Research On Amplitude-Phase Modulation Classification Technology

With the rapid development of modern information interaction technology,it is very important to ensure the integrity and security of the content in the process of information transmission.The types of signal modulation are becoming more and more abundant with the iteration of communication technology,complex modulation types have the advantages of fast transmission efficiency and high spectrum utilization,but at the same time,they improve the difficulty of a series of signal processing related work,including modulation type identification technology.As a key technology in the field of signal processing,modulation type recognition technology determines whether the subsequent information analysis can be carried out smoothly.At present,most of the modulation type recognition algorithms that have been proposed stay at the theoretical research level,and some algorithms have large computational complexity or unstable feature extraction process,which makes them unable to be effectively applied to real scenes.Therefore,how to effectively classify and process modulation type signals in the actual environment is the bottleneck that needs to be broken in the field of modulation type recognition technology at present,and the research on this technology has very important practical significance.Traditional modulation type classification techniques usually make specific judgments on different signal types by extracting multi-dimensional features of signals,comparing the differences of the same feature among different signals.In the satellite communication environment,when the signal encounters strong noise interference,the general feature extraction method will encounter problems such as difficulty in feature extraction.This paper proposes an amplitude and phase modulation signal recognition algorithm based on the combination of multi-level blind digital receivers.Based on the principle of digital receiver,the algorithm extracts stable signal information through the hierarchical cooperation of receivers with different structures,makes full use of the structural characteristics of constellation,counts different types of features,and judges the signal types one by one.The experimental data obtained from the actual collected satellite communication signals show that the recognition algorithm proposed in this paper has excellent performance in practical application scenarios.At present,deep learning has made achievements in various fields of biometrics,and has also made good exploration in the application of communication recognition.Conventional digital modulation signals can be classified by using the constellation features of baseband signals.The commonly used image classification network is mainly convolutional neural network,but convolutional neural networks pay more attention to local features,and do not effectively retain spatial information in the process of feature information extraction.In the low SNR environment,the constellation points are seriously affected by noise,and the problem that convolutional neural network ignores the overall constellation point distribution is more prominent,which affects the classification effect.In this paper,a ViT network with local feature window fusion is proposed.In the process of forward propagation,a local feature window fusion module is added to the Transformer output sequence,so that each local feature window contains information of other local feature windows,strengthens the correlation between local features,and solves the problem that traditional ViT networks ignore local feature correlation.At the same time,the network proposed in this paper has the global receptive field characteristics that CNN network does not have.In different SNR environments,compared with the traditional ViT network and CNN network,it shows that the local feature window fusion ViT model is better for constellation classification.