Source number estimation in single channel blind source separation

The paper proposed a new algorithm to estimate source number based on cyclic spectrum and curve clustering. The cyclic amplitude spectrum expressions of common digital communication signals are analyzed. And the vanishing characteristics of the interference in the estimation of cyclic correlation using finite observations are derived. Then the estimation algorithm based on cyclic amplitude spectrum curve clustering is proposed, and its suitable conditions are given. Finally several simulations are carried out, which show the efficiency and feasibility of this algorithm.