Abstract: Modulation recognition under non-cooperative reception conditions generally requires sophisticated preprocessing and has limited classification set. In the present paper, a modulation classification scheme based on cyclic frequency features and support vector machine classifier is proposed to improve the classification performance and expand the recognition set under blind reception conditions. The signal’s cyclic frequency features of the cyclic cumulants are used to discriminate digital modulation signals including FSK, PSK, QAM, OQAM, CPOFDM, ZPOFDM, etc. The proposed method alleviates the preprocessing needs such as estimation of parameters and synchronization. Theoretical derivation is presented and simulations performed, showing
effectiveness of the method.

Key words:  modulation classification, non-cooperative reception, SVM classifier, cyclic frequency feature