A practical classification algorithm for M-ary frequency shift keying signals

In this paper, the issue of automatically classifying M-ary frequency shift keying (MFSK) signals contaminated by additive white Gaussian noise (AWGN) is considered, which is blind in nature since the modulation parameters of a received MFSK signal such as frequency deviation that may vary over a large range are unknown in advance. Based on the properties of the spectra of MFSK signals, we present a new classification algorithm which only requires some knowledge that can be obtained through the front-end processing. In simulation, the proposed algorithm can classify 2-, 4-, 8-, 16-, and 32-FSK signals with an overall success rate higher than 99% when the signal-to-noise ratio (SNR) is greater than or equal to 5 dB. To examine its effectiveness with MFSK data received from real-world communications systems, it has been applied to 2-, 4-, and 8-FSK data available in a Web site, resulting in an overall success rate of 89.1%. It is clear that the proposed algorithm has made a step towards practical and blind MFSK signal classification.