Support vector machine based classifier for digital modulations in presence of HF noise

Support vector machines (SVMs) deal with challenging classification problems. One such a challenge is to classify the modulation type of a signal transmitted over High frequency (HF) band. The noise distribution in this band has time varying nature. SVM can mitigate these variations with correct choice of features and kernel functions. This paper presents a feature-based classification method utilizing SVM for the classification of 10 types of modulations in the presence of Gaussian as well as non-Gaussian noise disturbances. The proposed method is able to classify type and order of modulation at relatively low signal-to-noise ratios (SNRs) for both simulated as well as actual data.