Hierarchical hypothesis and feature-based blind modulation classification for linearly modulated signals

Abstract

<div lang="en" class="abs"><p class="abs_header">Abstract</p><p>This paper presents a hierarchical hypothesis test and a feature-based blind modulation classification (BMC) algorithm for linearly modulated signals. The proposed BMC method is based on the combination of elementary cumulant (EC) and cyclic cumulants. The EC is used to decide whether the constellations are from real, circular, or rectangular class, which is referred to as macro classifier. The cyclic cumulant is used to classify modulation within a subclass, which is referred to as micro classifier. For the micro classification, we use positions of nonzero cyclic frequencies (symbol rate frequency or carrier frequency) of the received signals. A hierarchical hypothesis-based theoretical framework has been developed to find the probability of error for the proposed classification. The method works over a flat fading channel without any knowledge of the signal parameters. The proposed method is more robust than the one based on EC and at the same time it requires lower complexity than the maximum likelihood approach. To validate the proposed scheme, measurement is carried out in realistic scenarios. The performance of the new algorithm is compared with the existing methods. In this paper, we have considered a six-class problem including binary phase-shift keying, quadrature phase-shift keying (QPSK), offset-QPSK, π/4-QPSK, minimum shift keying, and 16-quadrature amplitude modulation.</p></div>