Identification of frequency hopping signals based on clustering

In order to identify frequency hopping signals,clustering algorithms and searching parameter’ histogram peak values are studied.The histogram algorithm is firstly modified to achieve insensitivity to the estimating error of frequency hopping signal parameters.Secondly,the K Harmonic-Means(KHM) algorithm is modified to achieve insen-sitivity to the random initialization of clustering centers.And finally,a new method of estimating the exact optimum clustering number is proposed.Combining the improved KHM algorithm with the method of estimating the clustering number,the frequency hopping signals can be identified with easily known signals’ durations,directions and powers.The proposed algorithm is shown good in performance.