Modulation Identification Based On Twice Constellation Clustering And Fuzzy Compensation Support Vector Machine

In modern world,people are more and more demanding on the quality of communication and the efficiency of communication.However,the transmission environment of wireless communication is becoming more and more complex.MQAM and MPSK signals can make better use of frequency band and channel resources,so it is concerned by satellite communication and network communication.Therefore,the research on modulation recognition of MQAM and MPSK signals has important practical value.Based on the twice Constellation Clustering and fuzzy support vector machine,we propose a new algorithm to identify the MQAM and MPSK signals.The main research results are as follows:1.Based on the analysis of the defects of the traditional single clustering algorithm and the comparison of the different MPSK and MQAM constellation,we came up with a method for twice constellations clustering,which is combining DENCLUE(density based clustering algorithm)and K-means algorithm and extracting the feature values of the constellation of the signal based on the distance.We can solve the problem that the traditional clustering algorithm is dependent on the initial value,unstable and easy to fall into local extreme points.And finally,the feature extraction module can be constructed.2.Aiming at the deficiency of the twice Constellation Clustering Algorithm,The particle swarm optimization algorithm which has good effect in selecting parameters is introduced to optimize improvements: In order to ensure that there is a corresponding optimal clustering radius in different modulation order M,the module of DENCLUE which gains clustering radius is reconstructed,Therefore,the stability of the algorithm is improved under low SNR.MATLAB simulation verifies the effectiveness of the algorithm.At SNR=2dB,the recognition rate of BPSK,8PSK and 4QAM is up to 99% and the recognition rate of 32 QAM is up to 80%,which is higher than the traditional clustering algorithm and fully embodies the superiority of the algorithm.3.Based on the analysis of the deficiency of the traditional multi-class support vector machine,a support vector machine algorithm based on fuzzy compensation and binary tree multi classification is proposed to build the classification of modulation recognition module: By using fuzziness and fuzzy compensation,the paper presents new constraint condition,reconstructs the Lagrange formula of traditional support vector machine,and gets a excellent discrimination function.Simulation results show: At SNR=0dB,the recognition rate of BPSK,8PSK,4QAM,and 16 QAM has reached 98%,the recognition rate of 32 QAM and 64 QAM which traditional support vector machine is difficult to distinguish has reached 80% recognition rate.It fully reflects the robustness of the algorithm in the low SNR,and has a broad space for development in the future.