DOA estimation for quasi-stationary signals without knowing the number of sources

In many practical scenarios such as speech and audio processing applications, quasi-stationary signals are frequently encountered, and in this paper, we present a novel method for solving the problem of direction of arrival (DOA) estimation of quasi-stationary signals without considering the number of sources. Based on Khatri-Rao product, if the number of sensors is N, through virtual expansion, we can identify up to 2N-2 sources DOA. According to the fact that the eigenvalue belonging to noise subspace is far less than the eigenvalue belonging to signal subspace, we can modify traditional Capon algorithm and improve its resolution. Finally, exploiting modified Capon algorithm, we realize DOA estimation of quasi-stationary signals even with no consideration the number of sources. Simulation results demonstrate the effectiveness and performance of the proposed method. © 2015 by Binary Information Press.