CN114362795B - Signal detection method of nonlinear millimeter wave MIMO communication system - Google Patents
Signal detection method of nonlinear millimeter wave MIMO communication system Download PDFInfo
- Publication number
- CN114362795B CN114362795B CN202111394394.XA CN202111394394A CN114362795B CN 114362795 B CN114362795 B CN 114362795B CN 202111394394 A CN202111394394 A CN 202111394394A CN 114362795 B CN114362795 B CN 114362795B
- Authority
- CN
- China
- Prior art keywords
- signal
- real
- nonlinear
- neural network
- signal vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
Abstract
The invention relates to the technical field of communication, in particular to a signal detection method of a nonlinear millimeter wave MIMO communication system. Compared with the prior art, the deep neural network of the method has a simpler structure, greatly reduces the number of hyper-parameters, reduces overfitting, reduces complexity, is easy to train, and can achieve better detection performance by only needing a small amount of training data.
Description
Technical Field
The invention relates to the technical field of communication, in particular to a signal detection method of a nonlinear millimeter wave MIMO communication system.
Background
MIMO communication, i.e., multiple-input multiple-output communication, is a communication method for transmitting and receiving information using a plurality of transmitting and receiving antennas. MIMO communication is a promising technology for millimeter wave band communication, because it can meet the high transmission rate requirements in the scenes of 5G wireless communication, even B5G, etc. Millimeter waves generally require high power transmission because of their physical properties, which tend to be attenuated. However, millimeter wave high power amplifiers have non-linear characteristics that are difficult to avoid. In order to achieve high-rate transmission, it is often required to use a high-order modulation method, which is however very sensitive to nonlinear distortion, resulting in a severe impact on communication performance. It is necessary to address the combined effects of non-linear distortion and inter-channel interference in MIMO communications.
One of the existing methods to solve this problem is to linearize the non-linear distortion of the power amplifier using a Volterra series or to compensate for the non-linearity of the power amplifier. The memory polynomial is a simplified Volterra series method, can reduce part of complexity to compensate nonlinear distortion, but is easy to fall into the problem of data instability due to the fact that the polynomial coefficient is solved, and therefore the neural network is an optional scheme for simulating and removing the power amplifier nonlinear distortion. However, the existing deep neural network-based methods have complex structures, redundant over-parameter amounts, and require excessive training data.
Disclosure of Invention
Aiming at the problems in the prior art, the invention aims to provide a signal detection method of a nonlinear millimeter wave MIMO communication system, wherein a deep neural network has a simpler structure, the number of hyper-parameters is greatly reduced, overfitting is reduced, complexity is reduced, training is easy, and good detection performance can be achieved only by a small amount of training data.
In order to achieve the purpose, the invention is realized by adopting the following technical scheme.
A signal detection method of a nonlinear millimeter wave MIMO communication system comprises the following steps:
step 4, inputting the training data into the deep neural network, and updating the hyper-parameters of the deep neural network; inputting test data into the trained deep neural network, and verifying the performance of the trained deep neural network; finally obtaining a deep neural network after the hyper-parameters are solidified;
and 5, detecting the real-valued received signal vector y' at a receiving end by using the deep neural network after the hyper-parameters are solidified, removing the nonlinear distortion of the signal and the interference between channels, and finishing the signal detection.
Compared with the prior art, the invention has the following beneficial effects: the deep neural network of the method has a simpler structure, greatly reduces the number of hyper-parameters, reduces overfitting, reduces complexity, is easy to train, and can achieve better detection performance by only needing a small amount of training data.
Drawings
The invention is described in further detail below with reference to the figures and specific embodiments.
FIG. 1 is a schematic flow chart illustrating a signal detection method of a nonlinear millimeter wave MIMO communication system according to the present invention;
FIG. 2 is a schematic diagram of a deep neural network model of a signal detection method of the nonlinear millimeter wave MIMO communication system according to the present invention;
FIG. 3 is a graph of bit error rate versus signal-to-noise ratio for the 5 methods of simulation test 1;
fig. 4 is a graph of the relationship between the bit error rate and the amount of training data for the 4 methods of simulation test 2.
Detailed Description
Embodiments of the present invention will be described in detail below with reference to examples, but it will be understood by those skilled in the art that the following examples are only illustrative of the present invention and should not be construed as limiting the scope of the present invention.
Referring to fig. 1, a schematic flow chart of a signal detection method of a nonlinear millimeter wave MIMO communication system according to the present invention is shown, where the signal detection method of the nonlinear millimeter wave MIMO communication system includes the following steps:
specifically, the modulated digital signal is subjected to power amplifier nonlinear distortion through a power amplifier, and the distorted signal intensity a and phase Φ are respectively represented as:
wherein x represents the transmitted complex signal; alpha is alpha a Is the signal strength gain; alpha is alpha φ And beta φ Is the phase gain, σ a Is the signal strength nonlinear order; x is the number of sat To suppress strength; q. q.s 1 And q is 2 Is the phase non-linear order;
the final signal after nonlinear distortion is expressed as:
s=f(x)=A(|x|)e j(angle(x)+Φ(|x|))
wherein f (x) represents a non-linear function; a (| x |) represents the signal intensity of the signal x after nonlinear intensity transformation; phi (| x |) represents the phase of x after nonlinear phase transformation; angle (x) represents the phase of x;
the signal after nonlinear distortion is transmitted by using K transmitting antennas at the same time and is expressed as a transmitting signal vector s; the transmitted signal vector s passes through the millimeter wave channel, and is received by N receiving antennas at the receiving end, and the m-th received signal vector may be represented as:
y(m)=Hs(m)+n(m)=Hf(x(m))+n(m)
wherein H represents a millimeter wave channel matrix; n represents a noise vector;
respectively regarding the real part and the imaginary part of the received signal as two mutually independent real part characteristics, and performing two-dimension on the signal, wherein the specific formula is as follows:
wherein the content of the first and second substances,representing the operation of the real part; />Representing an imaginary part taking operation; y' represents the received signal vector after real number processing; h' represents a channel matrix after real number processing; s' represents a signal vector after being subjected to nonlinear distortion after being subjected to real number; n' represents the noise vector after real number;
the operation of step 1 is performed on the plurality of modulated digital signals to obtain a plurality of real-valued received signal vectors y'.
step 4, inputting the training data into the deep neural network, and updating the hyper-parameters of the deep neural network; inputting test data into the trained deep neural network, and verifying the performance of the trained deep neural network; finally obtaining a deep neural network after the hyper-parameters are solidified;
a deep neural network, specifically, referring to fig. 2, the deep neural network includes an input layer, a first hidden layer, a second hidden layer and an output layer;
between the input layer and the first hidden layer is a fully connected network with a special structure, and the weight of the fully connected network is expressed as follows:
wherein, W 11 And W 12 Is a matrix of size K × N; k is the number of users, and N is the number of receiving antennas;
the activation function of the first hidden layer is linear, each output unit is connected with a single hidden layer neural sub-network, and the hidden layer of the sub-network is the second hidden layer; finally forming a partially connected network;
the output of the second hidden layer is represented as:
wherein, g 2 () Is the activation function of the second hidden layer, which is non-linear; w is a 2,Re And w 2,Im Is the weight of the second hidden layer;is the output of the ith neuron of the first hidden layer; b 2,Re And b 2,Im Is an offset; m is a time index;
the ith cell output of the output layer is expressed as:
wherein w 3,Re And w 3,Im Is the output layer weight;
for K neuron output signals of 1 st to Kth hidden layers, the K connected neural sub-networks have the same weight and bias, and the weights w are the input layer weights of the neural sub-networks 2,Re And bias b 2,Re Output layer weights w of neural subnetwork 3,Re ;
For K neuron output signals of the K +1 th to the 2K th hidden layers, the K connected neuron sub-networks have the same weight and bias respectivelyInput layer weights w for neural sub-networks 2,Im And bias b 2,Im Output layer weights w of neural sub-networks 3,Im ;
The loss function of the deep neural network is:
wherein the content of the first and second substances,E k (m)=(e k (m)) 2 ,/>e k (m) is the prediction signal of the kth transmitting antenna->And true signal t' k An error of (2); m 0 The number of training data;
randomly initializing hyper-parameters of the deep neural network, including the weight and bias of connection between layers, inputting training data into the deep neural network, and updating the hyper-parameters of the deep neural network by utilizing gradient descent and back propagation minimum loss functions; inputting test data into the trained deep neural network, and verifying the performance of the trained deep neural network; and finally, curing the hyper-parameters of the deep neural network to obtain the deep neural network after curing the hyper-parameters.
And 5, detecting the real-valued received signal vector y' at a receiving end by using the deep neural network after the hyper-parameters are solidified, removing the nonlinear distortion of the signal and the interference between channels, and finishing the signal detection.
Simulation test
The test model is a multi-sending and multi-receiving millimeter wave communication system model with nonlinearity. The Saleh-Valenzuela model is used to simulate the millimeter wave channel, and the channel vector between the specific kth transmitting antenna and the N receiving antennas can be expressed as:
wherein, theta kq Is the angle of incidence for the q-th path,is a steering vector with an antenna spacing d, λ is the carrier wavelength, Q k For the number of paths of the Kth transmitting antenna, beta kq The gain of the q-th path, N, is the number of receive antennas. The nonlinear model parameter is set to alpha φ =2560,β φ =0.114,σ a =0.81,x sat =0.58,q 1 =2.4,q 2 =2.3。
And (4) performing hyper-parameter initialization and batch gradient descent iterative update on the deep neural network by using a Tensorflow tool. The batch size is 50, the number of epochs is 300, the learning rate is 0.01, and the number of hidden layer neurons of the single hidden layer subnet is 30. 20% of the data set was used as training data and 80% as test data.
Test 1: training data was set to 1000, and 5 different methods were used at different signal-to-noise ratios:
The simulation result of experiment 1 is shown in fig. 3, and it can be seen that under different signal-to-noise ratios, the method of the present invention has the lowest bit error rate, which proves the correctness and effectiveness of the present invention under the condition of existence of nonlinear distortion and inter-channel interference.
Test 2: the SNR was set to 20dB, and under the condition of different amounts of training data, the remaining settings were consistent with those of experiment 1, using methods 1, 3, 4 and 5 described in experiment 1.
The simulation result of experiment 2 is shown in fig. 4, and it can be seen that the method of the present invention can achieve the bit error rate effect of the method of the present invention with the shared weight cancelled in 2000 training data only with about 1000 training data, and both methods are far superior to the zero forcing algorithm of the known channel and the traditional deep neural network algorithm; the method of the invention can achieve better effect by using less training data and can effectively reduce communication overhead.
Although the invention has been described in detail in this specification with reference to specific embodiments and examples, it will be apparent to those skilled in the art that certain changes and modifications can be made thereto without departing from the scope of the invention. Accordingly, it is intended that all such modifications and alterations be included within the scope of this invention as defined in the appended claims.
Claims (4)
1. A signal detection method of a nonlinear millimeter wave MIMO communication system is characterized by comprising the following steps:
step 1, transmitting modulated digital signals through a power amplifier and K transmitting antennas simultaneously to obtain a transmitting signal vector s; the transmitting signal vector s passes through a millimeter wave channel to obtain a real number receiving signal vector y'; processing the plurality of modulated digital signals to obtain a plurality of real-numbered received signal vectors y';
step 2, acquiring a real-number input signal vector x ' corresponding to the real-number received signal vector y ', and pairing the real-number received signal vector y ' with the real-number input signal vector x ' corresponding to the real-number received signal vector y ', so as to generate a data set with a pairing number of M;
step 3, randomly dividing the data in the data set into training data and testing data;
step 4, inputting the training data into the deep neural network, and updating the hyper-parameters of the deep neural network; inputting test data into the trained deep neural network, and verifying the performance of the trained deep neural network; finally obtaining a deep neural network after the hyper-parameters are solidified;
the deep neural network in step 4, specifically, the deep neural network includes an input layer, a first hidden layer, a second hidden layer and an output layer;
between the input layer and the first hidden layer is a fully connected network with a special structure, and the weight of the fully connected network is expressed as follows:
wherein, W 11 And W 12 Is a matrix of size K × N; k is the number of users, and N is the number of receiving antennas; the activation function of the first hidden layer is linear, each output unit is connected with a single hidden layer neural sub-network, and the hidden layer of the sub-network is the second hidden layer; finally forming a partially connected network;
the output of the second hidden layer is represented as:
wherein, g 2 () Is an activation function of the second hidden layer, which function is non-linear; w is a 2,Re And w 2,Im Is the weight of the second hidden layer;is the output of the ith neuron of the first hidden layer; b 2,Re And b 2,Im Is an offset; m is a time index;
the ith cell output of the output layer is expressed as:
wherein w 3,Re And w 3,Im Is the output layer weight;
for K neuron output signals of 1 st to Kth hidden layers, the K connected neural sub-networks have the same weight and bias, and the weights w are the input layer weights of the neural sub-networks 2,Re And bias b 2,Re Output layer weights w of neural subnetwork 3,Re ;
For K neuron output signals of K +1 th to 2K th hidden layers, the K connected neural sub-networks have the same weight and bias, and the weights are input layer weights w of the neural sub-networks 2,Im And bias b 2,Im Output layer weights w of neural subnetwork 3,Im ;
And 5, detecting the real-valued received signal vector y' at a receiving end by using the deep neural network with the cured hyper-parameters, removing the nonlinear distortion of the signal and the interference between channels, and finishing the signal detection.
2. The signal detection method of the nonlinear millimeter wave MIMO communication system according to claim 1, wherein the loss function of the deep neural network is:
3. The signal detection method of the nonlinear millimeter wave MIMO communication system according to claim 1, wherein a signal vector s is transmitted, specifically, the modulated digital signal is subjected to power amplifier nonlinear distortion by a power amplifier, and the distorted signal intensity a and phase Φ are respectively represented as:
wherein x represents the transmitted complex signal; alpha is alpha a Is the signal strength gain; alpha (alpha) ("alpha") φ And beta φ Is the phase gain, σ a Is the signal strength nonlinear order; x is the number of sat To inhibit strength; q. q.s 1 And q is 2 Is the phase non-linear order;
the final signal after nonlinear distortion is expressed as:
s=f(x)=A(|x|)e j(angle(x)+Φ(|x|))
wherein f (x) represents a non-linear function; a (| x |) represents the signal intensity of the signal x after nonlinear intensity transformation; phi (| x |) represents the phase of x after nonlinear phase transformation; angle (x) represents the phase of x;
the signal after nonlinear distortion is transmitted simultaneously using K transmit antennas, denoted as a transmit signal vector s.
4. The signal detection method of the nonlinear millimeter wave MIMO communication system according to claim 1, wherein the received signal vector y' after being subjected to the real-valued processing, specifically, the transmitted signal vector s is received by N receiving antennas at a receiving end through the millimeter wave channel, and the received signal vector at the m-th time can be represented as:
y(m)=Hs(m)+n(m)=Hf(x(m))+n(m)
wherein H represents a millimeter wave channel matrix; n represents a noise vector;
respectively regarding a real part and an imaginary part of a received signal as two mutually independent real part characteristics, and performing two-dimension on the received signal, wherein a specific formula is as follows:
wherein, the first and the second end of the pipe are connected with each other,representing the operation of the real part; />Representing the operation of taking an imaginary part; y' represents the received signal vector after real number processing; h' represents a channel matrix after real number processing; s' represents a signal vector after being subjected to nonlinear distortion after being subjected to real number; n' represents the noise vector after real number transformation. />
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111394394.XA CN114362795B (en) | 2021-11-23 | 2021-11-23 | Signal detection method of nonlinear millimeter wave MIMO communication system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111394394.XA CN114362795B (en) | 2021-11-23 | 2021-11-23 | Signal detection method of nonlinear millimeter wave MIMO communication system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114362795A CN114362795A (en) | 2022-04-15 |
CN114362795B true CN114362795B (en) | 2023-04-18 |
Family
ID=81096129
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111394394.XA Active CN114362795B (en) | 2021-11-23 | 2021-11-23 | Signal detection method of nonlinear millimeter wave MIMO communication system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114362795B (en) |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109995449A (en) * | 2019-03-15 | 2019-07-09 | 北京邮电大学 | A kind of millimeter-wave signal detection method based on deep learning |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20190273509A1 (en) * | 2018-03-01 | 2019-09-05 | Crowdstrike, Inc. | Classification of source data by neural network processing |
WO2021029891A1 (en) * | 2019-08-14 | 2021-02-18 | Google Llc | Communicating a neural network formation configuration |
CN111769975A (en) * | 2020-06-11 | 2020-10-13 | 南方科技大学 | MIMO system signal detection method and system |
CN111769862B (en) * | 2020-06-28 | 2022-11-11 | 南京理工大学 | Joint detection method in spatial modulation network based on deep learning |
CN113452420B (en) * | 2021-06-25 | 2022-12-20 | 东南大学 | MIMO signal detection method based on deep neural network |
-
2021
- 2021-11-23 CN CN202111394394.XA patent/CN114362795B/en active Active
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109995449A (en) * | 2019-03-15 | 2019-07-09 | 北京邮电大学 | A kind of millimeter-wave signal detection method based on deep learning |
Also Published As
Publication number | Publication date |
---|---|
CN114362795A (en) | 2022-04-15 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110336594B (en) | Deep learning signal detection method based on conjugate gradient descent method | |
CN109995449B (en) | Millimeter wave signal detection method based on deep learning | |
Wan et al. | Deep learning-aided off-grid channel estimation for millimeter wave cellular systems | |
CN112217545B (en) | Large-scale MIMO system detection model construction method | |
Ye et al. | Channel estimation for intelligent reflecting surface aided wireless communications using conditional GAN | |
Ma et al. | Antenna array diagnosis for millimeter-wave MIMO systems | |
CN114362795B (en) | Signal detection method of nonlinear millimeter wave MIMO communication system | |
CN107276657B (en) | Millimeter wave mixed beam forming method based on parallel algorithm | |
US11777585B2 (en) | Wireless receiving apparatus and method thereof | |
CN113612571B (en) | Multi-component safe transmission method based on generalized multi-fraction Fourier transform | |
Seyman | Symbol Detection Based on Back Tracking Search Algorithm in MIMO-NOMA Systems. | |
CN115913844A (en) | MIMO system digital predistortion compensation method, device, equipment and storage medium based on neural network | |
Senthil et al. | Improve Multidimensional 5G OFDM Based MIMO Sushisen Algorithms Merge Multi-Cell Transmission | |
Chen et al. | Particle swarm optimization assisted B-spline neural network based predistorter design to enable transmit precoding for nonlinear MIMO downlink | |
CN101924718B (en) | Hybrid wavelet neural network blind equalization method controlled by fuzzy neutral network | |
Ramezanpour et al. | DNN‐based interference mitigation beamformer | |
Cherif et al. | Autoencoder‐based deep learning for massive multiple‐input multiple‐output uplink under high‐power amplifier non‐linearities | |
CN114337745A (en) | Large-scale MIMO signal detection method for model-driven deep learning | |
Liang et al. | A novel RTRLNN model for passive intermodulation cancellation in satellite communications | |
Shukla et al. | Variable Step-size Zero Attractor LMS based Channel Estimator for Millimeter Wave Hybrid MIMO System with Hardware Impairments | |
CN114499601B (en) | Large-scale MIMO signal detection method based on deep learning | |
MK et al. | Beamforming and combining for multi‐user large MIMO communication system | |
CN112202510B (en) | Signal detection method based on regularization stream neural network | |
CN114584236B (en) | RIGS algorithm-based large-scale MIMO system detection model construction method | |
Saxena et al. | MIMO semi-blind channel estimation by using Tikhonov regularized MMSE and MAP algorithms with Householder transformation based QR decomposition |
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 |