Abstract:

In order to identify frequency hopping signals, clustering algorithms are mainly studied. The K HarmonicMeans (KHM) algorithm is first modified to achieve insensitivity for random initialization of clustering centers. An existing cluster validity function is used to acquire the range of the optimum clustering number. A new method to estimate the exact optimum clustering number is proposed. Combining the improved algorithm of KHM with the method proposed to estimate the clustering number, frequency hopping signals are identified with easyknown signals' durations, directions and power. The proposed algorithm is shown good in performance. 

Key words: frequency hopping signals, identification, clustering, K Harmonic-Means clustering