Parallel computation of SVD for high resolution DOA estimation

In the study of array signal processing, the extensively researched MUSIC (Multiple Signal Characterization) algorithm can be realized based on SVD (Singular Value Decomposition) and many advantages are expected to be achieved. In the application of SVD, how to compute SVD on a parallel array becomes a key issue. In this paper, we discussed several SVD algorithms and their implementation on array processors. Detailed analyses demonstrate some modifications of Hestenes SVD method and Franklin SVD algorithm can make them more suitable for array processing since the array architecture may be substantially simplified.