Automatic Digital Modulation Recognition Using Feature Subset Selection

Modulation type is one of the most important characteristics used in signal identification and monitoring. Modulation recognition systems should correctly classify the incoming signal's modulation scheme in the presence of noise. Automatic digital modulation recognition (ADMR) can be used for both military applications and civilian applications. Some examples are surveillance, electronic warfare, threat assessment, signal confirmation, interference identification and spectrum management.In this paper, a new automatic digital modulation recognition method using ECOC-SVMs and GA is introduced. A new feature set combined statistical and spectral feature subset is used for modulation classification to make the SVMs classifier more robust to Gaussian noise. Moreover, GA is used to perform feature subset selection to reduce the input dimension and increase the performance of the ECOC-SVMs classifier. Compared to the conventional ANN method and the decision theoretic algorithm, the proposed method can recognize more digital modulation types. Furthermore, significant improvements can be seen particularly at a low SNR.