receiving end without connection via a wire or a cable, and can be performed by radio waves, microwaves, infrared rays, and the like. The large-scale MIMO technology can fully mine space resources, obviously improve the frequency spectrum efficiency, and play a very important role in improving the data throughput of the modern wireless communication system. In a large-scale MIMO system, multiple data streams sent simultaneously on multiple antennas pass through a channel to a receiving end, so a detection technology for quickly and accurately extracting a desired signal from a superposed and distorted received signal is a core technology of a multi-antenna communication system, and implementation complexity and error code performance are two key indexes for evaluating the detection technology and are important bases for measuring whether the detection technology can be applied to an actual system.
In the classical MIMO detection technology, the linear detection method has an advantage in implementation complexity, and the basic idea is to consider other signals than the desired signal as interference, and detect the desired signal by minimizing the interference, and the simplest linear detection method includes a Matched Filter (MF), zero Forcing (ZF), and Minimum Mean Square Error (MMSE). The linear detection method has the advantages of simple implementation, low complexity and the like, but the performance is poor. For a large-scale MIMO system, the rapidly increased number of antennas and modulation order make the classical detection scheme face great challenges in the aspect of complexity, and the intelligent detection method based on deep learning provides a new idea for breaking through the bottleneck, and some methods are provided.
However, existing intelligent detection methods based on deep learning are all designed for a lower modulation scheme (such as BPSK or QPSK), and when a higher modulation scheme is met, the signal detection accuracy is greatly reduced.
Disclosure of Invention
The invention aims to provide a signal detection method and a signal detection device based on deep learning in a large-scale MIMO system, so as to improve the accuracy of signal detection and reduce the error rate in a higher modulation mode.
The invention adopts the following technical scheme:
the embodiment of the invention provides a signal detection method based on deep learning in a large-scale MIMO system, which comprises the following steps:
detecting the received signal by adopting a zero forcing detection method to obtain a preliminary estimation signal;
reconstructing an interference signal based on the preliminary estimated signal;
carrying out interference cancellation on the received signal by taking the interference signal as a reference;
and estimating the received signals after the interference cancellation by adopting a deep learning network to obtain accurate estimation signals of each transmitting antenna.
Optionally, the deep learning network comprises:
the number of nodes of the input layer is 2N r (ii) a Wherein, N r Is the number of receive antennas;
the number of nodes of the network output layer is M; wherein M is the modulation order of QAM;
7 hidden layers are arranged between the input layer and the output layer, and the number of nodes of the hidden layers is 128-64-64-32-32-64-64 in sequence.
Optionally, the detecting the received signal by using a zero-forcing detection method includes:
solving an objective function of the zero forcing detection method to obtain a solution of the objective function:
and calculating and obtaining a preliminary estimation signal based on the solution of the objective function.
Optionally, the objective function is:
wherein y is the received signal, x is the transmitted signal, | · | | calving 2 Denotes a 2 norm, H is the channel matrix, H T Is the transpose of the channel matrix (.) -1 Representing the inversion, u being the solution of the objective function, N t The number of transmit antennas.
Optionally, calculating and obtaining the preliminary estimation signal includes:
and mapping the solution of the objective function and the points in the constellation set of the transmitted signal to obtain a preliminary estimated signal.
Optionally, mapping the solution of the objective function to a point in the constellation set of the transmitted signal is implemented as follows:
wherein the content of the first and second substances,
in order to make a preliminary estimation of the signal,
representing the minimum euclidean distance between the transmit signal and the solution to the objective function.
Optionally, reconstructing the interference signal based on the preliminary estimated signal comprises:
and calculating the sum of products of the preliminary estimation signals of other transmitting antennas except the k-th transmitting antenna and the corresponding channel vector to obtain the interference signal.
Optionally, the interference cancellation of the received signal comprises:
the interference cancellation of the received signal is realized according to the following calculation mode;
where y is the received signal,
as an interference signal, y k Which represents the received signal of the kth receiving antenna after interference cancellation.
The second embodiment of the present invention provides a deep learning based signal detection apparatus in a massive MIMO system, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor executes the computer program to implement the deep learning based signal detection method in the massive MIMO system as described in any one of the above.
The invention has the beneficial effects that: the signal detection method based on deep learning in the large-scale MIMO system provided by the invention obtains a preliminary estimation value of a transmitting signal through interference elimination in the first step; secondly, based on the obtained preliminary estimation value, carrying out interference elimination on the received signal of each receiving antenna; and thirdly, detecting the received signal subjected to interference elimination through a deep neural network to obtain an accurate estimation value of the transmitted signal. The method combining the interference elimination and the deep neural network greatly improves the detection precision of the signal under the condition of reducing the complexity of the algorithm.
Drawings
Fig. 1 is a schematic diagram of a massive MIMO communication system according to an embodiment of the present invention;
fig. 2 is a flowchart illustrating steps of a signal detection method based on deep learning in a large-scale MIMO system according to an embodiment of the present invention;
FIG. 3 shows a block diagram of a block diagram N according to an embodiment of the present invention r ＝N t When =24, comparing error code performance of different detection algorithms;
FIG. 4 shows a block diagram of a computer system according to an embodiment of the present invention r ＝N t When the antenna is used, comparing error code performance of PIC-DNN under different antenna numbers;
FIG. 5 shows a block diagram of a block diagram N according to an embodiment of the present invention r When the error rate is increased, a PIC-DNN error code performance effect graph is obtained;
FIG. 6 shows a block diagram of a computer system according to an embodiment of the present invention r ＝64,N t The error code performance comparison chart of several algorithms under different modulation modes is defined as 32;
FIG. 7 shows a block diagram of a computer system according to an embodiment of the present invention r ＝N t And when =24, comparing the performance of the PIC-DNN in different training modes.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
With reference to fig. 1, the implementation principle of the signal detection method based on deep learning in the massive MIMO system provided by the present application is as follows:
firstly, detecting a received signal by adopting a simple linear detection algorithm to obtain a preliminary estimation signal of each transmitting signal; then, reconstructing an interference signal of a receiving end through the preliminary estimation signal; then, interference cancellation is carried out to cancel interference signals from the total received signals, the process simplifies the MIMO system into a plurality of parallel SIMO systems, the output of each SIMO system is the received signals after corresponding interference cancellation, and the signals do not contain the interference signals of other multiple transmitting antennas; and finally, detecting the received signal after the interference cancellation by adopting a DNN network, and estimating an accurate estimation signal of the transmitted signal. The implementation process is shown in fig. 1.
The received signal model corresponding to the system model shown in fig. 2 is:
assume that a massive MIMO system has N t Root transmitting antenna, N r (N r ≥N t ) The complex signal column vector of the transmitting antenna is
X c Representing a finite set of constellation points, x c Element x in (1) j (1≤j≤N t ) Represents the modulation signal on the jth transmitting antenna, and the complex number domain receiving signal of the system
Is expressed as shown in the following formula (1):
y c ＝H c x c +n c (1)
wherein the content of the first and second substances,
then y is i (1≤i≤N r ) The received signal of the ith receiving antenna.
Representing a complex channel transmission matrix, H c Element h in (1) i，j (1≤i≤N r ，1≤j≤N t ) Indicating the ith transmitting antenna and the ith receiving antenna of the transmitting terminalj receives the channel gain between the antennas.
Is a complex gaussian noise that is a function of,
is N r ×N r The unit matrix of (2).
If the complex domain system model is subjected to a real number operation, the received signal of the real domain of the system can be represented as:
wherein the content of the first and second substances,
is a real number field 2N r ×2N t The channel matrix of the dimension(s),
is a real number field 2N t X 1-dimensional transmit signal vector, x being x c A finite set of constellation points for the in-phase or quadrature branch of (a),
is a real number field 2N r The x 1-dimensional received signal vector. Denotes H as
Wherein, the first and the second end of the pipe are connected with each other,
for the channel vector of the jth receive antenna,
for real domain gaussian noise, equation (2) can be further expressed as:
based on the system model and the received signal model, an embodiment of the present invention provides a signal detection method based on deep learning in a massive MIMO system, as shown in fig. 2, including the following steps:
s101, detecting a received signal by adopting a zero forcing detection method to obtain a primary estimation signal;
step S102, reconstructing an interference signal based on the preliminary estimation signal;
step S103, taking the interference signal as a reference, and carrying out interference cancellation on the received signal;
and step S104, estimating the received signals after the interference cancellation by adopting a deep learning network to obtain accurate estimation signals of each transmitting antenna.
In a specific embodiment, step S101 employs simple zero-forcing (ZF) detection on the received signal, and the objective function of the zero-forcing detection method is:
wherein y is the received signal, x is the transmitted signal, | · | | non |, Y 2 Denotes a 2 norm, H is the channel matrix, H T Is the transpose of the channel matrix (.) -1 Representing the inversion, u being the solution of the objective function.
Further, mapping the solution of the objective function and the point in the constellation set of the transmission signal to obtain a preliminary estimation signal of the transmission signal is realized according to the following calculation mode:
wherein the content of the first and second substances,
in order to initially estimate the signal(s),
representing the minimum euclidean distance between the transmit signal and the solution to the objective function.
Step S102, calculating the sum of products of the preliminary estimation signals of other transmitting antennas except the kth transmitting antenna and the corresponding channel vector to obtain an interference signal, for example:
is a jamming signal in which, among other things,
preliminary estimation of the signal for the jth receiving antenna, h j And the channel vector corresponding to the jth receiving antenna.
Step S103, at the receiving end, the interference signal is reconstructed by using the preliminary estimation signal of the detected transmitting signal and is cancelled from the total receiving signal, namely 2N t ×2N r The MIMO system is simplified into 2N t 1 x 2N in parallel r SIMO system of (3).
For example: if the signal on the kth transmitting antenna of the transmitting end is the desired signal, after interference cancellation, the received signal of the kth SIMO system can be represented as:
wherein, y is a received signal,
as an interference signal, y k Which represents the received signal of the kth receiving antenna after interference cancellation.
Step S104, the received signal y after the interference elimination is carried out k Adopting DNN network to detect, and obtaining accurate estimation value of kth transmitting antenna by using strong nonlinear mapping capability of DNN
In one embodiment, DNN is a classifier, and the input vector is y k Dimension of 2N r X 1, the net output vector can be set to z = [ z ] 1 z 2 … z M ] T The dimension of the signal x is Mx 1, M and the k transmitting antenna k Modulation order of the constellation modulation.
During the DNN network training phase, the loss function of the network employs a cross-entropy loss function, which may be defined as:
wherein H (p, q) represents cross entropy, p = [ p ] 1 p 2 … p M ]Is the k-th transmitting antenna signal x k One-hot encoded vector of (1), q = [ q ] ([ q ]) 1 q 2 … q M ]For output of z after Softmax, softmax is an activation function, i.e.
The network model training is completed through a minimum loss function, and the accurate estimation value of the kth transmitting antenna can be obtained by carrying out hard decision on the maximum output branch of the network output vector on a constellation in a test stage based on the trained DNN network
Wherein argmax (·) represents the index corresponding to the returned maximum value, and D (·) represents the hard decision on the constellation, i.e., mapping from the modulation index to the constellation point value is implemented.
In a specific embodiment, the method of steps S101 to S104 can be summarized as follows:
the method mainly includesThe first stage realizes the pre-processing of the received signal, and the zero forcing detector is first used to estimate the transmitted signal initially
Then the interference cancellation is utilized to obtain the equivalent receiving signal y of the kth SIMO system k . In the second stage, DNN is applied to the preprocessed signals for signal detection, the DNN is trained, and then the trained network is used for detection, so that accurate estimation of the expected signals is achieved
Through two estimation stages, the technical effect of improving the estimation precision of the transmitted signal is achieved, and the specific implementation mode is shown in table 1.
TABLE 1
It should be noted that the PIC-DNN network is a deep fully-connected network, and includes: the number of nodes of the input layer is 2N r Wherein, 2N is r Is the number of receive antennas; the node number of the network output layer is M, wherein M is the modulation order of QAM; 7 hidden layers are arranged between the input layer and the output layer, and the number of nodes of the hidden layers is 128-64-64-32-32-64-64 in sequence. The specific node numbers of the hidden layers and other related parameter settings are shown in table 2.
The PIC in this embodiment actually refers to Parallel Interference Cancellation (PIC), the DNN refers to Deep Neural Networks (DNN), and the PIC-DNN refers to Deep Neural Networks (PIC-DNN) based on Parallel Interference Cancellation.
TABLE 2
An experimental simulation result and performance analysis based on the method are as follows:
in the experimental environment, a large-scale MIMO system is considered, an MIMO channel is an independent and identically distributed Rayleigh fading (Rayleigh fading) channel, channel Noise is zero-mean i.i.d. gaussian Noise, variance is set according to a Signal-to-Noise Ratio (SNR) of work, channel coding is not considered, and a modulation mode and the number of antennas are determined according to each experiment. The simulation software platform adopts a Pycharm IDE environment, the deep learning framework adopts Tensorflow, and a Maximum Likelihood (ML) detection algorithm realizes optimization of the package by means of Gurobi.
The performance of the proposed PIC-DNN algorithm with other classical MIMO detection algorithms was evaluated and compared by some related experiments. The method mainly evaluates the Error code performance and the implementation complexity, wherein the Error code performance adopts Symbol Error Rate (SER) as an evaluation index, and the implementation complexity adopts universal time complexity for analysis.
With reference to fig. 3, the number of transmitting antennas N is given t And number of receiving antennas N r 24, under the condition that the modulation mode is QPSK, the error code performance of the proposed PIC-DNN algorithm is compared with the error code performance of other classic ZF, MMSE, PIC, SDR and ML algorithms, and the figure shows that ZF has the performance of simple structure and basically fails in a large-scale MIMO system, the MMSE algorithm gives consideration to the influence of noise while decorrelating channels, the detection performance of the MMSE algorithm is superior to that of ZF, but the performance of the large-scale MIMO system is still poor. The ML algorithm is implemented based on an exhaustive search mode, and its performance is optimal, but the implementation complexity increases exponentially with the modulation order and the number of antennas, which is not acceptable for a large-scale MIMO system, and is usually only used as a theoretical reference. The SDR algorithm is a compromise between performance and complexity of the ML algorithm, and its error performance is suboptimal, but its complexity is still high. Therefore, the PIC-DNN algorithm provided by the invention mainly comprises three stages in implementation, wherein the first stage adopts ZF algorithm, the first two stages are the same as the PIC, and the difference between the third stage and the PIC is that the PIC adopts ZF algorithm, the known channel state information is needed in implementation and the channel needs to be carried outAnd the PIC-DNN adopts a deep neural network to estimate channel state information through training without matrix pseudo-inverse operation, and the PIC-DNN performance is obviously superior to ZF, MMSE and PIC algorithms. When SER =10 -3 When the signal strength is higher than that of the signal strength, the PIC-DNN is about 9dB lower than the SNR required by MMSE, and the SER of the PIC-DNN is close to 10 when the SNR =18dB -5 。
In conjunction with fig. 4, the advantage of MIMO technology is that multiple antennas can achieve significant improvement in system performance through spatial multiplexing and spatial diversity. Therefore, the proposed PIC-DNN algorithm was further evaluated experimentally for performance at different antenna counts. For the application scene of MIMO spatial multiplexing, N of the transmitting terminal t Root transmitting antenna sending N in parallel t The independent data streams can greatly improve the transmission rate of the system and the effectiveness of the system, and the number of the antennas at the receiving end is not less than N t Thus, FIG. 4 shows that at N r ＝N t The error code performance of the PIC-DNN can be seen, the error code performance of the system is obviously improved along with the reduction of the number of the antennas, and when SER =10 -4 The number of antennas 24 has a performance improvement of about 5dB over the number of antennas 64, so that a compromise between system capacity and error performance can be made depending on the application requirements.
In conjunction with FIG. 5, for the application scenario of MIMO spatial diversity, N is used r >N t The receiving end can obtain diversity gain and improve the reliability of the system, and fig. 5 shows that the PIC-DNN algorithm under QPSK modulation is in N r >N t The error code performance is compared with the graph, and the graph shows that N is fixed t With N r The performance of the PIC-DNN algorithm is obviously improved. When N is present r Than N t Doubled, SER =10 -3 The required SNR is reduced by about 6dB. Therefore, in order to improve the reliability of the system, the number N of receiving antennas is increased appropriately r Is an effective method.
It should be noted that in the MIMO system, the Link adaptation (Link adaptation) technology can dynamically adjust parameters such as the transmission power, the coding mode, the modulation mode, and the modulation order according to the channel condition. When the channel condition is not good, namely the SNR is low, low-order modulation is adopted; when the channel conditions are good, i.e. highHigh-order modulation is adopted under SNR, so that the transmission rate is improved to the maximum extent, and the error rate is reduced. For this reason, in FIG. 6, N is compared r ＝64、N t =32, performance under different modulation schemes in QPSK, 16QAM, and 64QAM modulation schemes, respectively. For example in the presence of diversity gain (N) r >N t ) In the case of (1), the modulation mode is QPSK modulation, the performance of SDR is superior to that of ZF, MMSE and PIC, but the difference between the SDR and ML is increased along with the increase of the modulation order, the SDR gradually approaches to ZF, and the performance is unstable. Under the condition of high signal-to-noise ratio, the performance of MMSE is gradually close to ZF, main factors influencing the performance of MMSE and ZF are decorrelation processing of a channel, the influence of noise on the performance is gradually weakened, and the MMSE algorithm is gradually reduced to ZF. In theory, AMP is asymptotically optimal for gaussian channels, but in the case of 64QAM high order modulation, there is a problem of poor robustness. In comparison, the performance of the proposed PIC-DNN algorithm is close to ML and is superior to ZF, MMSE, PIC, AMP and SDR algorithms no matter in low-order modulation or high-order modulation, and the algorithm has good robustness.
In order to reduce the training time of the PIC-DNN algorithm and the robustness of the verification algorithm under different SNRs, the PIC-DNN network is trained under two different training modes, namely a different-SNR training mode and a fixed-SNR training mode. In different SNR training modes, for a given SNR value range, training a PIC-DNN model under each SNR, and then evaluating error code performance of test data under the SNR by using the trained model; in the fixed SNR training mode, the training process is only executed once in a given SNR value interval, namely, the model is only selected to be trained under a certain SNR, and then the trained model is used for evaluating the error code performance of the test data under all the SNR values in the interval.
As shown in fig. 7, the number of receiving antennas N r And number of transmitting antennas N t When the modulation modes are 24, under the condition that the modulation mode is QPSK, the PIC-DNN algorithm is compared in the performance of different SNR training modes and the performance of the PIC-DNN algorithm under the fixed SNR training mode, and the SNR of the fixed SNR training mode is 15dB, and experimental results show that compared with the different SNR training modes, the performance of the PIC-DNN algorithm is lost but the PIC-DNN algorithm is lost in the fixed SNR training modeThe method only needs to train once, so that the training time of the model is obviously reduced, the implementation complexity is reduced, and the method still has obvious error code performance advantage compared with ZF and PIC algorithms, when the SNR (signal to noise ratio) is higher>At 15dB, PIC-DNN is superior to MMSE, and SER can reach 10 at SNR =24dB -4 . It can be seen that the PIC-DNN algorithm has better robustness on test data under different SNRs, and the fixed SNR training mode can reduce the implementation complexity and maintain the error code performance superior to that of the PIC.
The second embodiment of the present invention provides a deep learning based signal detection apparatus in a massive MIMO system, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor executes the computer program to implement the deep learning based signal detection method in the massive MIMO system as described in any one of the above.
It should be noted that the signal detection method based on deep learning in the large-scale MIMO system implemented by the apparatus is consistent with the above embodiments, and details are not repeated here.
The device can be a desktop computer, a notebook computer, a cloud server and other computing equipment. The apparatus may include, but is not limited to, a processor, a memory. Those skilled in the art will appreciate that more or fewer components may be included, or certain components may be combined, or different components may be included, such as input output devices, network access devices, etc.
The Processor may be a Central Processing Unit (CPU), or other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory may in some embodiments be an internal storage unit of the extraction device, such as a hard disk or a memory of the extraction device. The memory may also be an external storage device of the extracting apparatus in other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), and the like, which are provided on the extracting apparatus. Further, the memory may also include both an internal storage unit and an external storage device of the extraction apparatus. The memory is used for storing an operating system, application programs, a BootLoader (BootLoader), data, and other programs, such as program codes of the computer programs. The memory may also be used to temporarily store data that has been output or is to be output.
Claims (9)
1. A signal detection method based on deep learning in a large-scale MIMO system is characterized by comprising the following steps:
detecting the received signal by adopting a zero forcing detection method to obtain a preliminary estimation signal;
reconstructing an interference signal based on the preliminary estimated signal;
performing interference cancellation on the received signal by taking the interference signal as a reference;
and estimating the received signals after the interference cancellation by adopting a deep learning network to obtain accurate estimation signals of each transmitting antenna.
2. The method as claimed in claim 1, wherein the deep learning network comprises:
the number of nodes of the input layer is 2N r (ii) a Wherein, N r Is the number of receive antennas;
the number of nodes of the network output layer is M; wherein M is the modulation order of QAM;
7 hidden layers are arranged between the input layer and the output layer, and the number of nodes of the hidden layers is 128-64-64-32-32-64-64 in sequence.
3. The method as claimed in claim 1, wherein the detecting the received signal by the zero-forcing detection method comprises:
solving an objective function of the zero forcing detection method to obtain a solution of the objective function:
and calculating and obtaining a preliminary estimation signal based on the solution of the objective function.
4. The method as claimed in claim 3, wherein the objective function is:
wherein y is the received signal, x is the transmitted signal, | · | | computationally idle 2 Denotes a 2 norm, H is the channel matrix, H T Is the transpose of the channel matrix (.) -1 Denotes the inversion, u is the solution of the objective function, N t The number of transmit antennas.
5. The method as claimed in claim 3, wherein the step of calculating and obtaining the preliminary estimation signal comprises:
and mapping the solution of the objective function and points in a constellation set of the transmitted signals to obtain the preliminary estimation signal.
6. The method according to claim 5, wherein the mapping between the solution of the objective function and the point in the constellation set of the transmitted signal is implemented as follows:
wherein, the first and the second end of the pipe are connected with each other,
in order to be able to estimate the signal preliminarily,
representing the minimum euclidean distance to solve for the transmitted signal and the objective function.
7. The method as claimed in any of claims 2-5, wherein reconstructing the interference signal based on the preliminary estimation signal comprises:
and calculating the sum of products of the preliminary estimation signals of other transmitting antennas except the kth transmitting antenna and the corresponding channel vector to obtain the interference signal.
8. The method of claim 1, wherein the interference cancellation of the received signal comprises:
the interference cancellation of the receiving signals is realized according to the following calculation mode;
wherein y is the received signal, and y is the received signal,
for the interference signal, y k Indicating the received signal of the kth receiving antenna after the interference cancellation.
9. An apparatus for deep learning based signal detection in massive MIMO system, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for deep learning based signal detection in massive MIMO system as claimed in any one of claims 1-8 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211348083.4A CN115694571A (en) | 2022-10-31 | 2022-10-31 | Signal detection method and device based on deep learning in large-scale MIMO system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211348083.4A CN115694571A (en) | 2022-10-31 | 2022-10-31 | Signal detection method and device based on deep learning in large-scale MIMO system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115694571A true CN115694571A (en) | 2023-02-03 |
Family
ID=85045493
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211348083.4A Pending CN115694571A (en) | 2022-10-31 | 2022-10-31 | Signal detection method and device based on deep learning in large-scale MIMO system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115694571A (en) |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080137782A1 (en) * | 2006-12-06 | 2008-06-12 | Electronics And Telecommunications Research Institute | Signal detection method and receiving apparatus in mimo system |
| US20140140425A1 (en) * | 2012-11-19 | 2014-05-22 | Beihang University | Mimo communication system signal detection method |
| CN108429574A (en) * | 2018-01-24 | 2018-08-21 | 西安科技大学 | Extensive mimo system emitting antenna selecting method |
| CN110460359A (en) * | 2019-07-08 | 2019-11-15 | 南京邮电大学 | A kind of mimo system signal acceptance method neural network based |
| CN110474663A (en) * | 2019-08-08 | 2019-11-19 | 广州大学 | One kind is neural network based can iterative intelligent signal detecting method |
| CN111630787A (en) * | 2020-04-07 | 2020-09-04 | 东莞理工学院 | MIMO multi-antenna signal transmission and detection technology based on deep learning |
| CN113517941A (en) * | 2021-07-06 | 2021-10-19 | 西安电子科技大学广州研究院 | Simulation method and system for channel estimation and iterative detection of large-scale MIMO system |
| CN113938234A (en) * | 2021-12-03 | 2022-01-14 | 东南大学 | Low-complexity sparse large-scale MIMO detection method |
| CN114499601A (en) * | 2021-12-23 | 2022-05-13 | 西安电子科技大学 | Large-scale MIMO signal detection method based on deep learning |
| CN114759997A (en) * | 2022-04-08 | 2022-07-15 | 山东大学 | Dual-drive MIMO system signal detection method based on data model |
| US20220303159A1 (en) * | 2021-03-18 | 2022-09-22 | San Diego State University (SDSU) Foundation, dba San Diego State University Research Foundation | Systems and methods for optimization and estimation of nonlinear mimo systems with deep neural networks |
-
2022
- 2022-10-31 CN CN202211348083.4A patent/CN115694571A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080137782A1 (en) * | 2006-12-06 | 2008-06-12 | Electronics And Telecommunications Research Institute | Signal detection method and receiving apparatus in mimo system |
| US20140140425A1 (en) * | 2012-11-19 | 2014-05-22 | Beihang University | Mimo communication system signal detection method |
| CN108429574A (en) * | 2018-01-24 | 2018-08-21 | 西安科技大学 | Extensive mimo system emitting antenna selecting method |
| CN110460359A (en) * | 2019-07-08 | 2019-11-15 | 南京邮电大学 | A kind of mimo system signal acceptance method neural network based |
| CN110474663A (en) * | 2019-08-08 | 2019-11-19 | 广州大学 | One kind is neural network based can iterative intelligent signal detecting method |
| CN111630787A (en) * | 2020-04-07 | 2020-09-04 | 东莞理工学院 | MIMO multi-antenna signal transmission and detection technology based on deep learning |
| WO2021203242A1 (en) * | 2020-04-07 | 2021-10-14 | 东莞理工学院 | Deep learning-based mimo multi-antenna signal transmission and detection technologies |
| US20220303159A1 (en) * | 2021-03-18 | 2022-09-22 | San Diego State University (SDSU) Foundation, dba San Diego State University Research Foundation | Systems and methods for optimization and estimation of nonlinear mimo systems with deep neural networks |
| CN113517941A (en) * | 2021-07-06 | 2021-10-19 | 西安电子科技大学广州研究院 | Simulation method and system for channel estimation and iterative detection of large-scale MIMO system |
| CN113938234A (en) * | 2021-12-03 | 2022-01-14 | 东南大学 | Low-complexity sparse large-scale MIMO detection method |
| CN114499601A (en) * | 2021-12-23 | 2022-05-13 | 西安电子科技大学 | Large-scale MIMO signal detection method based on deep learning |
| CN114759997A (en) * | 2022-04-08 | 2022-07-15 | 山东大学 | Dual-drive MIMO system signal detection method based on data model |
Non-Patent Citations (9)
| Title |
|---|
| HENGTAO HE: "A Model-Driven Deep Learning Network for MIMO Detection", 《COMPUTER SCIENCE》, 25 September 2018 (2018-09-25) * |
| KANG XIAOFEI: "Deep Learning for Joint Pilot Design and Channel Estimation in MIMO-OFDM Systems", 《SENSORS》, 15 June 2022 (2022-06-15) * |
| LIANG-XIAN ZHANG;: "Optimization Arrangement of angle rings in Converter Transformer Valve Winding End Insulation Structure", 《2020 IEEE INTERNATIONAL CONFERENCE ON HIGH VOLTAGE ENGINEERING AND APPLICATION (ICHVE)》, 15 December 2020 (2020-12-15) * |
| TIMOTHY J.: "Deep Learning Based MIMO Communications", 《COMPUTER SCIENCE》, 25 July 2017 (2017-07-25) * |
| 康晓非: "MIMO-OFDM系统中一种改进的盲信道估计算法", 《应用科学学报》, 31 December 2016 (2016-12-31) * |
| 康晓非: "基于支持向量机的多用户检测算法研究", 《信息科技》, 15 September 2005 (2005-09-15) * |
| 康晓非: "超宽带系统中接收技术研究", 《信息科技》, 15 February 2013 (2013-02-15) * |
| 朱啸豪;姜述超;孙超;: "MIMO检测算法与深度学习方法探讨", 微型电脑应用, no. 01, 20 January 2020 (2020-01-20) * |
| 朱啸豪;孙超;姜述超;: "基于深度学习和AMP的MIMO检测算法", 微型电脑应用, no. 05, 20 May 2020 (2020-05-20) * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109921882B (en) | Deep learning-based MIMO decoding method, device and storage medium | |
| CN109995449B (en) | Millimeter wave signal detection method based on deep learning | |
| RU2303330C1 (en) | Method for receiving signal in communication system with several channels for transmitting and receiving | |
| CN109951214B (en) | Signal detection method suitable for large-scale MIMO system | |
| CN106982086B (en) | Spatial modulation method based on receiving and transmitting antenna selection | |
| CN108650056A (en) | A kind of mixed iteration detection method in extensive mimo system uplink | |
| CN108833060B (en) | Large-scale MIMO system detection method based on EP-NSA algorithm | |
| CN104301267A (en) | Multi-stage iterative detection method and device of MIMO wireless communication receiver | |
| CN116318311B (en) | Transmission method based on reconfigurable intelligent surface anti-phase index modulation | |
| CN101227254B (en) | Method for detecting V-BLAST in MIMO system | |
| CN101834693B (en) | Multiple-input multiple-output signal detection method, device and system | |
| CN102487309B (en) | Signal detecting method and device under MIMO (Multiple Input Multiple Output) system | |
| EP1609265B1 (en) | Signal processing apparatus and method | |
| CN109818891B (en) | Lattice reduction assisted low-complexity greedy sphere decoding detection method | |
| CN115694571A (en) | Signal detection method and device based on deep learning in large-scale MIMO system | |
| Kim | Performance of decremental antenna selection algorithms for spatial multiplexing MIMO systems with linear receiver over correlated fading channels | |
| Huang et al. | Deep learning based parallel detector for MIMO systems | |
| CN106899388B (en) | Joint-detection and coding/decoding method of the LDPC code under mimo channel | |
| CN109818663A (en) | A kind of low complex degree difference quadrature spatial modulation detection method | |
| CN114640561A (en) | Communication signal transmission method and device | |
| CN107248876B (en) | Generalized spatial modulation symbol detection method based on sparse Bayesian learning | |
| CN101924601A (en) | Method and device for realizing MIMO signal detection | |
| KR102441253B1 (en) | Apparatus and method for receiving signal in wireless communication system | |
| KR100939357B1 (en) | Apparatus for enhanced signal detection in wireless communication system with multiple antennas and method therof | |
| WO2021109768A1 (en) | Decoding result determining method and device, storage medium, and electronic device |
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 |