Wireless Signal Classification Based on High-Order Cumulants and Machine Learning

摘要

Wireless Signal Classification (WSC) is mainly based on Automatic Modulation Classification (AMC) and an intermediate process between signal detection and demodulation, which can identify the modulation from the received signal. The higher-order cumulants (HOCs) are derived from the received signals and input as features to the selected classifier. HOCs have good anti-noise performance, which can classify the modulation methods commonly used in current wireless communication systems. In this paper, the classification of {OFDM, BPSK, QPSK, GFSK, 16QAM, 64QAM} is realized by MATLAB programming based on the characteristic of HOCs. A new feature parameter is proposed according to 2th order cumulant and 6th order cumulant. The simulation results show that under the SNR of 3dB, the DT classifier's classification accuracy is over 99.4%. Those 6 kinds of signals are the main modulation signals used in wireless communication systems. Therefore, this paper has engineering application value in some degree.