Research On Bandwidth-limited Information Transmission Method Based On Self-learning

The current wireless mobile communication transmission data has increased exponentially,resulting in the current situation of tight spectrum resources.The traditional wireless communication development is based on the OSI seven-layer model.The physical layer focuses on approaching communication capacity from the perspective of the physical spectrum bandwidth.At the same time,the research on traditional source channel joint coding has been relatively mature,and to a certain extent,the channel capacity has been improved from the perspective of error-free data transmission.However,there are still many problems with traditional source channel joint coding.On the one hand,the traditional source-channel joint coding problem is often decomposed into several steps such as source coding and channel coding through the separation law.The basic theory of the law of separation is to decompose the parent problem of communication into multiple sub-problems that are easy to solve.However,the optimal solution of the sub-problem does not necessarily represent the optimal solution of the global problem,that is,a communication system designed with separate source compression and channel transmission may not obtain the optimal performance of the overall communication system.On the other hand,different data have different requirements for source and channel coding in practical applications.The smallest bit error rate does not mean that the loss of information obtained by the receiving end is the smallest,that is,the same bit error is encoded by the source and reconstructed after channel transmission,which may cause a single pixel to be different or cause the entire picture to be garbled.These problems cannot be solved if channel coding and source coding are only considered by their own optimality.Secondly,the law of source channel separation does not hold in the presence of side information or limited packet length,so it is meaningful to study new information transmission methods under conditions such as limited packet length or limited bandwidth.Therefore,in order to solve the above problems,this paper proposes an intelligent information transmission method based on self-learning and a feature compression algorithm based on weight adjustment,providing a learning paradigm that does not Decompose and select the processing method.Acquire from the learning model to obtain the raw data to the semantic features.Output mapping,transforming error-free transmission of bit data at the physical layer into key features at the information level The effective transmission is to extract the key characteristic parameters of the information source in a self-learning manner from the source point of view.It could achieve channel capacity improvement through the effective design of the information knowledge base at both ends of the transceiver.In order to effectively extract the key characteristics of the real communication between the sender and the receiver under the condition of insufficient prior knowledge,this paper uses the principle of zero-sum game to propose a method of intelligent information transmission and a method of generative adversarial network joint design.It uses an unsupervised method to automatically learn and transmit.The key characteristic parameters are obtained by using the iterative self-learning method to converge to the optimal solution for communication resource scheduling.The background knowledge base is gradually constructed to optimize the scalability of the system,so that the overall bandwidth occupation of the transmission system has a time dimension gain.This paper models the joint processing method based on iterative self-learning.The minimum information loss convergence process is the approximation process of the optimal solution for communication resource scheduling.The evaluation method is more similar to the human visual evaluation standard.Simulation experiments compare different features.The performance of the model under the mode,selection range,and time factors further validates that the intelligent definition mode based on the information loss rate can capture the key feature parameters to the greatest extent,indicating that its bandwidth occupation has a time dimension performance gain;in addition,by comparing the human vision and The data transmission volume is different.Under the condition that the basic visual loss is met,the system can meet the 99% information transmission accuracy requirement with the characteristic data transmission volume of 8% of the original data in the Facades training set.Transmission characteristics and traditional image compression transmission methods can effectively reduce the amount of data transmission and reduce bandwidth consumption.Aiming at the problem that the static information extraction method cannot meet the efficient allocation of resources due to the dynamic changes of communication channel parameters,this paper uses the feature priority adaptive ranking algorithm to sort the extracted main feature parameters that affect the real information transmission of the sending and receiving parties.Adjusted adaptive feature compression algorithm.The algorithm improves the flexibility and adaptability of the system.It can solve the optimal solution of communication resource scheduling in a dynamic channel where the parameters such as signal-to-noise ratio and transmission rate change in real time,thereby converging on the optimal solution Gradually reduce bandwidth usage.This article theoretically analyzes the theoretical basis of the joint loss function based on PSNR and perceptual loss,which can guide the iterative self-learning process to dynamically adjust the feature weights in the main feature parameters during transmission to achieve flexible adjustment of feature weights.Simulation experiments are designed to compare the effects of communication bandwidth environment,dynamic channel conditions,and adaptive strategies on performance.From the perspective of visual and semantic loss,it proves that the adaptive feature compression algorithm based on weight adjustment can meet the needs of dynamic channel feature adjustment,adaptively adjust the high-level image Low-level feature weights balance the relationship between the amount of data transmitted and the quality of the reconstructed image,and reduce the amount of data transmitted while achieving effective transmission of information.