합성곱 신경망 기반 위상 오프셋에 강인한 변조 분류

Abstract

Automatic modulation classification (AMC) is one of the important technologies in non-cooperative contexts such as cognitive radio. In this paper, we propose an AMC robust to phase offset by using convolution neural network. For AMC, we adopt polar coordinate images to which two-dimensional discrete Fourier transform is applied. We analyze the classification performance of the proposed method and show that the proposed AMC is not affected by phase offset.