CN-121984823-A - 5G base station signal uplink and downlink detection method, system, equipment and medium based on deep learning network

CN121984823ACN 121984823 ACN121984823 ACN 121984823ACN-121984823-A

Abstract

The invention discloses a 5G base station signal uplink and downlink detection method, a system, equipment and a medium based on a deep learning network, wherein the detection method comprises the following steps: the method comprises the steps of collecting base station signals, preprocessing the base station signals to achieve frame head synchronization, dividing the signals into time slots based on synchronization results, generating a time slot level image, inputting the time slot level image into a first deep learning network, identifying the time slot level image as a downlink time slot, an uplink time slot or a flexible time slot, dividing the identified flexible time slot into orthogonal frequency division multiplexing symbols, generating a symbol level image, inputting the symbol level image into a second deep learning network, identifying the symbol level image as a downlink symbol or an uplink symbol, and finally synthesizing a complete uplink and downlink resource allocation diagram. The invention combines physical layer signal characteristics with deep learning, gets rid of dependence on complex high-layer signaling analysis, reduces equipment cost and complexity, improves reliability and efficiency of synchronization and identification, and can be widely applied to repeater and network monitoring equipment.

Inventors

YING CHAO
JI JUN
WANG ZHENGYI
LING WEI
GAO XINJIAN
LI HAICHAO
JIANG MIAO

Assignees

昆山九华电子设备厂

Dates

Publication Date: 20260505
Application Date: 20260305

Claims (10)

1. The 5G base station signal uplink and downlink detection method based on the deep learning network is characterized by comprising the following steps of: s1, signal acquisition and preprocessing, namely acquiring signals with preset time length on a base station, and preprocessing the acquired signals; step S2, frame head synchronization, namely performing correlation operation on the preprocessed signals to realize frame head synchronization and determining the initial boundary of a wireless frame; S3, generating a time-frequency image, namely dividing the preprocessed signal into a plurality of time slots based on a frame header synchronization result, converting time-domain signals of a plurality of Orthogonal Frequency Division Multiplexing (OFDM) symbols contained in each time slot into frequency-domain signals, and generating a time-frequency level image representing time-frequency energy distribution of the time slot; s4, time slot level uplink and downlink identification, namely inputting time slot level images into a first deep learning network trained in advance, wherein the first deep learning network identifies that each time slot level image belongs to a downlink time slot or an uplink time slot or a flexible time slot; S5, symbol-level uplink and downlink identification, namely, for the flexible time slot identified in the step S4, further dividing the flexible time slot into a plurality of Orthogonal Frequency Division Multiplexing (OFDM) symbols, generating symbol-level images of the plurality of Orthogonal Frequency Division Multiplexing (OFDM) symbols, inputting the symbol-level images into a pre-trained second deep learning network, and identifying that each Orthogonal Frequency Division Multiplexing (OFDM) symbol belongs to a downlink symbol or an uplink symbol by the second deep learning network; And S6, synthesizing uplink and downlink configuration, namely synthesizing a complete uplink and downlink resource configuration diagram in a signal acquisition time window according to the time slot level identification result in the step S4 and the symbol level identification result in the step S5, and completing synchronization with a base station signal.
2. The method for detecting uplink and downlink signals of a 5G base station based on a deep learning network according to claim 1, wherein the step S2 comprises the following detailed steps: s21, performing sliding correlation on the preprocessed signals by using a main synchronization signal (PSS) sequence to realize frame head coarse synchronization and carrier frequency offset estimation and correction; Step S22, performing sliding correlation near a coarse synchronization point by using an auxiliary synchronization signal (SSS) sequence to realize frame head fine synchronization and position a synchronous broadcast block (SSB); step S23, extracting demodulation reference signals (DMRS) of Physical Broadcast Channels (PBCH) in the synchronous broadcast blocks (SSB), correlating the extracted demodulation reference signals (DMRS) with a local demodulation reference signal (DMRS) sequence set, determining synchronous broadcast block (SSB) indexes, and further calculating to obtain accurate frame head positions by combining protocol mapping relations so as to realize frame head synchronization.
3. The method for detecting uplink and downlink of 5G base station signals based on deep learning network of claim 1, wherein in step S3, after determining the frame header position, the time domain signal of each time slot is divided according to the corresponding time slot length, for each time slot, cyclic Prefix (CP) is removed first, a plurality of Orthogonal Frequency Division Multiplexing (OFDM) symbols are separated, and then Fourier transform (FFT) is performed on each Orthogonal Frequency Division Multiplexing (OFDM) symbol to convert the time domain sampling sequence into frequency domain complex vector.
4. The method for detecting uplink and downlink signals of a 5G base station based on a deep learning network according to claim 1, wherein the step S5 comprises the following detailed steps: Step S51, for the signals identified as flexible time slots, removing the Cyclic Prefix (CP) and then separating out a plurality of Orthogonal Frequency Division Multiplexing (OFDM) symbols contained in the signals; step S52, carrying out Fourier transform (FFT) on the time domain signal of each Orthogonal Frequency Division Multiplexing (OFDM) symbol to generate a frequency domain energy vector of a single symbol and visualizing the frequency domain energy vector into a symbol-level image; Step S53, inputting the symbol-level image to a pre-trained second deep learning network, and outputting the classification result of the image belonging to the downlink symbol or the uplink symbol by the second deep learning network.
5. The method for detecting uplink and downlink of 5G base station signals based on deep learning network of claim 1, wherein the method for constructing the first deep learning network comprises the following steps: s10, data preparation and preprocessing, namely collecting signals with known time slot configuration, manually marking labels of downlink time slots or uplink time slots or flexible time slots for each time slot of the signals, and generating a time-frequency image for each time slot through Fourier transform (FFT) and visualization processing after frame header synchronization and time slot division are completed; step S20, designing and constructing a network architecture, namely selecting a first network architecture, designing an input layer adapting to the image size, designing a feature extraction layer for feature extraction, and outputting the probability of a downlink time slot or an uplink time slot or a flexible time slot; And step S30, performing iterative training and parameter optimization, and storing a model with the best performance to obtain a first deep learning network.
6. The method for detecting uplink and downlink of 5G base station signals based on deep learning network of claim 1, wherein the method for constructing the second deep learning network comprises the following steps: Step S100, data preparation and preprocessing, namely collecting signals of flexible time slots, manually marking labels of downlink symbols or uplink symbols for each time slot of the signals, separating all Orthogonal Frequency Division Multiplexing (OFDM) symbols, and generating single-symbol time-frequency images through Fourier transform (FFT); Step 200, designing and constructing a network architecture, namely selecting a second network architecture, designing a feature extraction layer, and outputting the probability of a downlink symbol or an uplink symbol; And step 300, performing iterative training and parameter optimization, and after completing supervisory training on a symbol-level data set, storing a model with optimal performance to obtain a second deep learning network.
7. The method for detecting uplink and downlink of 5G base station signals based on deep learning network of claim 1, wherein the first deep learning network is a Convolutional Neural Network (CNN) or a visual transformer (ViT), and the second deep learning network is a Convolutional Neural Network (CNN) or a visual transformer (ViT).
8. The 5G base station signal uplink and downlink detection system based on a deep learning network is characterized in that the system is used for executing the 5G base station signal uplink and downlink detection method based on the deep learning network as set forth in any one of claims 1 to 7, and the method comprises the following steps: The signal acquisition module is used for acquiring base station signals; the synchronization module is used for carrying out frame synchronization on the acquired base station signals; An image generation module for generating a time-frequency image of a time slot and a time-frequency image of an Orthogonal Frequency Division Multiplexing (OFDM) symbol according to a frame synchronization result; The time slot identification module is internally provided with a first deep learning network and is used for classifying time-frequency images of time slots and identifying the time-frequency images as downlink time slots or uplink time slots or flexible time slots; The symbol recognition module is internally provided with a second deep learning network and is used for classifying time-frequency images of Orthogonal Frequency Division Multiplexing (OFDM) symbols in flexible time slots and recognizing the time-frequency images as downlink symbols or uplink symbols; And the control output module is used for generating uplink and downlink resource allocation control signals according to the identification result.
9. An electronic device, characterized in that the electronic device comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, and the processor implements the 5G base station signal uplink and downlink detection method based on the deep learning network according to any one of claims 1-7 when executing the computer program.
10. A computer readable storage medium, wherein a computer program is stored on the computer readable storage medium, and when the computer program is executed by a processor, the method for detecting uplink and downlink signals of a 5G base station based on a deep learning network according to any one of claims 1 to 7 is implemented.

Description

5G base station signal uplink and downlink detection method, system, equipment and medium based on deep learning network Technical Field The invention belongs to the technical field of wireless communication, and particularly relates to a 5G base station signal uplink and downlink detection method, a system, equipment and a medium based on a deep learning network. Background In a 5G new air-interface time division duplex (NR TDD) system, in order to support ultra-low latency and flexible service adaptation, a highly dynamic uplink and downlink resource allocation scheme is introduced, and unlike a relatively fixed subframe configuration in 4G Time Division Duplex (TDD), configuration information of 5G is dispersed in a system information block (SIB 1), radio Resource Control (RRC) dedicated signaling, and dynamic scheduling information (such as DCI 2_0) of a Physical Downlink Control Channel (PDCCH), which means that if a conventional repeater or a monitoring device needs to acquire complete uplink and downlink configuration, it must continuously decode and associate multiple types of higher layer signaling, which has extremely high implementation complexity, severe requirements on hardware processing capability, high cost, and cannot guarantee reliability of synchronization when decoding of signaling fails. Therefore, there is an urgent need for a low-cost technical solution that is capable of quickly and accurately identifying uplink and downlink configurations directly from physical layer signal characteristics without relying on complex higher layer protocol parsing. Therefore, it is necessary to provide a method, a system, a device and a medium for detecting uplink and downlink signals of a 5G base station based on a deep learning network to solve the above technical problems. Disclosure of Invention The invention mainly aims to provide a 5G base station signal uplink and downlink detection method based on a deep learning network, which gets rid of dependence on complex high-layer signaling analysis by combining physical layer signal characteristics with the deep learning, reduces equipment cost and complexity, improves reliability and efficiency of synchronization and identification, and can be widely applied to a repeater and network monitoring equipment. The invention realizes the aim through the following technical scheme that the 5G base station signal uplink and downlink detection method based on the deep learning network comprises the following steps: s1, signal acquisition and preprocessing, namely acquiring signals with preset time length on a base station, and preprocessing the acquired signals; step S2, frame head synchronization, namely performing correlation operation on the preprocessed signals to realize frame head synchronization and determining the initial boundary of a wireless frame; S3, generating a time-frequency image, namely dividing the preprocessed signal into a plurality of time slots based on a frame header synchronization result, converting time-domain signals of a plurality of Orthogonal Frequency Division Multiplexing (OFDM) symbols contained in each time slot into frequency-domain signals, and generating a time-frequency level image representing time-frequency energy distribution of the time slot; s4, time slot level uplink and downlink identification, namely inputting time slot level images into a first deep learning network trained in advance, wherein the first deep learning network identifies that each time slot level image belongs to a downlink time slot or an uplink time slot or a flexible time slot; S5, symbol-level uplink and downlink identification, namely, for the flexible time slot identified in the step S4, further dividing the flexible time slot into a plurality of Orthogonal Frequency Division Multiplexing (OFDM) symbols, generating symbol-level images of the plurality of Orthogonal Frequency Division Multiplexing (OFDM) symbols, inputting the symbol-level images into a pre-trained second deep learning network, and identifying that each Orthogonal Frequency Division Multiplexing (OFDM) symbol belongs to a downlink symbol or an uplink symbol by the second deep learning network; And S6, synthesizing uplink and downlink configuration, namely synthesizing a complete uplink and downlink resource configuration diagram in a signal acquisition time window according to the time slot level identification result in the step S4 and the symbol level identification result in the step S5, and completing synchronization with a base station signal. Further, step S2 includes the following detailed steps: s21, performing sliding correlation on the preprocessed signals by using a main synchronization signal (PSS) sequence to realize frame head coarse synchronization and carrier frequency offset estimation and correction; Step S22, performing sliding correlation near a coarse synchronization point by using an auxiliary synchronization signal (SSS) sequence to realize frame head f