Markov Chain-based Frequency Correlation Processing Algorithm for Wideband DOA Estimation
Jun Zhang,Ming Bao,Zhifei Chen,Jing Zhao,Hong Hou,Jianhua Yang
DOI: https://doi.org/10.1016/j.sigpro.2023.108968
IF: 4.729
2023-02-12
Signal Processing
Abstract:For wideband direction-of-arrival (DOA) estimation, a Markov Chain-based frequency correlation processing algorithm is proposed in the sparse Bayesian learning (SBL) framework, called the MC-FC-SBL algorithm. The algorithm adopts a new frequency-domain structural correlation prior model, which can be adaptively changed to accommodate multi-wideband sources scenarios with different frequency characteristics. Specifically, the MC-FC-SBL algorithm separates the amplitudes and supports of the sparse coefficients through the spike-and-slab model, and judges the frequency correlation by the consistency of the supports at adjacent frequency points. The support prior is represented by a Gaussian mixture model, and the switching between the supports at adjacent frequency points is simulated by a Markov chain. The MC-FC-SBL algorithm performs the DOA estimation in the SBL framework to determine the adaptive prior of each coefficient by evaluating the appropriate frequency-correlation structural pattern. In addition, the MC-FC-SBL algorithm is processed in the real-domain, and the real and imaginary parts of complex signal are regarded as multi-snapshot data to implement joint sparse constraints, which can reduce the computational complexity and improve the algorithm performance. Numerical simulations demonstrate that the MC-FC-SBL algorithm is superior to the existing algorithms for wideband DOA estimation, and the results of field experiments show that this algorithm is still effective when the source is weak.
engineering, electrical & electronic