Multi Features-based Baseband Modulation Classification using Support Vector Machine

Abstract

This research discusses the implementation of machine learning for modulation classification purpose. In order to proof the concept, 6 types of modulation have been selected, i.e., BPSK, QPSK, 8-PSK, 16-QAM, BFSK, and 8-PAM. Machine learning algorithm that was used in this research is support vector machine (SVM) and implemented using MATLAB’s classification learner. Data sets were generated using an ADALM-PLUTO SDR, and processed at baseband frequency range. Regarding the input predictors to the SVM algorithm, this research proposes multi classification features, such as wavelet transform-based, spectral-based, and higher-order statistical-based features. SVM algorithm obtained a classification-rule model with 91.4% of accuracy without any optimization applied.