An antenna signal modulation method based on a neural network model

Abstract

The application relates to an antenna signal modulation method based on a neural network model, characterized in that the method at least comprises the following steps: creating a neural network through a BP neural network training method in artificial intelligence, constructing a first set of training sample sets with a sending signal vector x of a device transmitting end (1) and a sparse signal theta determined based on the sending signal vector x of the device transmitting end (1) as input and output respectively, training the neural network, and obtaining the first neural network model; performing compression dimension reduction processing on the signal after sparse representation; solving the sparse signal vector from the low-dimensional signal by using a compression sensing reconstruction algorithm; and finally, reconstructing the original signal by using a second neural network model on the received signal. The application reduces the dimension of the signal to be processed, multiplexes the parallel data stream exceeding the number of sending antennas onto a given sending antenna for sending, and greatly improves the signal multiplexing gain under the condition of a given MIMO system transceiving antenna number.

Description

[0001] The original basis for this divisional application is the patent application with application number (202080000961.6), application date of April 7, 2020, entitled "MIMO multi-antenna signal transmission and detection technology based on artificial intelligence". Technical Field

[0002] This invention relates to the field of mobile communication technology, and in particular to an antenna signal modulation method based on a neural network model. Background Technology

[0003] The research and development of communication technology can be considered a common asset of all mankind, but the global communication standards it brings are not just technical standards, but are related to industrial development and national strategies.

[0004] MIMO (Multiple-Input Multiple-Output) refers to a communication system that uses multiple antennas simultaneously at both the transmitting and receiving ends. This allows for a significant increase in system capacity and spectral efficiency without increasing bandwidth. In MIMO systems, multiple antennas are used at both the transmitting and receiving ends. The transmitted information stream is coded in space-time to form multiple information substreams, which are simultaneously transmitted to the channel. Each transmitted signal occupies the same frequency band, thus not increasing the system bandwidth. If the channel responses of each transmitting and receiving antenna are independent, the MIMO system can create multiple parallel spatial channels. Information is transmitted independently through these parallel spatial channels, inevitably increasing the data rate.

[0005] MIMO technology represents a significant breakthrough in wireless communication. In 2011, several companies developed commercial Wi-Fi or WiMAX systems based on MIMO. In 2012, all 4G communication system standards (such as TD-LTE, LTE-A, and WiMAX) adopted MIMO as one of their key technologies. MIMO systems are now widely used in wireless communication—mobile devices and networks commonly use multiple antennas to enhance connectivity, improve network speed, and enhance user experience. Massive MIMO is also a key factor in 5G's ultra-high data rates, enabling greater network capacity, wider signal coverage, and a better user experience, pushing the potential of 5G to a whole new level.

[0006] Methodologically, the upgrade from 4G to 5G goals is enormous. 4G focused on connecting people and emphasizing high transmission speeds, while 5G transcends these limitations to consider the interconnection of people and things, emphasizing not only higher transmission speeds but also massive connectivity and ultra-low latency. Sociologically, 4G changed our lives, while 5G is changing our social structure. To achieve 5G's three key performance indicators (KPIs)—ultra-high bandwidth, massive connectivity, and ultra-low latency—the existing network architecture needs to be upgraded and intelligently adjusted according to demand. Therefore, Artificial Intelligence (AI) can empower 5G. Conversely, 5G technology can generate more application demands, which in turn accelerates the development of AI.

[0007] To date, there is still no clear and unified definition of artificial intelligence (AI). One textbook definition is "AI is a computer program that, based on its perception of the environment, makes reasonable actions and obtains maximum benefits." Another, more technology-oriented definition is "AI is a computer algorithm that can learn." From the perspective of technological application, especially in the development of communication technology, academia tends to favor the second definition, which technically equates AI with machine learning. Machine learning is essentially a key technology for extracting knowledge from data and is the driving force and engine of AI development. Currently, machine learning mainly solves classification, clustering, and regression problems and has been widely applied in character recognition, machine translation, speech recognition, search engines, facial recognition, and autonomous driving. There are currently hundreds of learning algorithms of varying sizes, and there is no unified framework to describe the design process of machine learning algorithms. However, there are currently two most powerful machine learning algorithms—deep learning and reinforcement learning.

[0008] Deep learning belongs to the artificial neural network system and is a new generation of neural networks developed from traditional neural networks. Neural networks are an intelligent science that mimics the structure and function of the human brain, the central nervous system. They possess rapid response capabilities, facilitating real-time control and processing; excellent self-organization and self-learning abilities; and the ability to thrive in complex environments. They can fully approximate arbitrary nonlinear systems, quickly obtaining optimal solutions to problems satisfying various constraints; and exhibit superior performance such as high robustness and fault tolerance, thus finding increasingly widespread applications in communication systems. Neurosurgery is a computer system formed by interconnecting multiple very simple processing units in a certain way. This system processes information through dynamic responses without external input.

[0009] Numerous invention patents, both domestically and internationally, have been developed to address MIMO issues using neural networks. Chinese patent application number 201510473741.6 (Complex Neural Network Channel Prediction Method, Xi'an University of Electronic Science and Technology) discloses a complex neural network channel prediction method, primarily addressing channel fading caused by time-varying channel dynamics in MIMO systems. The technical solution involves: 1. The base station measures the channel to obtain a channel coefficient training sequence containing estimation errors; 2. Based on the obtained channel coefficient sequence, corresponding training samples and expected outputs are obtained; 3. The training samples are input to train a complex wavelet neural network, yielding the final network weights; 4. The base station uses the trained complex wavelet neural network to predict channel coefficients. This method is simple, effective, and suitable for reducing the impact of time-varying channel dynamics on MIMO system channels.

[0010] Chinese patent application number 201810177829.7 (A Wireless Channel Modeling Method Based on Neural Networks, Southeast University) discloses a wireless channel modeling method based on neural networks. This invention first processes the received signal fed back by the user to obtain estimated channel parameters; then, it obtains the three-dimensional geographic information of the scatterers from a two-dimensional image and clusters them; finally, it uses the channel parameters and geographic information as input to the neural network and the received signal as output to train a nonlinear time-varying neural network model. This method obtains a more accurate channel model within acceptable complexity, meeting the channel modeling requirements of future 5G communication systems such as massive MIMO technology, bandwidth expansion, and high mobility scenarios.

[0011] Chinese patent application number 201810267976.3 (A Deep Neural Network for Detecting Large-Scale MIMO Systems Based on BP Algorithm, Southeast University) provides a deep neural network method for detecting large-scale MIMO systems based on the BP algorithm. This method constructs a deep neural network for large-scale MIMO system detection by expanding and mapping the factor graph of the confidence propagation iterative algorithm onto a neural network structure. The neurons in the deep neural network correspond to the nodes in the factor graph of the iterative algorithm, and the number of neurons in each layer is equal to the number of symbol nodes in the factor graph. The mapping function between hidden layers is the update formula for confidence information in the iterative algorithm, and the number of hidden layers equals the number of iterations of the iterative algorithm. Specifically, this invention also provides MIMO detection methods that construct two deep neural networks based on damped confidence propagation and maximum sum confidence propagation iterative algorithms, respectively. This invention achieves a lower bit error rate without increasing online computational complexity and is robust to various channel conditions and antenna configurations.

[0012] Chinese patent application number 201910063733.2 (An Optimized MIMO Detection Method Based on Deep Learning, Shanghai University) constructs an end-to-end MIMO transmission model. It uses the signal y(t) received at the MIMO device receiver and estimated imperfect channel state information to obtain the complex time-domain vector of the model as input to a deep neural network (DNN). The DNN is used to obtain an estimate of the bitstream at the device transmitter. Compared to existing technologies that obtain the estimate of the transmitted bitstream through hard decision-making, this invention improves accuracy and detection rate under imperfect channel information, ensuring low bit error rate detection performance with low-complexity algorithms, while also exhibiting good robustness even with inherent channel errors.

[0013] Chinese patent application number 201610327115.0 (A Deep Learning-Based Codebook Selection Method for Large-Scale MIMO, Chongqing University of Posts and Telecommunications) relates to a deep learning-based codebook selection method for large-scale MIMO. The method includes: collecting pilot information from the test area to construct a pilot training sequence, thereby obtaining pilot training samples; performing iterative learning on the pilot training samples using a neural network to obtain the final network weight values; and selecting the optimal codeword from the complete codebook based on the channel output by the learned neural network. Then, the unknown area is matched with the test area to obtain its wireless channel, and thus the codeword corresponding to the wireless channel is obtained. This invention can effectively, accurately, and quickly establish a wireless channel model and codebook lookup, avoiding channel estimation in unknown areas and greatly reducing the complexity of codebook selection in unknown areas.

[0014] Chinese patent application number 201811626005.X (A Signal Detection Method for Low-Complexity MIMO-NOMA Systems Based on Improved Gradient Projection Method, Chongqing University of Posts and Telecommunications) discloses a signal detection method for low-complexity MIMO-NOMA systems based on an improved gradient projection method, relating to wireless communication technology. Based on the sparse characteristics of active users in the system, the method utilizes the idea of ​​convex optimization algorithms to transform the system model into a rigorous quadratic programming problem. This problem is then iteratively solved, and the results of each iteration are preprocessed to achieve effective detection of active users and their signals. This invention overcomes the problem of slow convergence speed in traditional detection methods. Preprocessing the results of each iteration not only enables rapid convergence of detection results but also allows for the detection of the active user set. Its implementation is simple and its application range is wide.

[0015] Chinese patent application number 201910014714.0 (A Beamforming Matrix Design Method for MIMO Systems Based on Deep Learning, Nanjing University of Posts and Telecommunications) discloses a beamforming matrix design method for MIMO systems based on deep learning. The steps are as follows: First, a training sample set required for the deep learning network is obtained using a known algorithm; then, a deep learning neural network model is constructed, its parameters are initialized, and it is trained using the training sample set; next, the channel is acquired using pilot signals and fed into the neural network to predict the beamforming matrix coefficients; finally, the channel and beamforming matrix coefficients are combined to form the beamforming matrix. This method, using a deep learning neural network to obtain a beamforming matrix, can simultaneously balance performance and algorithm complexity, reducing latency while ensuring performance, enabling the MIMO system to provide real-time services.

[0016] Chinese patent application number 201810182937.3 (A Machine Learning-Based Adaptive Transmission Method for MIMO Links, Southeast University) discloses a machine learning-based adaptive transmission method for MIMO links. This method uses an unsupervised learning autoencoder algorithm for feature extraction and dimensionality reduction, incorporating deep learning concepts to reduce feature dimensionality and computational complexity while preserving key state information. The invention utilizes logistic regression to construct a mapping relationship between channel state information and transmission parameters. Unlike previous fixed parameterized models, this method can be trained based on sample data. With high-quality datasets covering all states, it can better establish the mapping relationship between channel state information and transmission parameters, and compared to the traditional single equivalent signal-to-noise ratio, it can more fully utilize channel state information. Furthermore, this invention performs CQI selection based on the channel matrix. By studying the channel matrix and noise variance, the machine learning-based adaptive transmission method for MIMO links is not constrained by receiver design and has universality.

[0017] Chinese patent application number 201710495044.X (A Deep Learning-Based Joint Precoding and Antenna Selection Method for MIMO Systems, Zhejiang Sci-Tech University) discloses a deep learning-based joint precoding and antenna selection method for MIMO systems, comprising the following steps: First, generating the training dataset required for deep learning using existing antenna selection methods; next, establishing a deep learning model, training the deep learning model using the training data, and saving the model; then, using the saved deep learning model to complete antenna selection; finally, performing optimal precoding design on the selected MIMO subsystem. This invention utilizes deep learning technology to design joint precoding and antenna selection for MIMO systems, achieving low computational complexity while obtaining a good system signal-to-noise ratio.

[0018] Chinese Patent Application No. 201910242525.9 (A Deep Signal Detection Method for High-Speed ​​Rail, Shenzhen University) proposes a deep signal detection method for high-speed rail. First, data is collected by gathering transmitted and received signals in various scenarios along the high-speed rail line, based on different environmental types. Second, scenarios are divided into multiple regions through data analysis to ensure neural network compatibility. Third, a deep high-speed rail signal detection neural network model is established. Next, the neural network is trained offline. Finally, online real-time signal detection is performed. During operation, the high-speed rail's location is determined via GPS, its region is identified, and an appropriate neural network model is selected. The received signals are then input into the trained neural network, which outputs the base station's transmitted signals in real time. This invention significantly improves system performance, reduces the signal detection bit error rate, and enhances algorithm robustness. The method eliminates the need for channel estimation, saving pilot overhead.

[0019] Chinese Patent Application No. 201810279530.2 (A Method for Anti-interference and Noise Reduction of Visible Light Communication MIMO Based on BP Neural Network, South-Central University for Nationalities) discloses a method for anti-interference and noise reduction of visible light communication MIMO based on BP neural network, relating to MIMO antenna technology in the field of visible light communication. The system includes a system transmitter, a system device receiver signal processing section, and a BP neural network signal processing section connected in sequence. The method is as follows: 1) An electrical signal is loaded onto an LED array and transmitted in the form of an optical signal; 2) A photodetector at the device receiver converts the optical signal into an electrical signal; 3) Multiple electrical signals are filtered by a low-pass filter to remove high-frequency interference; 4) After training, the BP neural network performs noise reduction and interference elimination processing on the multiple signals, and finally converts them into a binary serial data stream through parallel-to-serial conversion. This invention improves the transmission performance of existing MIMO technology; combines neural networks with visible light MIMO technology to leverage the advantages of neural networks in noise reduction and interference reduction in wireless communication; and uses neural network receiver processing technology to make the entire VLC system more stable.

[0020] Chinese Patent Application No. 201710213235.2 (A Joint Equalization Method for Visible Light Channels Based on Orthogonal Mapping and Probabilistic Neural Networks, Sun Yat-sen University) discloses a joint equalization method for visible light channels based on orthogonal mapping and probabilistic neural networks. The method includes a transmitter and a receiver. The signal is transmitted from the transmitter to the receiver via a visible light MIMO channel. The visible light MIMO channel is a multiple-input multiple-output (MIMO) channel. The joint equalization combines pre-equalization and post-equalization. This invention employs a joint equalization scheme combining pre-equalization and post-equalization techniques, namely, a joint equalization method for visible light MIMO channels based on orthogonal mapping and probabilistic neural networks. This method effectively suppresses interference between channels in visible light MIMO communication systems and improves data transmission reliability.

[0021] Chinese Invention Patent No. 201910125325.5 (A Deep Learning-Based MIMO Decoding Method, Apparatus, and Storage Medium, Shenzhen Baolian Artificial Intelligence Technology Co., Ltd.) discloses a deep learning-based MIMO decoding method, apparatus, and storage medium. The invention involves constructing a training dataset for MIMO decoding, including multiple training data sets; then training a neural network based on this dataset to obtain a trained neural network model; upon receiving a MIMO signal to be decoded, inputting the signal into the neural network model for MIMO decoding, and finally obtaining the MIMO decoding result output by the neural network model. Through the implementation of this invention, a neural network model for joint MIMO detection and channel decoding is designed based on deep learning, treating MIMO detection and channel decoding as a joint decoding process. Furthermore, training improves the approximation of the neural network model's output, ensuring the overall performance of MIMO decoding and achieving higher decoding accuracy and faster decoding speed.

[0022] Chinese patent application number 201810757547.4 (A Machine Learning-Assisted Downlink User Scheduling Method for Large-Scale MIMO, Southeast University) discloses a machine learning-assisted downlink user scheduling method for large-scale MIMO, comprising the following steps: S1: The base station obtains the feature mode energy coupling matrix in the feature direction through the uplink probe signal sent by the user; S2: The base station uses the feature mode energy coupling matrix to perform sum rate calculation under various user and beam combinations with the assistance of machine learning methods; S3: A greedy algorithm is used to implement user scheduling based on the sum rate maximization criterion to obtain the optimal user beam pairing combination. This invention obtains statistical channel information through uplink probe signals and uses the sum rate maximization criterion for user scheduling. Even when the base station only has statistical channel information, through targeted feature extraction and neural network design, it accurately achieves approximate calculation of the sum rate, greatly reducing the complexity of user scheduling under large-scale antennas, and achieving near-optimal performance, with good applicability and robustness.

[0023] Chinese Invention Patent No. 201610353881.4 (A Modulation Recognition Method for MIMO Correlated Channels Based on Machine Learning Algorithms, Beijing University of Posts and Telecommunications) is a modulation recognition method for MIMO correlated channels based on machine learning algorithms, belonging to the field of communications. The specific steps are as follows: First, each data stream at the communication transmitter is space-time coded, and each codeword is transmitted through Nt transmit antennas. Then, the MIMO channel matrix H is calculated based on the correlation matrix at the receiver and the correlation matrix at the transmitter. Based on the MIMO channel matrix H, the received signal on each receiving antenna is calculated and corrected. Finally, features are extracted from the corrected signal at each receiving antenna, and the extracted feature values ​​are used for training and testing to calculate the modulation recognition mode to which the sample belongs. The advantages are: strong robustness and generalization ability to non-Gaussian channels; modulation scheme recognition can be achieved in more complex environments through parameter iteration; and by extracting features of higher-order moments and higher-order cumulants, the signal feature differences are obvious at higher signal-to-noise ratios, facilitating classification by machine learning algorithms.

[0024] The application of MIMO technology makes space a resource that can be used to improve performance and increase the coverage of wireless systems. MIMO technology has become one of the key technologies in the field of wireless communication, and through continuous development in recent years, it has been increasingly applied to various wireless communication systems. However, with the increase in the number of antennas used, the complexity of implementing MIMO technology increases significantly, thus limiting the number of antennas that can be used and preventing the full realization of MIMO's advantages. Meanwhile, since the formal introduction of artificial intelligence as a discipline in 1956, it has made significant progress over the past 50 years, becoming a broad interdisciplinary and cutting-edge science. As an important branch of artificial intelligence, as can be seen from the above description, neural networks, implemented through computers to create mathematical networks, are widely recognized as a solution to some of the bottleneck problems currently encountered in communication.

[0025] Based on current domestic and international research on key technologies related to neural networks and communication, numerous research results have been achieved, proposing feasible solutions from the perspectives of MIMO channel estimation and signal detection. However, how to improve system capacity and reduce the algorithmic and implementation complexity of MIMO technology while maintaining certain system performance has become a significant challenge for the industry. Currently, existing research on MIMO conditions has not yet explored solutions for improving system multiplexing gain using artificial intelligence techniques, given a fixed number of transmit and receive antennas.

[0026] Furthermore, on the one hand, there are differences in understanding among those skilled in the art; on the other hand, the inventors studied a large number of documents and patents when making this invention, but due to space limitations, not all details and contents were listed in detail. However, this does not mean that the present invention does not possess the features of these prior art. On the contrary, the present invention already possesses all the features of the prior art, and the applicant reserves the right to add relevant prior art to the background art. Summary of the Invention

[0027] Regarding the question of how to improve system capacity and reduce the algorithmic and implementation complexity of MIMO technology while ensuring certain system performance, existing research on MIMO conditions has not yet explored solutions for improving signal transmission and detection by combining artificial intelligence technology under the given number of transmitting and receiving antennas.

[0028] This application, combining the latest research progress in artificial intelligence and compressed sensing technologies, proposes a signal multiplexing transmission and detection scheme in MIMO systems based on artificial intelligence and compressed sensing technologies. On the one hand, compared with traditional MIMO technology schemes, this application reduces the dimensionality of the signal to be processed by introducing a sparsification module and a compressed sensing multiplexing module at the transmitting end of the device. This allows parallel data streams exceeding the number of transmitting antennas to be multiplexed onto a given transmitting antenna for transmission, thereby significantly improving the signal multiplexing gain under the given number of transmitting and receiving antennas in the MIMO system, and better meeting the application requirements of MIMO systems for broadband transmission. On the other hand, the compressed multiplexing matrix proposed in this invention does not rely on channel state information. The technical solution adopted in this invention can reconstruct the compressed multiplexed signal without modifying the existing MIMO technology scheme. This is achieved by adding sparsification and compressed sensing multiplexing steps at the transmitting end of the device and adding a mature optimized reconstruction algorithm in the field of compressed sensing at the receiving end of the device. This requires minimal modification to the existing MIMO system and has the advantage of being easy to implement.

[0029] The MIMO device based on artificial intelligence and compressed sensing technology proposed in this application mainly consists of a sparsification module and a compression multiplexing module at the transmitting end, and a demultiplexing module at the receiving end. Preferably, the transmitting end also includes a random number generator (or raw information bit generation module), a bit-level processing module, and a modulation module, while the receiving end includes a channel estimation module.

[0030] Preferably, such as Figure 5 As shown, at the transmitting end of the device: the raw data generated by the random number generator passes through the bit-level processing module and the modulation module to generate a modulated signal. The modulated signal then passes through the sparsification module, which expresses it as a sparse signal based on the first neural network model. The sparsified signal then passes through the compression and multiplexing module, which performs compression, dimensionality reduction, and multiplexing processing on the transmitted signal. The compressed and multiplexed signal is then transmitted through the transmitting antenna. At the receiving end of the device: the channel estimation module performs channel estimation on the received signal. The estimated channel state information passes through the demultiplexing module, which reconstructs and outputs the original transmitted data stream x based on the second neural network model.

[0031] The above process specifically includes the following:

[0032] An AI-based MIMO multi-antenna signal transmission and detection device includes: a sparsification module that uses a first neural network model to sparsely represent the original signal; a compression and multiplexing module that performs compression and dimensionality reduction processing on the sparsely represented signal; and a receiving end that processes the received signal to reconstruct the target signal. The device also includes a demultiplexing module, configured to use a compressed sensing reconstruction algorithm at the receiving end to obtain the sparse representation vector from the low-dimensional signal. Finally, the original signal x is reconstructed from the received signal using a second neural network model.

[0033] According to a preferred embodiment, the input and output of the second neural network model are opposite to the input and output of the first neural network model used by the sparsification module at the device transmitter to extract the sparse signal θ from the low-dimensional signal to be transmitted to the compression and multiplexing module for processing.

[0034] According to a preferred embodiment, the receiving end of the device further includes a channel estimation module, which is configured to perform channel estimation on the signal received by the receiving end of the device and use the resulting channel parameter matrix as input to the compressed sensing reconstruction algorithm.

[0035] According to a preferred embodiment, the demultiplexing module obtains the sparse representation vector by solving for the compressed sensing multiplexing matrix and the channel parameter matrix, which do not depend on channel state information. of.

[0036] According to a preferred embodiment, the sparsification module creates a neural network using the BP neural network training method in artificial intelligence. It constructs a first set of training samples and trains the neural network to obtain the first neural network model by using the transmitted signal vector x from the transmitter of the device and the sparse representation θ determined based on the transmitted signal vector x from the transmitter of the device as inputs and outputs, respectively.

[0037] According to a preferred embodiment, the demultiplexing module constructs a redundant dictionary D by using all possible combinations of the transmitted signal vector x as different column vectors of the redundant dictionary, thereby realizing a sparse representation θ of the transmitted signal vector, i.e., x = Dθ.

[0038] According to a preferred embodiment, the sparse representation θ after sparsification is compressed into ρl signals by the compression multiplexing module, and then the device transmits the compressed and multiplexed signal z through the transmitting antenna.

[0039] According to a preferred embodiment, the compressed and multiplexed signal z is obtained by calculating z = Aθ, where A is N. t A compressed dimension reduction matrix with m rows and m columns. This represents the compression ratio.

[0040] The system is an AI-based MIMO multi-antenna signal transmission and detection system. The system constructs an end-to-end MIMO transmission model. The transmitting end uses its constructed first neural network model to extract the sparse signal from the low-dimensional signal before it is transmitted to the compression and multiplexing module for processing. The receiving end uses its constructed second neural network model, which has the opposite input and output to the first neural network model of the sparsification module, to reconstruct the original signal from the received signal.

[0041] According to a preferred embodiment, the input to the second neural network model is a sparse representation vector obtained by the compressed sensing reconstruction algorithm. Attached Figure Description

[0042] Figure 1 This is a schematic block diagram of the signal processing flow of the signal transmission and detection system of the MIMO multi-antenna system based on artificial intelligence and compressed sensing technology provided by the present invention.

[0043] Figure 2 This is a schematic block diagram of the preferred signal compression multiplexing and detection processing procedure provided by the present invention;

[0044] Figure 3 These are the bit error rate performance curves of the classic detection algorithm ZF and the multi-antenna signal transmission and detection technology of the MIMO system of this invention under different transmit and receive antenna configurations;

[0045] Figure 4 These are the bit error rate performance curves of the classic detection algorithm ZF and the multi-antenna signal transmission and detection technology of the MIMO system of this invention under different transmit and receive antenna configurations; and

[0046] Figure 5 This is a schematic diagram of the module connection of the signal transmission and detection system of the MIMO multi-antenna system based on artificial intelligence and compressed sensing technology provided by the present invention.

[0047] List of reference numerals

[0048] 1: Device transmitter 101: Random number generator

[0049] 102: Bit-level processing module; 103: Modulation module

[0050] 104: Sparsification Module 105: Compression and Multiplexing Module

[0051] 2: Device receiver 201: Channel estimation module

[0052] 202: Demultiplexing module Detailed Implementation

[0053] The present invention will now be described in detail with reference to the accompanying drawings.

[0054] The MIMO device based on artificial intelligence and compressed sensing technology proposed in this application mainly consists of a sparsification module 104 and a compression multiplexing module 105 at the transmitter 1, and a demultiplexing module 202 at the receiver 2. Preferably, the transmitter 1 also includes a random number generator 101 (or raw information bit generation module), a bit-level processing module 102, and a modulation module 103, while the receiver 2 includes a channel estimation module 201.

[0055] Preferably, such as Figure 5 As shown, at the transmitting end 1, the raw data generated by the random number generator 101 is processed by the bit-level processing module 102 and the modulation module 103 to generate a modulated signal. The modulated signal then passes through the sparsification module 104, which expresses it as a sparse signal based on the first neural network model. The sparsified signal then passes through the compression and multiplexing module 105, which performs compression, dimensionality reduction, and multiplexing processing on the transmitted signal. The compressed and multiplexed signal is then transmitted through the transmitting antenna. At the receiving end 2, the channel estimation module 201 performs channel estimation on the received signal. The estimated channel state information is then processed by the demultiplexing module 202, which reconstructs and outputs the original transmitted data stream based on the second neural network model.

[0056] like Figure 2 As shown, the technical solution adopted in this invention is a signal sparsity design, compressed multiplexing matrix design, and signal detection method based on compressed sensing in a MIMO system. The technical solution is as follows:

[0057] 1. Signal processing at transmitter 1 of the device:

[0058] For a MIMO communication system equipped with l transmit antennas and l receive antennas, the system transmitter obtains l signals x after channel coding and signal modulation. The modulated l signals x are then mapped to a sparse signal θ by a pre-trained first neural network model. m×1 The first neural network model training algorithm uses the backpropagation (BP) algorithm.

[0059] 2. Compression and multiplexing processing at transmitter 1 of the device:

[0060] The sparse signal θ obtained after sparsification m×1 After passing through the compression and multiplexing module 105, the signal is compressed into ρl channels, and then the compressed and multiplexed signal z is transmitted through the transmitting antenna. The compression processing of the input signal by the compression and multiplexing module 105 can be expressed as: z = Aθ, where A is N. t A compressed dimension reduction matrix with m rows and m columns. This represents the compression ratio. The compression ratio ρ is determined by the size of the measurement matrix A in compressed sensing technology. Preferably, a Gaussian random matrix is ​​selected as the compressed sensing multiplexing matrix / compression dimensionality reduction matrix A at the transmitter 1 of the device.

[0061] 3. Signal detection at receiver 2 of the device:

[0062] The process of the receiver 2 link is roughly the reverse of the transmitter 1 link. The signal received by receiver 2 is: y = Hz + n = HAθ + n. Here, y is an Nr×1 received signal vector, which is the compressed and multiplexed ρl-path modulation symbols received by receiver 2 from n transmitting antennas; n is an Nr×1 Gaussian white noise vector, whose elements are independent and identically distributed complex Gaussian variables with a mean of 0 and a variance of 1; H is an Nr×Nt channel propagation matrix, and it is a deterministic matrix that remains unchanged within a coherent time interval. Receiver 2 can estimate the channel propagation matrix H based on the pilot signals inserted in the transmitted data.

[0063] Assuming the channel matrix H follows a Gaussian distribution, it can be proven that the new matrix HA also follows a Gaussian distribution, satisfying the conditions required for a compressed sensing measurement matrix. Based on the compressed sensing multiplexing matrix HA, the decompression and multiplexing process is achieved by solving the following optimization problem, calculating and determining the transmitted sparse vector: min||θ||0 sty=H·A·θ. Here, the Bayesian Compressive Sensing (BCS) reconstruction algorithm can be used to obtain the sparse signal.

[0064] 4. Based on deep learning training methods, a second neural network model is generated using sparse signals. Mapped to Finally, for Demodulate the signal to restore the original transmitted data stream.

[0065] Compared with the prior art, the beneficial effects of the present invention are:

[0066] The greatest advantage of the technical solution adopted in this invention is that, based on existing MIMO systems, it uses a sparsity module 104 and a compressed sensing multiplexing module. In contrast, the traditional signal transmission process involves codewords being modulated and then transmitted, followed by signal detection and reconstruction of the original signal at the receiving end 2. This application reduces the dimensionality of the signal to be processed, allowing parallel data streams exceeding the number of transmitting antennas to be multiplexed onto a given transmitting antenna for transmission. This significantly improves the signal multiplexing gain under the given number of transmitting and receiving antennas in a MIMO system, better meeting the application requirements of MIMO systems for broadband transmission. Simultaneously, by combining the BP neural network training method from artificial intelligence, sparse representation and reconstruction of the signal are achieved.

[0067] Second, the compressed multiplexing matrix proposed in this invention does not rely on channel state information. The technical solution adopted in this invention can reconstruct the compressed multiplexing signal without modifying the existing MIMO technology solution. It adds sparsification and compressed sensing multiplexing steps to the transmitting end 1 of the device and adds a mature optimized reconstruction algorithm in the field of compressed sensing to the receiving end 2 of the device. It requires little modification to the existing MIMO system and has the advantage of being easy to implement.

[0068] Third, the artificial intelligence method used in this invention sparses the signal representation and reconstructs the original signal based on the sparse signal, ensuring the feasibility of compressing and transmitting multiple data streams while taking into account the bit error rate.

[0069] The specific implementation steps of the above process are explained below:

[0070] Regarding the sparse representation of signals: Compressed sensing, also known as compressed sampling or sparse sampling, is a method for finding sparse solutions to underdetermined linear systems. Compressed sensing is one of the most remarkable achievements in signal processing since the beginning of the 21st century, and it has been effectively applied in fields such as magnetic resonance imaging, image processing, and wireless communication systems.

[0071] Converting analog signals into digital signals that computers can process inevitably involves a sampling process. To ensure signal integrity, Nyquist provided the answer—the sampling frequency should be twice the highest frequency of the signal. The Nyquist sampling theorem has long been considered the golden rule of digital signal processing. Candès was the first to recognize the possibility of a breakthrough and, with the assistance of Terence Tao and Donoho, proposed compressed sensing theory. This theory posits that if a signal is sparse, it can be reconstructed from a much smaller number of sampling points than required by the sampling theorem.

[0072] Compressed sensing theory differs from the traditional Nyquist sampling theorem. If a signal x is compressible or sparse in a certain transform domain D, then a high-dimensional sparse signal obtained from the transformation can be projected onto a low-dimensional space using an observation matrix A independent of the transform basis D. Then, by solving an optimization problem, the original signal can be reconstructed with high probability from these few projections. Within this theoretical framework, the sampling rate is not determined by the signal bandwidth, but by the structure and content of the information in the signal. Compressed sensing theory mainly includes three aspects: sparse representation of signals, coded sampling, and reconstruction algorithms. Since signals commonly found in nature are generally not sparse, sparse representation of a signal means that when the signal is projected onto a certain transform domain D, only a few elements are non-zero. The resulting transform vector is then said to be sparse or approximately sparse, i.e., x = Dθ, where θ is a concise expression for the original signal x. This is a priori condition for compressed sensing.

[0073] Finding the optimal sparse domain for a signal is the foundation and prerequisite for the application of compressed sensing theory. Only by selecting a suitable basis to represent the signal can the sparsity of the signal be guaranteed, thereby ensuring the accuracy of signal recovery. When studying the sparse representation of a signal, the sparse representation capability of the transform basis can be measured by the decay rate of the transform coefficients. Candès et al., in their paper "Nearoptimal signal recovery from random projections: Universal encoding strategies," IEEE Trans. Information Theory, vol. 52, no. 12, pp. 5406-5425, 2006, pointed out that signals with power-law decay can be recovered using compressed sensing theory. In recent years, a hot topic in sparse representation research has been the sparse decomposition of signals under a redundant dictionary. This is a completely new signal representation theory: replacing the basis functions with an overcomplete library of redundant functions, called the redundant dictionary D, where the elements are called atoms. In his paper "Nonlinear Methods of Approximation," IMI Research Reports, Dept. of Mathematics, University of South Carolina, 2001, Temlyakov points out that the choice of dictionary D should conform as closely as possible to the structure of the signal being approximated, and its composition can be unrestricted. Finding a small number of atoms with the best linear combination to represent a signal x = Dθ from a redundant dictionary is called sparse approximation or highly nonlinear approximation of the signal, where only K elements of θ are non-zero. The composition of the overcomplete redundant dictionary D should conform as closely as possible to the inherent characteristics of the signal itself, which is crucial for the sparse representation of the signal. The closer the structure of the overcomplete redundant dictionary is to the characteristics of the signal, the fewer atoms are needed, the sparser θ becomes, the fewer measurements are required, and the more accurate the reconstruction performance. Theoretically, it is always possible to find a transform domain D that achieves a sparse representation of the signal.

[0074] The concept of deep learning originates from research on artificial neural networks. Neural networks, a concept proposed in 1986 by scientists led by Rumelhart and McClelland, are multi-layered feedforward neural networks trained using an error backpropagation algorithm. Currently, in practical applications of artificial neural networks, most neural network models employ BP networks and their variations. It is also the core component of feedforward networks, embodying the essence of artificial neural networks. Here, we use the BP algorithm to train the first neural network model (NN1) to achieve sparse signal representation.

[0075] Regarding the "observation matrix in a compressed sampling system": In compressed sensing theory, it's necessary to design the observation matrix A of the compressed sampling system, and how to sample a small number of observations while ensuring the original signal can be reconstructed from them. Clearly, if the observation process destroys information in the original signal, the reconstruction quality cannot be guaranteed. To ensure that the linear projection of the signal preserves its original structure, the projection matrix must satisfy the Restricted Isometry Property (RIP) condition. Then, the linear projection measurement of the original signal is obtained by multiplying the original signal by the measurement matrix. The RIP condition is defined as follows: If there exists a constant δ... K For all signals θ of sparsity K ∈ (0, 1), matrix A satisfies the following equation: Then matrix A is said to satisfy the constrained isometry property of order K, where sparsity K refers to the number of non-zero elements in the signal θ. A is N t A compressed dimension reduction matrix with row 1 and column 1. The advantage of compressed sensing technology is that even N... t <l (l refers to the length of the signal), we can still start from N r (N r =N t The original data of length l is recovered from the measurements. Let... This represents the compression ratio. According to the principle of compressed sensing, as long as the measurement A satisfies the RIP (Reduced Indicator Perspective), even if A is a matrix with a number of rows much smaller than the number of columns, projecting the signal θ onto a space with a reduced dimension, the original signal can still be completely recovered from the number of measurements much smaller than the signal dimension using the compressed sensing reconstruction algorithm. ρ determines the number of transmit and receive antennas that can be reduced and the reconstruction performance of the receiver. M. Davenport, in Theorem 3.5 of his doctoral dissertation "Randomobservation on random observations: Sparse signal acquisition and processing," states that A satisfies the 2Kth order RIP constant δ. 2K A matrix ∈ (0, 1] only needs to If C is a constant approximately equal to 0.28, the original signal can be recovered. Meanwhile, Donoho, in his paper "Extensions of compressed sensing," *Signal Processing*, vol. 86, no. 3, pp. 533-548, 2006, gives three necessary conditions for an observation matrix, pointing out that most uniformly distributed random matrices possess these three conditions and can be used as observation matrices, such as partial Fourier sets, partial Hadamard sets, and uniformly distributed random projection sets. His papers "decoding by linear programming," *IEEE Transactions on Information Theory*, vol. 51, no. 12, pp. 4201-4215, 2005, and "Stable signal recovery from incomplete and inaccurate measurements," *Communications on Pure and Applied Mathematics*, vol. 59, no. 8, pp. 1207-1223, 2006, prove that when the measurement matrix A is a Gaussian random matrix, A can satisfy RIP with a relatively high probability. Therefore, a Gaussian random matrix is ​​selected as the compressed sensing multiplexing matrix A in this application.

[0076] Regarding the "signal reconstruction at receiver 2": the underdetermined problem is finally solved using a reconstruction algorithm from compressed sensing. Donoho, in his paper "For most large underdetermined systems of linear equations, the minimal l0-norm solution is also the sparsest solution," Communications on Pure and Applied Mathematics, vol. 59, no. 6, pp. 797-829, 2006, points out that the minimal l0-norm problem is an NP-hard problem, requiring exhaustive enumeration of all possible permutations of non-zero values ​​in θ, and therefore cannot be solved. In light of this, researchers in this field have proposed a series of algorithms to obtain suboptimal solutions, mainly including the following four categories:

[0077] (1) Greedy pursuit algorithm: This type of method gradually approximates the original signal by selecting a local optimum solution in each iteration. These algorithms include the piecewise OMP algorithm proposed by Donoho in the literature "Sparse solution of underdo-termined linear equations by stagewise orthogonal matching pursuit", Technical Report, 2006;

[0078] (2) Convex relaxation method: This type of method finds the approximation of the signal by transforming the non-convex problem of solving the l0 norm into a convex problem. Examples include the interior point method proposed in the paper "A method for large scale regularized least squares", IEEE Journal on Selected Topics in Signal Processing, vol.4, no.1, pp.606-617, 2007; the gradient projection method proposed in the paper "Gradient projection for sparse reconstruction: Application to compressed sensing and other inverse problems", Journal of Selected Topics in Signal Processing: Special Issue on Convex Optimization Methods for Signal Processing, vol.1, no.4, pp.586-598, 2007; and the iterative thresholding method proposed by Daubechies in the paper "Aniterative thresholding algorithm for linear inverse problems with a sparsity constraint", Comm. Pure Appl. Math., vol.57, no.11, pp.1413-1457, 2004.

[0079] (3) Bayesian Compressive Sensing Reconstruction (BCS) Algorithm: This type of method uses Bayesian priors to give the solution signal a reasonable prior distribution and then derives the original signal, such as the BCS (Bayesian Compressive Sensing) algorithm proposed in "Bayesian compressive sensing using laplacepriors", IEEE Trans. Image Process, vol.19, no.1, pp.53-63, 2010;

[0080] (4) Combinatorial algorithms: These methods require that the sampling of the signal supports rapid reconstruction through group testing, such as the Fourier sampling proposed in the literature "Improved time bounds for near optimal sparse Fourier representation", Proceedings of SPIE, Wavelets XI, Bellingham WA: International Society for Optical Engineering, 2005, and the HHS (Heavy Hitters on Steroids) tracking proposed in the literature "One sketch for all: Fastalgorithms for compressed sensing", Proceedings of the 39th Annual ACMSymposium on Theory of Computing, New York: Association for Computing Machiner, pp. 237-246, 2007.

[0081] As can be seen above, each algorithm has its inherent drawbacks. The convex relaxation method requires the fewest observations to reconstruct the signal, but it often has a heavy computational burden. The greedy pursuit algorithm falls between these algorithms in terms of running time and sampling efficiency, but its noise resistance is unstable. A suitable reconstruction algorithm can be selected based on different environments; once the sparse representation vector is obtained, the original signal can be recovered.

[0082] Regarding "signal reconstruction at receiver 2": For example... Figure 1As shown, compared with the traditional signal transmission process: the codeword is modulated and then sent, and the receiving end 2 of the device performs signal detection and reconstructs the original signal, this application adds a sparsification module 104 and a compression multiplexing module 105 to the transmitting end 1 of the device. First, the modulated signal is expressed as a sparse signal through deep learning to realize the nonlinear expression of the signal; then the sparse signal is compressed and reduced in dimension. The selection of the compression and reduction matrix does not require channel state information. The measurement matrix in the compressed sensing technology can be selected as the signal compression matrix to complete the compression, reduction and multiplexing of the transmitted signal. The receiving end 2 of the device reconstructs the signal in two steps: (1) the high-dimensional sparse signal is decomposed from the low-dimensional received signal through the compressed sensing reconstruction algorithm. (2) The original signal is reconstructed using the second neural network model (NN2) trained by deep learning algorithms.

[0083] This invention, based on existing MIMO system signal multiplexing techniques, introduces a sparsity module 104, a compression multiplexing module 105, and a demultiplexing module 202 at the receiving end 2 of the device. Compared to traditional MIMO schemes, it reduces the required number of antennas while simultaneously transmitting the same amount of data, thereby improving the multiplexing gain and capacity of the MIMO system. Compared to existing MIMO spatial multiplexing techniques, we no longer focus solely on eliminating interference between adjacent data streams, but rather on how to multiplex more data streams to the receiving end 2 within a given number of transmit antennas, while ensuring the detection performance of the receiving end 2, thus achieving multiplexing gain and transmission capacity exceeding the inherent multiplexing gain of the MIMO system.

[0084] Example

[0085] This embodiment integrates the MIMO multi-antenna signal transmission and detection technology based on artificial intelligence and compressed sensing proposed in this application, and provides a detailed example of the specific implementation steps of the present invention.

[0086] First, the information source is generated by using a random number generator 101 to produce a 0,1 bit sequence.

[0087] Modulation is the modulation of bit data, including BPSK, QPSK, 16QAM and 64QAM.

[0088] This embodiment uses BPSK modulation as an example for illustration.

[0089] According to such Figure 2 The signal processing flow at the transmitting end 1 and receiving end 2 of the device shown is as follows:

[0090] S1: Signal processing at device transmitter 1.

[0091] S11: Use random number generator 101 to generate a 0,1 bit sequence to form the original data.

[0092] S12: After BPSK modulation, signal x is generated. The transmitted data for each group can be different.

[0093] S13: Train the neural network model. Create a neural network using the BP neural network training method in artificial intelligence. Use the transmitted signal vector x from the transmitter 1 of the device and the sparse representation θ determined based on the transmitted signal vector x from the transmitter 1 of the device as inputs and outputs, respectively, construct a first set of training samples, train the neural network, and obtain the first neural network model.

[0094] When the transmitted signal is x, the signal received at device receiver 2 is y. Theoretically, we can always find a suitable basis to achieve a sparse representation of the signal. Temlyakov, in his paper "Nonlinear Methods of Approximation," IMI Research Reports, Dept. of Mathematics, University of South Carolina, 2001, points out that the choice of dictionary D should best conform to the structure of the approximated signal, and its composition can be unrestricted. Here, all possible combinations of x are used as different column vectors of the redundant dictionary D to form the redundant dictionary D, and the sparse representation corresponding to x is found: a sparse vector θ where all positions except the corresponding index position are 1.

[0095] S14: Sparse representation of the signal. The first neural network model is obtained through training using the BP neural network method in artificial intelligence. When the input is a signal x, the output is the corresponding sparse signal θ.

[0096] S15: Compression and multiplexing processing. The sparse signal θ is transmitted by multiplying the data by the compression and multiplexing matrix A to obtain the data vector z, i.e., z = Aθ.

[0097] The compression multiplexing matrix A mentioned here is chosen as a Gaussian matrix. Wherein, This represents the compression ratio, indicating the percentage by which the number of antennas is reduced.

[0098] S16: Transmit the data through the channel.

[0099] S2: Signal detection at receiver 2 of the device.

[0100] S21: The received signal is y = Hz + n = HAx + n, where n represents noise. The signal matrix H is estimated using the channel estimation module 201.

[0101] S22: Given the multiplexing matrix A and the channel matrix H, solve the following optimization problem and calculate the sparse vector θ:

[0102] min||θ||0 sty=HA·θ

[0103] In the above formula, ||·||0 is the l0 norm of the vector, which represents the number of non-zero elements in the sparse vector θ.

[0104] Assuming the channel matrix follows a Gaussian distribution, it can be proven that the new matrix HA also follows a Gaussian distribution, satisfying the conditions required for compressed sensing measurement matrices. The underdetermined problem in compressed sensing can be solved using reconstruction algorithms. In this embodiment, the Bayesian compressed sensing BCS reconstruction algorithm is used. Based on the multiplexing matrix A and channel matrix H known at receiver 2, the sparse representation vector is obtained.

[0105] S23: Train the neural network model. Swap the input and output order of the first set of training samples to form the second set of training samples. Create a neural network using the BP neural network training method in artificial intelligence, using the sparse representation θ and the transmitted signal vector x as input and output respectively, construct the second set of training samples, train the neural network, and obtain the second neural network model.

[0106] S24: According to the second neural network model, when the input is a signal At that time, the reconstructed transmission data is obtained.

[0107] The following further explains steps S13 and S23: The neural network model consists of three parts: an input layer (layer1), intermediate layers (layer2, ..., L-1), and an output layer (layerL). The input layer acts as a signal transmission layer, responsible for receiving external input information. Each unit in the input layer represents a feature. The intermediate layers can be single or multiple, and they act as internal information processing layers, responsible for information transformation. The output layer outputs information to the outside world, and each unit in the output layer represents a category. In this application, a backpropagation (BP) neural network is used to simulate a mapping function that maps input space data to output space. The BP neural network will try to fit a function between the input and output of the transmitted signal vector x of the device transmitter 1 and the sparse representation θ determined based on the transmitted signal vector x of the device transmitter 1, and the input and output of the sparse representation θ and the original transmitted data stream, respectively. The function between; the mapping function model generated by the trained BP neural network can obtain the sparse representation θ of the sparse original signal x on the overcomplete redundant dictionary D according to the modulated transmitted signal vector x of the transmitter 1 of the device, and reconstruct the original transmitted data x according to the sparse representation θ calculated by the Bayesian reconstruction algorithm.

[0108] This application uses a cost function to measure the difference between the output of the BP neural network and the true output. The BP neural network is trained so that the output of the network's input (the signal received by receiver 2) is as close as possible to the theoretical output. To minimize the cost function, gradient descent is used to solve for the neural network parameters. Once the optimal neural network weights are found, a first or second neural network model is established. The BP neural network is created by collecting a large amount of sample data and manually labeling the correct classification results. This labeled data is then used to train the created neural network. During this process, each layer in the neural network continuously adjusts its weights and biases based on the difference between the current output value and the correctly labeled target value until it can accurately output the target value.

[0109] Further explanation regarding the two parameters required for training a neural network—weights and biases: In this application, the weight parameter matrix between each layer of the neural network is used... This indicates that the superscript of the weight parameter w represents the layer number, and the subscript represents the nth node in each of the two adjacent layers. For example, This represents the weight of the line segment between the first node of Layer 1 and the second node of Layer 2. These weights determine the model's function; the goal of the neural network is to calculate these weights using samples. Each node in the intermediate and output layers represents a Logistic function g(z) = a. For example... This represents the input value of the first node in Layer 2, which is then substituted into the Logistic function to obtain the output.

[0110] The bias parameter matrix between each layer of the neural network is: B = [b1b2...b n ] T Given that the input to the neural network is: Y = [y1y2...y...] n ] T The output of the neural network is: Introduce a nonlinear operator: Then we can deduce that:

[0111] Initialize weight parameters: Randomly initialize the weight parameters w to a value between [-ε, ε], where ε is a preset, sufficiently small value.

[0112] Training a neural network model: The process of training a neural network model mainly consists of two steps: first, calculating the cost function J(θ); and second, adjusting the parameters θ to minimize the cost function value J(θ). The forward propagation algorithm is used below to calculate the output of each sample under the current neural network model, obtain the cost function, and then update the weight parameters based on the output. The cost function J(θ) is defined, where m is the number of samples. Since the neural network has K outputs, its cost function also calculates the cost of K outputs accordingly. The calculation formula is as follows:

[0113]

[0114] The backpropagation algorithm is used to adjust the parameter θ to minimize the cost function value J(θ). The backpropagation algorithm updates the weight coefficients by calculating the partial derivatives of the cost function with respect to each weight coefficient. For example, first, the gradient of the last layer is calculated: (1) the gradient of the cost function value with respect to the nonlinear operator, (2) the gradient of the neural network output with respect to the bias and the weights between adjacent layers. The gradient is updated in the negative direction of the gradient. Second, the gradient of the penultimate layer is calculated: (1) the gradient of the backpropagation error of the upper layer with respect to the nonlinear operator, (2) the gradient of Hn-1 (H is the output of each layer after the activation function) with respect to the bias and the weights between adjacent layers, and the gradient is updated in the negative direction of the gradient. Finally, after backpropagation layer by layer, the gradient of the first layer is calculated: (1) the gradient of the backpropagation error of the second layer with respect to the nonlinear operator, (2) the gradient of H1 with respect to the bias and the weights between adjacent layers. The gradient is updated in the negative direction of the gradient. Thus, after the first backpropagation process is completed, the above two steps of forward propagation to obtain the output and backpropagation to update the parameters are continued until the mean square error is minimized, which completes the training process of the neural network model.

[0115] Figure 3 The bit error rate (BER) performance of a MIMO system with different transmit and receive antenna configurations after employing compressed sensing and neural network signal transmission and detection techniques is presented. A flat fading channel is assumed here. In the conventional scheme, only four transmit antennas can transmit four data symbols simultaneously. Applying the scheme proposed in this application, the signal is first modulated using BPSK to obtain the original signal x. 4 ×1 After passing through the first neural network model, the sparse signal θ is obtained. 16×1 A random Gaussian matrix A 4ρ×16As a compression and dimensionality reduction matrix, z is obtained. If ρ = 0.5, only 2 transmitting antennas are needed to transmit the original data. If ρ = 0.75, only 3 transmitting antennas are needed. Device receiver 2, also using 2 or 3 receiving antennas, employs the BCS reconstruction algorithm and a second neural network model trained by the neural network to obtain the reconstructed transmitted signal. The bit error rate performance of this scheme is shown in Figure (2×2)-4. The first number in parentheses represents the number of transmit antennas, the second number represents the number of receive antennas, and the last number represents the original data length. With an increased number of receive antennas, this scheme is denoted as (3×3)-4. Compared with the classic detection algorithm ZF (Zero Forcing), the scheme proposed in this application can reduce the number of transmit and receive antennas while maintaining a high SNR.

[0116] Figure 4 The bit error rate (BER) performance of a MIMO system with different transmit and receive antenna configurations after employing compressed sensing and neural network signal transmission and detection techniques is presented. A flat fading channel is assumed here. In the conventional scheme, 20 transmit antennas can only transmit 20 data symbols simultaneously. Using the scheme of this application, BPSK modulation of the signal is first employed, and then x... 20×1 Divided into 5 groups, the vector x in each group i The length of the sequence (i = 1, 2, 3, 4, 5) is 4. Each group of signals is processed by the first neural network model to obtain a sparse signal θ. i 16×1 (i = 1, 2, 3, 4, 5). Choose a random Gaussian matrix A. 4ρ×16 As a compression and dimensionality reduction matrix, z is obtained. i z i Concatenating them yields the vector to be sent. If ρ = 0.5, only 10 transmit antennas are needed to transmit the original data. If ρ = 0.75, 15 transmit antennas are needed. Receiver 2 also uses 10 receive antennas. The bit error rate performance of this scheme is shown in Figure (10×10)-20. The first number in parentheses represents the number of transmit antennas, the second number represents the number of receive antennas, and the last number represents the length of the original data. For the same data and the same number of packets, ρ = 0.75, the compression and dimensionality reduction matrix is ​​A. 3×4 When the number of receiving antennas increases, this scheme is denoted as (15×15)-20. Compared with the classic detection algorithm ZF (Zero Forcing), the scheme proposed in this application can reduce the number of transceiver antennas required while maintaining a high SNR. Therefore, our proposed scheme can reduce the number of transceiver antennas required while maintaining a high bit error rate.

[0117] As described above, the enhanced spatial multiplexing method proposed in this invention can, based on existing MIMO systems, combine artificial intelligence to transmit the same amount of data while reducing the number of antennas, thereby reducing the required number of antennas and improving multiplexing gain and system capacity.

[0118] It should be noted that the specific embodiments described above are exemplary, and those skilled in the art can devise various solutions inspired by the disclosure of this invention. These solutions all fall within the scope of this invention and its protection. Those skilled in the art should understand that this specification and its accompanying drawings are illustrative and not intended to limit the scope of the claims. The scope of protection of this invention is defined by the claims and their equivalents.

Claims

1. An antenna signal modulation method based on a neural network model, characterized in that, The method includes at least: A neural network is created using the BP neural network training method in artificial intelligence, with the transmitted signal vector of the device transmitter (1) as the basis. and the transmission signal vector based on the transmitter (1) of the device The determined sparse signal Using these as inputs and outputs respectively, a first set of training samples is constructed to train the neural network, resulting in a first neural network model. After channel coding and signal modulation, the model is obtained. Path transmission signal vector Modulated Path transmission signal vector The first neural network model, generated through pre-training, is mapped to a sparse signal. ; The sparsely represented signal is compressed and its dimensionality reduced. Sparse signal vectors are obtained from low-dimensional signals using compressed sensing reconstruction algorithms. Finally, the original signal is reconstructed from the received signal using a second neural network model that has the opposite input and output to the first neural network model. Based on deep learning training methods, a second neural network model is generated, consisting of sparse signal vectors. Mapped to Finally, regarding Demodulate the signal to restore the original transmitted data stream; The weight parameter matrix between layers of a neural network is used This indicates that the weight parameters are... The superscript indicates the layer number, and the subscript indicates the nth node of each of two adjacent layers; The bias parameter matrix between each layer of the neural network is as follows: The input to the neural network is known to be: The output of the neural network is: By introducing a nonlinear operator, we can derive: ;in, Represents the bias vector of each node; The dimension is The input vector; The dimension is The output vector; F represents the activation function; This represents the weight parameter between the k-th node and the i-th node; Initialize weight parameters: Set the weight parameters It is randomly initialized to a value between [-ε, ε], where ε is a preset value.

2. The antenna signal modulation method based on a neural network model according to claim 1, characterized in that, Methods for compressing and reducing the dimensionality of sparsely represented signals also include: The sparse signal obtained after sparsification The compressed and multiplexed signal is compressed into ρ1 channels by the compression and multiplexing module, and then the compressed and multiplexed signal is... Transmitted via transmitting antenna; The compression processing of the input signal by the compression multiplexing module can be represented as follows: ,in, yes OK The column compression and dimensionality reduction matrix, This represents the compression ratio.

3. The antenna signal modulation method based on a neural network model according to claim 1, characterized in that, Sparse signal vectors are obtained from low-dimensional signals using compressed sensing reconstruction algorithms. The methods include: Set the received signal as follows: ; Where y is Nr The received signal vector is ×1, which is the received signal vector. The compressed and multiplexed ρl-path modulation symbols transmitted by the root transmitting antenna; yes Nr A Gaussian white noise vector of size ×1, whose elements are independent and identically distributed complex Gaussian variables with mean 0 and variance 1. yes Nr × Nt The channel propagation matrix; Based on the compressed sensing multiplexing matrix The decompression and multiplexing process is achieved by solving the following optimization problem, which calculates and determines the sparse vector to be transmitted: ; For vectors Norm, representing sparse signals The number of non-zero elements in the neutron; The sparse signal vector is obtained by solving the compressed sensing reconstruction algorithm based on Bayesian methods. .

4. The antenna signal modulation method based on a neural network model according to claim 3, characterized in that, The method further includes: Channel estimation is performed on the signal received by the receiving end (2) of the device, and the resulting channel parameter matrix is ​​used as the input of the compressed sensing reconstruction algorithm.

5. The antenna signal modulation method based on a neural network model according to claim 1, characterized in that, The method further includes: The transmitted signal vector The redundant dictionary D is constructed by using all possible combinations as different column vectors of the redundant dictionary, thereby realizing the transmitted signal vector. sparse signal That is, x=D .

6. The antenna signal modulation method based on a neural network model according to claim 1, characterized in that, The training methods for the second neural network model include: The input and output of the first set of training samples are swapped to form the second set of training samples; A neural network is created using the BP neural network training method in artificial intelligence to base the sparse signal. and the transmitted signal vector Using these as inputs and outputs respectively, construct the second set of training samples; The neural network is trained to obtain the second neural network model.

7. The antenna signal modulation method based on a neural network model according to claim 1, characterized in that, Other methods for training neural network models include: Calculate the cost function The forward propagation algorithm is used to calculate the output of each sample under the current neural network model and obtain the cost function. Adjust the parameter θ so that the cost function value To minimize this, the backpropagation algorithm is used to update each weight coefficient by calculating the partial derivative of the cost function with respect to each weight coefficient.

8. The antenna signal modulation method based on a neural network model according to claim 7, characterized in that, The steps of using the backpropagation algorithm to calculate the partial derivatives of the cost function with respect to each weight coefficient include: Calculate the gradient of the last layer: calculate the gradient of the cost function value with respect to the nonlinear operator, and calculate the gradient of the neural network output with respect to the bias and the weights between adjacent layers; Calculate the gradient of the penultimate layer: calculate the gradient of the backpropagation error of the upper layer with respect to the nonlinear operator, calculate the gradient of Hn-1 (H is the output of each layer after the activation function) with respect to the bias and the weights between adjacent layers, and update the gradient in the negative direction of the gradient. After backpropagation layer by layer, the gradient of the first layer is calculated; the gradient of the error backpropagated from the second layer with respect to the nonlinear operator is calculated; and the gradient of H1 with respect to the bias and the weights between adjacent layers is calculated.

Images