Passive localization and classification of mixed near-field and far-field sources based on high-order differencing algorithm

Abstract

Abstract A novel method called high-order differencing algorithm (HODA) is presented for localization of mixed sources. Five special cumulant matrices are constructed. The first one only contains the angle information. By modifying its rank and using an ESPRIT-like approach, the initial DOA set (IDOAS) is formed. The four others, pairwise, contain common far-field information. By two differencing operations, the far-field information is eliminated, and the difference cumulant matrices (DCMs) are obtained. After the rank modification is executed, the DCMs are reconstructed. By applying an ESPRIT-like approach to them, electrical angles are extracted. The extracted data is compared with IDOAS to obtain valid information of near-field sources (NFSs). A mechanism called kurtosis testing algorithm (KTA) is presented for identifying far-field sources (FFSs). KTA is able to identify even those FFSs that are located at the same angle with NFSs. To control the error of statistical differencing, an appropriate number of snapshots is considered. Analyses show that HODA prevents aperture loss; it does not require pairing, knowing the number of NFSs or FFSs, and heavy searches. The results confirm its good performance in terms of classification, the correct estimation of sources with different fields and the same DOAs, estimation accuracy and computational complexity.