Digital Signal Types Identification Using a Hierarchical SVM-Based Classifier and Efficient Features

Automatic digital signal type identification (ADSTI) is an important topic for both military and civilian communication applications. Most of proposed techniques (identifiers) can only recognize a few types of digital signal and usually need high levels of SNR. This paper presents a technique that includes a variety of digital signal types. In this technique a hierarchical support vector machine based structure is proposed for multi-class classification. Combination of higher order moments and higher order cumulants up to eighth are utilized as the effective features. Genetic algorithm is used to parameter selection in order to improve the performance of identifier. Simulation results show that proposed identifier has high performance even at low SNR values