Maximum likelihood approach to classification of digitally frequency-modulated signals

This paper presents a new method to classify M-ary frequency shift keying (MFSK) modulation using maximum likelihood (ML) criterion. This approach is used to identify the order of modulation for MFSK signals. The system is then analyzed theoretically and ML decision rule is obtained. Discrimination power of the proposed classifier is verified through simulations for automatic modulation recognition of MFSK (2, 4, and 8) signals in Gaussian channel. Sequential detection is also applied and analyzed. It is shown that the system complexity decreases in comparison with fixed sample size (FSS) method when sequential detection method is applied.