Forward-backward linear prediction to direction finding of coherent sources using higher-order cyclic statistics

This work proposes a hhigher-order cyclic statistic (HOCS)-based forward-backward linear prediction (LP) direction-finding approach for coherent sources and it is term as HOCS-FBLP. In the approach, the previously proposed HOCS-based LP approach is modified by performing both forward and backward LP with subarrays, and defining both the form of forward and backward fourth-order cyclic cumulants, thus incorporate subarrays with forward-backward LP model. It is proved that the proposed approach can de-correlate the coherency with fewer sensors. The simulations show that the improved approach has the advantages of larger array aperture and higher resolution under small length of data, as well as its powerful capability of interference suppression and signal selectivity duo to the use of HOCS.