Modulation Identification For OFDM Based On Support Vector Machine

The rapid development of digital communications results in the coexistence of many different communication systems, and the access technologies and the modulation methods of the communication systems are also different , which leads to a lot of disturbance for the communication among different communication systems. With the increasement of the communication both in kinds and in service, Orthogonal Frequency Division Multiplexing(OFDM), for its advantages such as in high frequency-band utilization and in strong anti-interferfernce, is widely used in communication domain. So the study on identification for OFDM is an urgent problem to be solved.This paper discusses the modulation recognition for OFDM signals based on Support Vector Machine in lower Signal Noise Ratio (SNR) and multipath circumstance. As the studies on signal recognition seldom take OFDM signals and the multipath circumstance into account, and the traditional recognition method either have the disadvantages in the case of limited train samples or have the low recognition rate in lower SNR, a new method of modulation identification for OFDM signals in multipath circumstance based on support vector machine (SVM) is presented. To an SVM classifier, how to select the feature parameters to construct a feature vector makes a great impact on its performance. In this paper, the cumulants and the wavelets used as classification feature vectors are investigated.In this paper, the Support Vector Machine (SVM) is firstly used to identify communication signals in multipath circumstance, and the 3th_order cumulants and its combination with wavelets used as classification feature vectors are also firstly used on SVM. Moreover, we propose a new type of SVM for signals recognition, and implements it by Tabu Search algorithm. Simulation results indicate that the proposed feature vectors have good performances in lower SNR and multipath circumstance, and the new type SVM classifier's ability proves to be good.