Research On Universality Of Modulation Recognition Algorithm Based On Multi-channel

With the rapid entry of wireless 5g technology and artificial intelligence technology into people’s daily life,wireless intelligence is becoming a new demand in all walks of life.Automatic modulation classification has become a key step for wireless intelligent devices to receive mixed signals with multiple modulation modes in the air and select the correct demodulation algorithm.Although a variety of deep learning methods in artificial intelligence have been applied in automatic modulation classification,the difference in algorithm performance makes it difficult for wireless intelligent devices with different needs to choose the appropriate classification algorithm.This paper will focus on how to select a modulation recognition algorithm that meets the general requirements.Firstly,on the basis of summarizing the research of modulation recognition algorithms at home and abroad,this paper summarizes four indexes to evaluate the universality of recognition algorithms: accuracy,robustness,computational efficiency and adaptability;Secondly,the theories of multiple modulation of wireless signals,channel transmission of signals,demodulation of wireless received signals and classification models of various neural networks(CNN,inception,RESNET and LSTM)involved in the modulation recognition algorithm are described in detail;Then,the general research model of modulation recognition algorithm performance is constructed.On the premise of providing the ranking basis for the four index evaluation methods respectively,the evaluation methods for selecting modulation recognition algorithms that meet the general requirements under different needs are obtained;Finally,three kinds of index evaluation and ranking of different algorithms under multichannel conditions are carried out by using simulation methods on four data sets.Combined with two demand scenarios,the modulation recognition algorithms that meet the universality requirements are given respectively,and the adaptability characteristics of the algorithms are simulated and analyzed.The simulation results show that although the four modulation recognition algorithms CNN,inception,RESNET and LSTM have obvious differences in the four indicators under different channels,once the scenario requirements are given,the intelligent receiving equipment can immediately select the algorithm that meets the universality requirements from a variety of modulation recognition algorithms.