CN116319195A - Millimeter wave and terahertz channel estimation method based on pruned convolutional neural network - Google Patents

Millimeter wave and terahertz channel estimation method based on pruned convolutional neural network Download PDF

Info

Publication number
CN116319195A
CN116319195A CN202310352756.1A CN202310352756A CN116319195A CN 116319195 A CN116319195 A CN 116319195A CN 202310352756 A CN202310352756 A CN 202310352756A CN 116319195 A CN116319195 A CN 116319195A
Authority
CN
China
Prior art keywords
neural network
channel
convolutional neural
millimeter wave
matrix
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
Application number
CN202310352756.1A
Other languages
Chinese (zh)
Other versions
CN116319195B (en
Inventor
韩充
胡正东
陈宇航
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Jiaotong University
Original Assignee
Shanghai Jiaotong University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shanghai Jiaotong University filed Critical Shanghai Jiaotong University
Priority to CN202310352756.1A priority Critical patent/CN116319195B/en
Publication of CN116319195A publication Critical patent/CN116319195A/en
Application granted granted Critical
Publication of CN116319195B publication Critical patent/CN116319195B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L25/00Baseband systems
    • H04L25/02Details ; arrangements for supplying electrical power along data transmission lines
    • H04L25/0202Channel estimation
    • H04L25/024Channel estimation channel estimation algorithms
    • H04L25/0254Channel estimation channel estimation algorithms using neural network algorithms
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04BTRANSMISSION
    • H04B10/00Transmission systems employing electromagnetic waves other than radio-waves, e.g. infrared, visible or ultraviolet light, or employing corpuscular radiation, e.g. quantum communication
    • H04B10/90Non-optical transmission systems, e.g. transmission systems employing non-photonic corpuscular radiation
    • YGENERAL 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
    • Y02TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02DCLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00Reducing energy consumption in communication networks
    • Y02D30/70Reducing energy consumption in communication networks in wireless communication networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Power Engineering (AREA)
  • Optics & Photonics (AREA)
  • Electromagnetism (AREA)
  • Radio Transmission System (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

A millimeter wave and terahertz channel estimation method based on a pruned convolutional neural network utilizes an approximate message transfer compressed sensing Algorithm (AMP) to process channel data in an off-line stage to obtain initial channel prediction, and then inputs the initial channel prediction into a convolutional neural network (DCNN) for training; then, on the premise that the parameters of the network remain unchanged after pruning operation is carried out on the trained convolutional neural network, the initial channel prediction is input into the neural network, and meanwhile, the real channel is used as a label for retraining; and performing millimeter wave and terahertz channel estimation by adopting the trained convolutional neural network in an online stage. The invention deletes redundant connection in the convolutional neural network by using the pruning algorithm through the channel estimation algorithm based on the pruning convolutional neural network, greatly compresses the scale of the neural network, improves the efficiency of channel estimation, and ensures the high precision of channel estimation.

Description

Millimeter wave and terahertz channel estimation method based on pruned convolutional neural network
Technical Field
The invention relates to a technology in the field of wireless communication, in particular to a millimeter wave and terahertz channel estimation method based on a pruned convolutional neural network.
Background
The existing improved channel estimation method is realized based on a deep learning technology and comprises model driving and data driving. Model-driven methods construct deep neural networks from models, typically using neural networks to simulate the iterative process of conventional algorithms. The deep learning network only needs to learn parameters required by the iterative algorithm to realize quick and efficient channel estimation. However, such methods are limited by the performance of the conventional algorithms that are simulated. In contrast, the data-driven method trains a deep neural network for channel estimation from a large amount of channel transmission data, is not limited by the conventional algorithm, and can realize high channel estimation accuracy. However, while approaching the best performance of channel estimation, the size of neural networks is also increasing at a very rapid rate, resulting in very high computational complexity, inefficiency and inapplicability to practical deployment.
Disclosure of Invention
Aiming at the problems that the existing terahertz ultra-large-scale channel estimation method based on the deep convolutional neural network is very large in scale, redundant in structure and low in efficiency and unsuitable for practical deployment, the invention provides the millimeter wave and terahertz channel estimation method based on the pruned convolutional neural network.
The invention is realized by the following technical scheme:
the invention relates to a millimeter wave and terahertz channel estimation method based on a pruned convolutional neural network, which utilizes an approximate message transfer compressed sensing Algorithm (AMP) to process channel data in an off-line stage to obtain initial channel prediction, and inputs the initial channel prediction into the convolutional neural network (DCNN) for training; then, on the premise that the parameters of the network remain unchanged after pruning operation is carried out on the trained convolutional neural network, the initial channel prediction is input into the neural network, and meanwhile, the real channel is used as a label for retraining; and performing millimeter wave and terahertz channel estimation by adopting the trained convolutional neural network in an online stage.
Technical effects
The invention combines the approximate message transmission compressed sensing algorithm and the convolutional neural network, and simultaneously uses the pruning algorithm to delete redundant convolutional kernels and characteristic diagrams in the deep convolutional network.
Drawings
FIG. 1 is a schematic diagram of a very large scale Multiple Input Multiple Output (MIMO) system;
FIG. 2 is a schematic diagram of a process for estimating a channel matrix from an approximate message passing compressed sensing algorithm and a convolutional neural network;
FIG. 3 is a schematic diagram of a process for pruning and retraining convolutional neural networks;
FIG. 4 is a diagram showing the accuracy of the designed AMP-DCNN channel estimation method in comparison with the conventional method;
FIG. 5 is a diagram showing the comparison of the accuracy of channel estimation after pruning at different scales;
fig. 6 is a schematic diagram of the number of parameters of the neural network after pruning at different scales.
Detailed Description
The embodiment relates to a millimeter wave and terahertz ultra-large-scale channel estimation method based on a pruned convolutional neural network, which comprises the following steps:
step 1, a channel estimation problem model is built according to a very large scale MIMO system as shown in figure 1, and a pilot (pilot) signal x [ k ] is transmitted at a transmitting end]=F RF F BB s[k]Based on the pilot signal known at the receiving end, from the receiving end signal
Figure BDA0004162208360000028
Figure BDA0004162208360000021
Recovering the channel matrix H [ k ]]The method comprises the steps of carrying out a first treatment on the surface of the Further based on the pilot signal and antenna precoding and combining matrices are known,according to sparsity H [ k ] of channel]=A R h[k]A T Thereby, when transmitting a plurality of pilot signals, a plurality of received signal results are collected in the channel coherence time to obtain a received signal Y [ k ]]=φH[k]+N[k]Wherein: k is different sub-carriers, s [ k ]]F for transmitted data RF And F BB Analog and digital beamforming matrices for the precoding process, respectively; w (W) RF And W is BB Respectively analog and digital combining matrices, n [ k ]]Representing the received noise; h [ k ]]As a sparse matrix, A R And A T A coefficient matrix relating to the channel multipath reception angle and the transmission angle, respectively; the observation matrix phi includes: pilot signals and antenna precoding and combining matrix information.
Step 2, as shown in FIG. 2, the received signal Y [ k ] obtained according to step 1]And measuring the matrix phi, and obtaining a preliminary estimated channel by adopting an approximate message transfer compressed sensing algorithm
Figure BDA0004162208360000022
As training set, convolutional neural network is input and channel estimation output +.>
Figure BDA0004162208360000023
The approximate message passing compressed sensing algorithm specifically comprises the following steps:
2.1 setting an initial value: t=0, h 0 =0,r -1 =0,b 0 =0,c 0 =0, m, n is the number of rows and columns of the matrix Φ, λ is a parameter set in advance.
2.2 calculating residual terms:
Figure BDA0004162208360000024
wherein: b t r t-1 And->
Figure BDA0004162208360000025
Is an Oncorhynchi correction term (Onsager Correction term) for accelerating the convergence rate.
2.3 calculating a predicted value: variance of
Figure BDA0004162208360000026
Wherein: η (eta) st Will be less than the threshold λσ as a function of setting the threshold t The term of (2) is set to 0.
2.4 update parameters:
Figure BDA0004162208360000027
2.5 repeating the steps 2.2-2.4 ten times to obtain a predicted result h T Recovering the original matrix
Figure BDA0004162208360000031
The convolutional neural network comprises 12 convolutional layers, the number of the convolutional kernels is 64 except for the last convolutional layer which is 2, and the convolutional layers are connected with each other through an activation function and a standardization layer. The convolutional neural network is to be input
Figure BDA0004162208360000032
As a 2-dimensional picture, 2 dimensions represent the real part and the imaginary part of the data, respectively,/output +.>
Figure BDA0004162208360000033
And->
Figure BDA0004162208360000034
The dimensions are the same.
The convolution layers all contain ReLU activation functions to enhance the nonlinear expression capacity of the neural network.
The convolutional neural network adopts a loss function of
Figure BDA0004162208360000035
Wherein: MSE is mean square error, lambda is the regularized coefficient, gamma is the scale coefficient of the normalized layer.
The scale factor of the standardized layer measures the importance of the convolutional neural feature map connected with the factor to a certain extent. Taking gamma as regularization term, the value is reduced as much as possible in the training process, so that the proportionality coefficient corresponding to the redundant characteristic diagram is close to 0.
The index used for measuring the channel recovery precision in the training of the convolutional neural network is normalized mean square error
Figure BDA0004162208360000036
Wherein: h is the real channel matrix, < >>
Figure BDA0004162208360000037
For the channel matrix obtained by AMP-DCNN prediction method, the adopted Norm is French Luo Beini Usneius Norm (Frobenius Norm), and +.>
Figure BDA0004162208360000038
Figure BDA0004162208360000039
The convolutional neural network adopts an Adam optimizer during training, the learning rate is 0.001, lambda is 0.0001, 6000 data are contained in a data set, 10% is a test set, and 90% is a training set.
Step 3, as shown in fig. 2, pruning the convolutional neural network trained in step 2 to obtain a lightweight optimized channel estimation network, so as to significantly reduce required memory and operation resources, including: and determining the proportion p to be pruned and the proportion coefficient threshold value of the standardized layer, wherein in the pruning process, the proportion coefficient of the characteristic diagram connection of the deconvolution neural network is smaller than the proportion coefficient threshold value of the standardized layer.
The proportion to be pruned refers to: the number of scaling factor thresholds less than the normalization layer is scaled by p.
And 4, retraining the pruned convolutional neural network on the data set obtained in the step 2 by taking a real channel as a label to finally obtain a lightweight neural network (PRINCE), and further carrying out millimeter wave and terahertz channel estimation by adopting the trained convolutional neural network in an online stage.
Through specific experiments, at the communication frequency of 0.3THz and the communication distance of 20m, 512 antennas are arranged at the receiving and transmitting ends respectively, as shown in fig. 4, and compared with the traditional method, the channel estimation accuracy is improved by 8dB. As shown in fig. 5 and 6, compared with the channel estimation algorithm based on the neural network, the channel estimation method based on pruning can reduce the complexity by 80%, ensure the high precision of channel estimation, and the error rate is as low as.10 dB under the condition that the signal-to-noise ratio is 10dB.
The foregoing embodiments may be partially modified in numerous ways by those skilled in the art without departing from the principles and spirit of the invention, the scope of which is defined in the claims and not by the foregoing embodiments, and all such implementations are within the scope of the invention.

Claims (7)

1. The millimeter wave and terahertz channel estimation method based on the pruned convolutional neural network is characterized in that channel data are processed by using an approximate message transfer compressed sensing Algorithm (AMP) in an off-line stage to obtain initial channel prediction, and the initial channel prediction is input into a convolutional neural network (DCNN) for training; then, on the premise that the parameters of the network remain unchanged after pruning operation is carried out on the trained convolutional neural network, the initial channel prediction is input into the neural network, and meanwhile, the real channel is used as a label for retraining; and performing millimeter wave and terahertz channel estimation by adopting the trained convolutional neural network in an online stage.
2. The method for estimating millimeter wave and terahertz channel based on a pruned convolutional neural network according to claim 1, wherein the initial channel prediction is obtained by:
step 1, establishing a channel estimation problem model according to a super-large-scale MIMO system, and transmitting pilot frequency (pilot) signals x [ k ] at a transmitting end]=F RF F BB s[k]Based on the pilot signal known at the receiving end, from the receiving end signal
Figure FDA0004162208350000011
Figure FDA0004162208350000012
Recovering the channel matrix H [ k ]]The method comprises the steps of carrying out a first treatment on the surface of the Further based on pilot signal and antenna precoding and combining matrix, based on sparsity of channel H [ k ]]=A R h[k]A T Thereby, when transmitting a plurality of pilot signals, a plurality of received signal results are collected in the channel coherence time to obtain a received signal Y [ k ]]=φH[k]+N[k]Wherein: k is different sub-carriers, s [ k ]]F for transmitted data RF And F BB Analog and digital beamforming matrices for the precoding process, respectively; w (W) RF And W is BB Respectively analog and digital combining matrices, n [ k ]]Representing the received noise; h [ k ]]As a sparse matrix, A R And A T A coefficient matrix relating to the channel multipath reception angle and the transmission angle, respectively; the observation matrix phi includes: matrix information of pilot signal and antenna precoding and combining;
step 2, according to the received signal Y [ k ] obtained in step 1]And measuring the matrix phi, and obtaining a preliminary estimated channel by adopting an approximate message transfer compressed sensing algorithm
Figure FDA0004162208350000013
As a training set;
the approximate message passing compressed sensing algorithm specifically comprises the following steps:
2.1 setting an initial value: t=0, h 0 =0,r -1 =0,b 0 =0,c 0 =0, m, n is the number of rows and columns of the matrix Φ, λ is a parameter set in advance;
2.2 calculating residual terms:
Figure FDA0004162208350000014
wherein: b t r t-1 And->
Figure FDA0004162208350000015
The correction term is an Oncorhynchi correction term and is used for accelerating the convergence rate;
2.3 calculating a predicted value: variance of
Figure FDA0004162208350000016
Wherein: η (eta) st Will be less than the threshold λσ as a function of setting the threshold t The term of (2) is set to 0;
2.4 update parameters:
Figure FDA0004162208350000017
2.5 repeating the steps 2.2-2.4 ten times to obtain a predicted result h T Recovering the original matrix
Figure FDA0004162208350000018
3. The method for estimating millimeter wave and terahertz channel based on a pruned convolutional neural network as claimed in claim 1, wherein the convolutional neural network includes 12 convolutional layers, the number of convolutional kernels is 64 except for the last convolutional layer which is 2, and the convolutional layers are connected by an activation function and a normalization layer, and the convolutional neural network connects the input convolutional layers
Figure FDA0004162208350000021
As a 2-dimensional picture, 2 dimensions represent the real part and the imaginary part of the data, respectively,/output +.>
Figure FDA0004162208350000022
And->
Figure FDA0004162208350000023
The dimensions are the same.
4. The method for estimating millimeter wave and terahertz channel based on a pruned convolutional neural network according to claim 1 or 3, wherein the convolutional layers each contain a ReLU activation function to enhance the nonlinear expression capability of the neural network;
the convolutional neural network adopts a loss function of
Figure FDA0004162208350000024
Wherein: MSE is mean square error, lambda is the regularized coefficient, gamma is the scale coefficient of the normalized layer.
5. The method for estimating millimeter wave and terahertz channel based on a pruned convolutional neural network as set forth in claim 4, wherein the convolutional neural network is trained by using normalized mean square error as an index for measuring channel recovery accuracy
Figure FDA0004162208350000025
Wherein: h is the real channel matrix, < >>
Figure FDA0004162208350000026
For the channel matrix obtained by AMP-DCNN prediction method, the adopted Norm is French Luo Beini Usneius Norm (Frobenius Norm), and +.>
Figure FDA0004162208350000027
Figure FDA0004162208350000028
6. The method for estimating millimeter wave and terahertz channel based on a pruned convolutional neural network according to claim 1, wherein the pruning operation is: pruning operation is carried out on the trained convolutional neural network to obtain a lightweight optimized channel estimation network so as to remarkably reduce required memory and operation resources, and the method specifically comprises the following steps: and determining the proportion p to be pruned and the proportion coefficient threshold value of the standardized layer, wherein in the pruning process, the proportion coefficient of the characteristic diagram connection of the deconvolution neural network is smaller than the proportion coefficient threshold value of the standardized layer.
7. The method for estimating millimeter wave and terahertz channel based on a pruned convolutional neural network according to claim 1, wherein the retraining means: and retraining the pruned convolutional neural network by taking a real channel as a label on initial channel prediction, and finally obtaining the lightweight neural network.
CN202310352756.1A 2023-04-04 2023-04-04 Millimeter wave and terahertz channel estimation method based on pruned convolutional neural network Active CN116319195B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310352756.1A CN116319195B (en) 2023-04-04 2023-04-04 Millimeter wave and terahertz channel estimation method based on pruned convolutional neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310352756.1A CN116319195B (en) 2023-04-04 2023-04-04 Millimeter wave and terahertz channel estimation method based on pruned convolutional neural network

Publications (2)

Publication Number Publication Date
CN116319195A true CN116319195A (en) 2023-06-23
CN116319195B CN116319195B (en) 2023-10-20

Family

ID=86788709

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310352756.1A Active CN116319195B (en) 2023-04-04 2023-04-04 Millimeter wave and terahertz channel estimation method based on pruned convolutional neural network

Country Status (1)

Country Link
CN (1) CN116319195B (en)

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190340497A1 (en) * 2016-12-09 2019-11-07 William Marsh Rice University Signal Recovery Via Deep Convolutional Networks
CN111030952A (en) * 2019-12-25 2020-04-17 内蒙古大学 Beam space channel estimation method and system of millimeter wave system
CN111224906A (en) * 2020-02-21 2020-06-02 重庆邮电大学 Approximate message transfer large-scale MIMO signal detection algorithm based on deep neural network
WO2020220278A1 (en) * 2019-04-30 2020-11-05 华为技术有限公司 Channel estimation model training method and device
CN112910806A (en) * 2021-01-19 2021-06-04 北京理工大学 Joint channel estimation and user activation detection method based on deep neural network
CN112968853A (en) * 2021-01-29 2021-06-15 北京交通大学 Multi-user access detection and channel estimation method for MMTC (multimedia messaging and control) system
CN113271270A (en) * 2021-06-28 2021-08-17 上海交通大学 Terahertz ultra-large scale channel estimation method based on deep convolutional neural network
CN115022134A (en) * 2022-06-28 2022-09-06 河南工业大学 Millimeter wave large-scale MIMO system channel estimation method and system based on non-iterative reconstruction network
CN115664898A (en) * 2022-10-24 2023-01-31 四川农业大学 OFDM system channel estimation method and system based on complex convolution neural network
US20230073228A1 (en) * 2020-05-15 2023-03-09 Vivo Mobile Communication Co., Ltd. Pilot information system sending method, channel estimation method, and communications device

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190340497A1 (en) * 2016-12-09 2019-11-07 William Marsh Rice University Signal Recovery Via Deep Convolutional Networks
WO2020220278A1 (en) * 2019-04-30 2020-11-05 华为技术有限公司 Channel estimation model training method and device
CN111030952A (en) * 2019-12-25 2020-04-17 内蒙古大学 Beam space channel estimation method and system of millimeter wave system
CN111224906A (en) * 2020-02-21 2020-06-02 重庆邮电大学 Approximate message transfer large-scale MIMO signal detection algorithm based on deep neural network
US20230073228A1 (en) * 2020-05-15 2023-03-09 Vivo Mobile Communication Co., Ltd. Pilot information system sending method, channel estimation method, and communications device
CN112910806A (en) * 2021-01-19 2021-06-04 北京理工大学 Joint channel estimation and user activation detection method based on deep neural network
CN112968853A (en) * 2021-01-29 2021-06-15 北京交通大学 Multi-user access detection and channel estimation method for MMTC (multimedia messaging and control) system
CN113271270A (en) * 2021-06-28 2021-08-17 上海交通大学 Terahertz ultra-large scale channel estimation method based on deep convolutional neural network
CN115022134A (en) * 2022-06-28 2022-09-06 河南工业大学 Millimeter wave large-scale MIMO system channel estimation method and system based on non-iterative reconstruction network
CN115664898A (en) * 2022-10-24 2023-01-31 四川农业大学 OFDM system channel estimation method and system based on complex convolution neural network

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
ZHENGDONG HU: "PRINCE: A Pruned AMP Integrated Deep CNN Method for Efficient Channel Estimation of Millimeter-wave and Terahertz Ultra-Massive MIMO Systems", 《IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS ( EARLY ACCESS )》, pages 1 - 6 *
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 *
邱佳锋;: "基于优化BM3D的毫米波大规模MIMO信道估计", 电信科学, no. 09 *

Also Published As

Publication number Publication date
CN116319195B (en) 2023-10-20

Similar Documents

Publication Publication Date Title
CN111030952B (en) Beam space channel estimation method and system of millimeter wave system
CN110648292B (en) High-noise image denoising method based on deep convolutional network
CN110336594B (en) Deep learning signal detection method based on conjugate gradient descent method
CN109525369B (en) Channel coding type blind identification method based on recurrent neural network
CN111160176A (en) Fusion feature-based ground radar target classification method for one-dimensional convolutional neural network
CN112600618B (en) Attention mechanism-based visible light signal equalization system and method
CN114157539B (en) Data-aware dual-drive modulation intelligent identification method
CN113014524B (en) Digital signal modulation identification method based on deep learning
CN114912486A (en) Modulation mode intelligent identification method based on lightweight network
CN110808932B (en) Multi-layer sensor rapid modulation identification method based on multi-distribution test data fusion
CN115982613A (en) Signal modulation identification system and method based on improved convolutional neural network
CN116319195B (en) Millimeter wave and terahertz channel estimation method based on pruned convolutional neural network
CN113938254A (en) Attention mechanism-based layered source-channel joint coding transmission system and transmission method thereof
CN114614920B (en) Signal detection method based on data and model combined driving of learning factor graph
CN116760491A (en) Signal-to-noise ratio estimation method based on deep learning
CN114422310B (en) Digital quadrature modulation signal identification method based on joint distribution matrix and multi-input neural network
CN115329821A (en) Ship noise identification method based on pairing coding network and comparison learning
CN114337883A (en) CNN cooperative spectrum sensing method and system based on covariance matrix Cholesky decomposition
CN114584441A (en) Digital signal modulation identification method based on deep learning
Zhao et al. An End-to-End Demodulation System Based on Convolutional Neural Networks
CN111985327A (en) Signal deep learning classification method based on sliding trainable operator
CN116367305B (en) Wireless positioning method and system based on mask self-encoder and signal guidance
CN115499278B (en) MIMO signal modulation identification method based on lightweight neural network
CN114157544B (en) Frame synchronization method, device and medium based on convolutional neural network
CN114301499B (en) Intelligent frequency hopping signal receiving method based on CNN-GRU

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