Automatic Modulation Classification Based on Cyclostationarity for Software Radios

摘要

Automatic modulation classification (AMC) is an important component for cognitive radio (CR) in order to identify unknown signals and is useful for CRs to fully grasp the condition of the radio environment. In this paper, signal cyclostationarity has been utilized for AMC. The spectral correlation density (SCD) is calculated, and the cyclic frequencies of the spectral peaks are collected as features. Taking hardware implementation into consideration, a decision tree is used as a classifier due to lower complexity. The targeted modulations in this paper are AM, BPSK, QPSK, 8PSK, π/2-DBPSK, π/4-DQPSK, OQPSK, MSK, GMSK, 2FSK and 4FSK signals. Simulation results show reasonable classification performance at SNRs of 6 dB and above.