AR-model-based data extension to improve the Performance of MUSIC

In this paper, we propose an improved version of the Multiple-Signal-Classification (MUSIC) method, which uses AR model based data extension. MUSIC is excellent as a super resolution DOA estimation method and applied on any array configuration. However, the performance of MUSIC degrades in severe environments. Especially for the case of small number of snapshots, MUSIC often fails in making spectrum peaks that lead to accurate DOA estimation. We employ data extension by using the AR model and try to estimate DOAs by increasing the number of snapshots virtually. Experimental results show that the proposed method provides better performance than the standard MUSIC method.