Advanced Features Generation Algorithm for MPSK and MQAM Classification in Flat Fading Channel

Abstract

The Automatic Modulation Classification (AMC) performance depends on the selected features. Conventionally, Higher-Order Cumulants (HOCs) are the well-known features due to their discrimination ability under different channel conditions. HOCs have good performance under the Additive white Gaussian noise (AWGN) channel, but their performance degrades under fading channel. This paper proposes an Advanced Features Generation Algorithm (AFGA) that generates mathematical forms of new features based on the maximum discrimination between the digital modulation types to overcome this performance limitation. These features have similar complexity to HOCs but better performance accuracy. The simulation results show that the proposed AFGA improves the performance accuracy up to 4.5% for a Signal-to-noise ratio (SNR) value of 10 dB under fading channel conditions with respect to conventional methods.