Modulation recognition of communication signals based on high order cumulants and support vector machine

Abstract Using high order cumulants as signal eigenvalues, signals of different modulation formats can be recognized with support vector machine (SVM). Since high order cumulants of different modulation signals are different from each other, fourth and sixth orders of cumulants were chosen and only four eigenvalues were constructed in the recognition in this paper, thus signals of different modulation format and signals of different modulation order with the same format were recognized in low signal-to-noise ratio (SNR). It was shown that the key of recognition of modulation format is how to choose proper high order cumulants as eigenvalues. So that acceptable accuracy of recognition could be obtained in low SNR. During the recognition process, one-to-one classification of LibSVM was used to simplify the usual binary separation process. An efficiency modulation recognition method was put forward with high recognition rate and less complexity.