Using combination of support vector machines for automatic analog modulation recognition

Abstract

In this paper, the usage of an automatic decision system based on the combination of support vector machines (SVM), which recognizes the some analog modulation types, is introduced. Here, automatic analog modulation recognition performance of this system is compared with a multi-layer perceptrons (MLP) classifier. The discrete wavelet transform (DWT) and wavelet entropy (WE) methods are used for the extraction effective features from analog modulation signals in the feature extraction stages of both these SVM and MLP systems. In this study, some experiments were performed for finding the optimal C (cost) and σ (sigma) kernel parameters of SVM. The analog modulated signals used in this study are amplitude modulation (AM), double side band (DSB), upper single band (USB), lower single band (LSB), frequency modulation (FM), and phase modulation (PM). The performance of both SVM and MLP classifiers is evaluated by using total 3240 analog modulated signals. These test results show the effectiveness of the system proposed in this paper. The rate of correct classification is about 96.419753% for the sample analog modulated signals.