Optimized Real-Time MUSIC Algorithm With CPU-GPU Architecture

Direction-of-arrival (DOA) estimation algorithm for uniform planar arrays has been applied in many fields. The multiple signal classification (MUSIC) obvious advantage high-resolution source scenarios. However, the MUSIC high computational costs, therefore it is hard to be used real-time scenes. Many studies are dedicated accelerating by parallel hardware, especially Graphics Processing Units (GPU). based on Central Unit (CPU) -GPU architecture acceleration rarely investigated previous literatures, and how well Algorithm with CPU-GPU could perform remains unknown. In this paper, we present evaluate a model of search architecture. proposed model, steering vector each candidate incident corresponding value 2D spatial pseudo-spectrum (SPS) function sequentially calculated single core GPU, subsequent calculation elevation or azimuth batches. Furthermore, order improve peak speed, propose new Coarse Fine Traversal (CFT) via CPU GPU acceleration. Across strategy comparison, utilizing processing, 150-160x performance gain achieved compared using only. Besides, resolution also analyzed.