Performance evaluation of cumulant feature based automatic modulation classifier on USRP testbec

Abstract

In this paper, a USRP based testbed has been developed for evaluating the performance of cumulant feature based automatic modulation classifier (AMC) in a real radio environment. The proposed testbed consists of conventional radio transmitter with a capability to choose any one of BPSK, QPSK, QAM16 and QAM64 modulation schemes. The receiver extracts appropriate order cumulants from the received signal which are then used as features by support vector machine (SVM) based machine learning classifier. Experimental results demonstrate that the Probability of correct classification (Pec) in varying signal-to-noise ratios (SNR) follow the same increasing pattern as in case of simulation results.