Automatic Modulation Classification in Cognitive Radio Using Multiple Antennas and Maximum-Likelihood Techniques

Abstract

The automatic modulation classification (AMC) is linked to the accurate identification of a received signal modulation. The AMC represents an important part of cognitive radio (CR) systems recently envisioned to be an appropriate platform to adjust against changing work conditions. The two main distinguished streams of AMC are either by using the likelihood based (LB) statistical tests or featured based (FB) recognitions. The LB is viewed to be optimum providing that all statistical signal descriptions are available to a receiver, while the FB usually viewed as suboptimal. In some practical situations, especially when the AMC process is carried out blindly, a signal enhancement is viewed necessary to boost the detection accuracy. The multiple-input multiple-output (MIMO) antennas configuration is widely accepted as key enhancer to signal and system performance. This paper is intended to explore opportunities of the AMC detection accuracy improvements using MIMO and diversity combining settings. The probability of error detection is reformulated using diagonalized MIMO channel through singular value decomposition (SVD) realization. Simulation results show that classification performance is improved by adopting multiple antennas with appropriate signal combining configuration.