Abstract: A new algorithm is proposed f
or classification and demodulation of MFSK signals. The ridge of time-frequency
representation of MFSK signals is analyzed. Modulation classification is realiz
ed by using the unsupervised clustering algorithm to determine the optimal numbe
r of clusters. In order to demodulate MFSK signals, the Haar wavelet transform i
s used to estimate the code-width and the centers of optimal clusters are used
to determine thresholds. Theoretical analysis and practical signal processing re
sults justify the robusticity and the efficiency of the new classification and d
emodulation algorithm.

Key words: signal interception, MFSK signal, modulation classification, fuzzy clustering