RAP-MUSIC GPU Implementation

RAP-MUSIC Algorithm GPU (CUDA 4.0) implementation is finished.

To realize RAP-MUSIC real-time source localization we reduce in a first step the computational costs by modifying and pre-calculating components of the subspace correlation. In a second step we apply to the accelerated algorithm a modified Powell’s Conjugate Gradient Method, which highly optimizes the search process. Since the subspace correlations of the RAP-MUSIC algorithm are independent, they are predestined to be computed on a many-core processor such as a graphics processing unit (GPU) in parallel. For the parallelization, we utilize the GPU and NVIDIAs Compute Unified Device Architecture (CUDA). The robustness of the localization accuracy is proved by the help of simulations, where we analyze the influence of the SNR and different regularization parameters on the localization accuracy and calculation time.