Abstract:

 A novel method using decision-making theory is proposed to identify digital modulation signals. The method is based on the cyclic spectrum of received signals and the statistical parameters. Using the frequency domain smoothing periodogram to estimate the cyclic spectrum, the characteristic spectral density of digital modulation
signals is analyzed on the cyclic frequency axis. Statistical parameters of digital modulation signals are then
extracted. Finally, using the decision-making theory, we present a flow chart and a determination threshold for
signal recognition by computer simulation. At 5 dB SNR, identification probability of 90% for 2ASK, 4ASK, 2FSK,
4FSK, 4PSK and 8PSK can be achieved. The results show effectiveness of the method even when signals are buried in strong noise.

Key words: modulation identification, cyclic spectral density, statistical parameter