DOA Estimation of Nonstationary and Close-Spaced Sources Based on Unified Parameterized Model

Time–frequency (TF) analysis can improve the performance of direction-of-arrival (DOA) estimation by only utilizing the valid TF points in the TF plane. However, when instantaneous frequencies (IFs) interact with each other in a low signal-to-noise ratio (SNR), the DOA estimation performance will degrade because of the difficulty in extracting valid TF points or IF ridges. We propose a novel method that adopts the unified general parameterized TF transform (GPTFT) with multiple sensors considering the sparseness of the signal in the angle domain to deal with severe situations in this article. First, a unified parameterized TF representation is proposed to enhance the ability to extract the IFs. Based on the sparseness of the signal in the angle domain, an evaluation of IF candidates is employed to select the most reliable IF candidate in GPTFT, which makes it more robust and accurate. With a specific IF, intrinsic chirp component decomposition (ICCD) in multiple channels is employed to derive the single-source signal and DOA can be estimated at last. The proposed method can solve the situation where IFs are complex in the TF plane and sources are close-spaced. Compared with the short-time Fourier transform and GPTFT, the energy ratio of the proposed method in the real IF band is increased by 12.17% and 6.79%, respectively. The proposed method exhibits a higher accuracy in DOA estimation in comparison with conventional methods when dealing with close-spaced and nonstationary sources. Simulations in different configurations of parameters are provided to verify the effectiveness of the proposed algorithm.