PROJECT TITLE :

An Optimized GPU Implementation of the MVDR Beamformer for Active Sonar Imaging

ABSTRACT:

The minimum variance distortionless response (MVDR) beamformer has recently been proposed as an engaging alternative to traditional beamformers in active sonar imaging. Unfortunately, it is terribly computationally complex because a spatial covariance matrix should be estimated and inverted for every image pixel. This might discourage its unnecessary use in sonar systems which are continuously being pushed to ever higher imaging ranges and resolutions. In this study, we show that for active sonar systems up to 32 channels, the computation time will be significantly reduced by performing arithmetic optimizations, and by implementing the MVDR beamformer on a graphics processing unit (GPU). We purpose out necessary hardware limitations for these devices, and assess the design in terms of how efficiently it is ready to use the GPU's resources. On a quad-core Intel Xeon system with a high-end Nvidia GPU, our GPU implementation renders additional than a million pixels per second (1 MP/s). Compared to our initial central processing unit (CPU) implementation, the optimizations described herein led to a speedup of additional than two orders of magnitude, or an expected 5 to 10 times improvement had the CPU received similar optimization effort. This throughput permits real-time processing of sonar information, and makes the MVDR a viable different to conventional ways in practical systems.