Classification of modulation modes using time-frequency methods

This paper proposes a new technique for feature extraction of modulated signals which is based on a pattern recognition approach. The new algorithm uses the cross Margenau-Hill distribution, autoregressive modeling, and amplitude variations to detect phase shifts, frequency shifts, and amplitude shifts, respectively. Our method is capable of classifying PSK2, PSK4, PSK8, PSK16, FSK2, FSK4, QAM8 and OOK signals. Unlike most of the existing decision-theoretic approaches, no explicit a priori information is required by our algorithm. Consequently, the method is suitable for application in a general noncooperative environment. Furthermore, our approach is computationally inexpensive. Simulation results on both synthetic and "real world" short-wave signals show that our approach is robust against noise up to a signal-to-noise ratio (SNR) of approximately 10 dB. A success rate greater than 94 percent is obtained.