Research On Modulation Recognition Algorithm Under Dynamic Uncertain Environment

With the continuous development of communication technology,dynamic uncertain characteristics of communication channel and the environment have become increasingly prominent and bring difficulties to the validity and reliability of the transmission of information.The modulation recognition technology,which is a key technology for reliable communication,are widely used in communications,spectrum management and military reconnaissance and other fields.The topic of this article is from the National Natural Science Foundation of China(project number:61471061).The problems what research mainly solves are modulation recognition algorithm under dynamic uncertain environment,especially time-varying fading channel.The main research work includes two parts.The first part is without consideration of dynamic characteristics of time-varying fading channel and based on the characteristics of a received signal to finish modulation recognition.In this part,two modulation recognition algorithms were studied.The first algorithm is first proposed and based on deep learning(DL),which abandons the traditional complex parameter or feature extraction processes and uses signal characteristics of time-domain magnitude and spectrum to make recognition by DL.Theory analysis and simulation show that the algorithm has high recognition accuracy.The second one is based on constellation clustering analysis,which has low computational complexity of the algorithm and a good performance at high SNR region.The second part takes full account of dynamic time-varying characteristics and is based on the likelihood function.In this one,an adaptive modulation recognition algorithm is presented.A new sequential Bayesian inference algorithm,which is based on dynamic state space model newly proposed and take full use of the prior transfer characteristics of time-varying channels,is designed and makes adaptive modulation recognition effect by joint estimation of channel status and modulation.Depending on the mapping rules,it is divided into two cases.One is that the channel gain and modulation are single-to-single(S2S)mapping.Compared with the average likelihood ratio test(ALRT)and the pilot assisted estimation scheme,feasibility of new algorithm is verified by simulation.The other is many-to-single(M2S)mapping.Compared to a solution of channel gain priority and a solution of priority modulation,simulation result shows the advantages and disadvantages of different options.Structure of the paper describes as follows:Chapter 1 is an introduction,which introduces the research background of modulation recognition technology,current research situation and content of the article.In the chapter 2,the two static modulation recognition algorithms are studied.Firstly,we present a recognition algorithm based on deep learning,which has a simple feature extraction and then use deep learning to make recognition.Simulation analysis proves the recognition accuracy is higher than the Back-Propagation(BP)neural network.Secondly,we study an algorithm based on constellation clustering analysis,which uses subtractive clustering.It is easy to understand and not so strong in resisting noise by simulation.Chapter 3 studies the traditional dynamic modulation recognition algorithm,including system modeling and algorithm flow of the ALRT algorithm and the pilot assisted estimation scheme.Chapter 4 presents the adaptive modulation recognition algorithm based on Bayesian sequential inference and expounds system model and sequential inference processes.Simulation result concludes that,in S2S case,neither under Rice or Rayleigh channel,recognition accuracy and computational complexity of the new algorithm is superior to ALRT and a little weaker than the pilot assisted estimation scheme.Chapter 5 studies the algorithm flow of two different solutions of new algorithm in M2S case.Simulation analysis found that,as the ratio of signal to noise ratio increase,scheme II which takes a full use of the channel information,makes faster growth of performance than scheme I which marginalizes channel gain.Chapter 6 makes the summaries and outlook of this paper.