Large Array DOA Estimation Method Based on VMD-NGOELM

In addressing the severe interference caused by noise issues in complex environments and the high-dimensional covariance matrix problem arising from large arrays during target azimuth estimation, this research introduces an innovative approach for Direction of Arrival (DOA) estimation in large-scale arrays, which integrates Variational Mode Decomposition with an optimized Extreme Learning Machine framework enhanced by the Northern Goshawk algorithm (VMD-NGOELM). This method leverages VMD to decompose the local features of each array signal, employing regularization and constraints to reduce the noise components in the signal, thereby mitigating the impact of noise information on covariance matrix construction. Subsequently, by harnessing the bio-inspired nature and local search capabilities of the NGO method, the algorithm of Extreme Learning Machine (ELM) undergoes a comprehensive optimization process in this study, enhancing the DOA estimation performance of the ELM model. The rapid training of ELM facilitates the swift acquisition of output weights between the covariance matrix and DOA, thereby accelerating the DOA estimation process for large arrays. Extensive numerical experiments reveal that the proposed methodology maintains superior accuracy in direction of arrival estimation, particularly in challenging low SNR environments.