CN-122027088-A - Multi-user robust synchronization method and system

Abstract

The invention discloses a multi-user robust synchronization method and system, and belongs to the technical field of wireless communication. The method comprises the steps of constructing an end-to-end semantic communication model, carrying out semantic coding, quantizing, adding a special synchronization head on a sending end, modulating signals, actively injecting random frame/symbol level and chip level synchronization errors, transmitting the signals through a channel to obtain a receiving signal, carrying out cross-correlation peak detection on the synchronization head on a receiving end to realize multi-user frame level coarse synchronization, extracting effective load, reconstructing data through a semantic decoder, calculating reconstruction loss and back propagation, and jointly optimizing semantic codec parameters to enable the parameters to learn robustness on the synchronization errors. The invention effectively resists the imperfect synchronization problem in actual deployment and improves the transmission reliability and semantic fidelity of multi-user uplink semantic communication under frame-level and chip-level errors through the cooperative mechanism of the synchronous head coarse alignment and the robust model fine correction.

Inventors

• ZHANG ZHI
• ZHANG NA
• WU DI
• MO KAIMING

Assignees

• 北京邮电大学

Dates

Publication Date

20260512

Application Date

20260323

Claims (10)

1. 1. A multi-user robust synchronization method, applied in a system comprising K terminal devices and a base station, comprising the steps of: the training stage, constructing and jointly optimizing an end-to-end semantic communication model, wherein the model at least comprises a semantic encoder, a channel transmission module, a frame synchronization module and a semantic decoder; The training phase specifically comprises the following steps: s101, each terminal device encodes and quantizes source data by using a semantic encoder to generate a data bit sequence; S102, distributing a unique synchronous head sequence for each terminal device, splicing the synchronous head sequence with a corresponding data bit sequence to form a transmission frame, and modulating; s103, in a channel transmission module, actively injecting random frame-level time offset and chip-level phase offset into the modulated signal of each terminal device, and simulating multi-user signal superposition, noise and fading to generate a receiving end signal; S104, in a frame synchronization module, carrying out cross-correlation peak detection by utilizing a synchronization head sequence of each terminal device aiming at a receiving end signal, and estimating the starting position of a data frame of each terminal device; s105, inputting the effective load signal into a semantic decoder after demodulation and dequantization to obtain reconstructed data, calculating loss between the reconstructed data and original data, and updating parameters of the semantic encoder and the semantic decoder through back propagation combination based on the loss; And in the communication stage, the trained semantic encoder, the frame synchronization module and the semantic decoder are used for carrying out actual multi-user semantic data transmission and reconstruction.
2. 2. The multi-user robust synchronization method according to claim 1, wherein the synchronization header sequences of the different terminal devices in step S102 are synchronization header sequences with low cross correlation.
3. 3. The multi-user robust synchronization method according to claim 1, wherein the length of the synchronization header sequence in step S102 is dynamically configured according to the number K of terminal devices.
4. 4. The multi-user robust synchronization method according to claim 1, characterized in that in step S103, the random frame-level time offset is sampled from a discrete uniform distribution for simulating inter-user transmission delay differences.
5. 5. The multi-user robust synchronization method according to claim 1, characterized in that in step S103, the random chip-level phase offset is sampled from a continuous uniform distribution for simulating a phase rotation caused by sampling clock bias.
6. 6. The method of claim 1, wherein the specific process of cross-correlation peak detection in step S104 is to perform cross-correlation calculation on the known synchronization header modulation version of the kth terminal device and the received signal in a sliding window, search an index corresponding to the maximum correlation value, and use the index as the estimated value of the data frame start position of the kth terminal device.
7. 7. The multi-user robust synchronization method of claim 1, wherein the semantic encoder and semantic decoder are deep neural network based models and the loss function is a semantic fidelity based loss function comprising at least one of a mean square error loss, a perceptual loss, or a multi-scale structural similarity loss.
8. 8. A multi-user robust synchronization system, comprising the following modules implementing the method according to any of claims 1 to 7: a plurality of terminal devices, each terminal device comprising at least: The semantic coding and quantizing module is used for converting the source data into a data bit sequence; a frame assembly module for splicing the allocated synchronization header sequence with the data bit sequence; A modulation module; base station apparatus, comprising at least: The frame synchronization module is used for detecting and estimating the starting position of each user data frame from the superposed received signals according to the synchronization head sequence of each terminal device; the payload extraction module is used for intercepting payload signals of all users according to the starting position; The semantic decoding and reconstructing module is used for processing the payload signal and reconstructing data; In the training mode, the system further comprises a channel transmission and error injection module for injecting random synchronization errors into the transmission signal during training.
9. 9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any one of claims 1 to 7 when the program is executed by the processor.
10. 10. A computer readable storage medium, characterized in that the storage medium has stored thereon a computer program which, when executed by a processor, implements the method according to any of claims 1 to 7.

Description

Multi-user robust synchronization method and system Technical Field The present invention relates to the field of wireless communications technologies, and in particular, to a method and a system for multi-user robust synchronization based on mode division multiple access (Model Division Multiple Access, MDMA). Background In conventional multi-user communication systems, system resources typically include time, frequency, space, and power, which should be allocated to different users. Multiple access techniques such as Time Division Multiple Access (TDMA), frequency Division Multiple Access (FDMA), space division multiple access, code Division Multiple Access (CDMA), and non-orthogonal multiple access (NOMA) are widely used. In semantic communications, which is considered to be a new paradigm for next-generation communication systems, high-dimensional features are extracted from signal sources in a model-based artificial intelligence approach, and model information spaces for the signal sources and channel features are constructed. By mining the sharing and personalizing information of semantic information from a high-dimensional semantic space, a new multiple access technology based on semantic domain resources is proposed, called Modular Division Multiple Access (MDMA). MDMA saves more bandwidth resources than conventional multiple access techniques. In MDMA systems, multiple users transmit high-dimensional semantic features in superposition on shared resources, which requires strict alignment among users, even small time offsets can affect user discrimination and semantic decoding. However, most of the existing MDMA schemes are designed under the condition of assuming perfect synchronization, and the problem of system performance degradation under the synchronization error is not fully considered. The problem of imperfect synchronization inevitably exists in actual deployment, and mainly comprises the following steps: (1) Frame level offset, namely, the initial positions of different user data frames are misplaced at a receiving end due to transmission delay caused by different distances between each terminal and a base station, processing delay difference and the like. (2) Chip level offset, which is a minute time deviation caused by hardware defects such as terminal crystal oscillator frequency deviation and sampling clock mismatch, is expressed as a phase rotation of a received symbol. Therefore, it is necessary to design a multi-user robust semantic transmission method under imperfect synchronization to alleviate performance degradation of the MDMA system in the presence of synchronization errors. In conventional bit communication, such synchronization errors are detected and corrected in the physical layer mainly by specific synchronization sequences, digital phase-locked loops and other technologies, and the aim is to recover a bit stream without dislocation. However, in semantic communications, the core value of information is its high-dimensional semantic features. The synchronization error not only can cause bit dislocation, but also can introduce nonlinear distortion into a high-dimensional feature vector generated by a semantic encoder, and the distortion can not be repaired through a traditional physical layer synchronization mechanism, so that a semantic decoder outputs nonsensical or extremely low-fidelity reconstruction information, and the overall performance of a system is seriously deteriorated. The existing semantic communication system based on deep learning mostly trains and evaluates under ideal synchronization assumption, and lacks the robustness consideration of actual synchronization errors. Few researches try to introduce an anti-interference mechanism, but generally fracture the synchronization problem and the semantic transmission problem, and cannot realize end-to-end joint optimization, so that the performance is greatly reduced in an actual imperfect synchronization scene. In summary, in an actual wireless communication system, due to factors such as propagation delay difference, drift of a terminal hardware clock, sampling offset and the like between a terminal device and a base station, frame-level synchronization errors and chip-level synchronization errors inevitably exist between different users. Whereas existing multi-user semantic communication systems typically assume that strict synchronization is achieved between users, this assumption is difficult to meet in practical deployments, resulting in significant degradation of semantic decoding performance. Therefore, a technical solution capable of effectively resisting synchronization errors, especially jointly processing frame-level and chip-level offsets, and guaranteeing reliability of multi-user uplink semantic communication is needed. Disclosure of Invention Aiming at the robustness problem faced by the multi-user uplink semantic communication system under the imperfect synchronization condition, the invention pr