The automatic dependent surveillance - broadcast (ADS-B) systems transmit ADS-B signals by randomly broadcasting in the same frequency band, which will lead to the overlapping of ADS-B signals and threaten aviation safety. At present, when Capon algorithm is used to separate ADS-B signal, the pulse characteristics of ADS-B signals are not taken into account, which will greatly degrade the performance of Capon algorithm. Therefore, this paper designs a signal separation algorithm based on cluster weighted covariance matrix for ADS-B overlapping signals. Firstly, the characteristics of the array response of ADS-B signal are analyzed according to its pulse characteristics. Then, three types of snapshots with only noise, only the first signal and only the second signal are screened

out by K-means clustering method. The covariance matrix of these three types is calculated respectively. Secondly,

the covariance matrix of noise snapshots and the covariance matrix of the signal to be suppressed are selected and

the sum of them is calculated to replace the covariance matrix estimated by all snapshots in the objective function of

Capon algorithm. Finally, combined with the Capon algorithm, overlapping ADS-B signals can be separated. The

results show that this algorithm can significantly improve the performance of ADS-B overlapping signals separation.

array signal processing, ADS-B, pulse characteristics, K-means clustering algorithm, signal separation