A Parallel Processing System for Broadband Robust Capon Beamformer
Robust capon beamformers has many advantages than conventional beamformers at the price of more computation complexity, which pose great challenges in embedded systems, especially when real-time must be satisfied. This paper demonstrates a multi-processor hybrid system which tries to meet these requirements. Fixed-point DSPs and FPGA are incorporated together to enhance the computation ability of the system. High-speed serial interfaces like SRIO and USB are employed to enhance the data transporting rate and as a result the throughput of the system. Hardware and software characteristics are discussed in details to explain how to get efficient speedup of the algorithm. By decomposing the algorithm into steps, the computation load in each step is analyzed and then the inherent parallelism of the algorithm is exploited and made use of. Other problems when implementing algorithms with fixed-point DSP, such as determining the scaling of each fixed-point number , computing nonlinear function value, and computing the eigen-decomposition of Hermitian matrix are also discussed in the paper. Algorithm prototypes which can help to identify the scaling and simulate the fixed point performance are developed individually with MATLAB and C as golden references.At the end, experiment results show the great ability of the system when real-time must be satisfied.
Ping Feng Mo-jiang Chen Xuan Li Xiao-chuan Ma
Institute of Acoustics,Chinese Academy of Sciences,Beijing 100190,P.R.China
国际会议
The 10th Western Pacific Acoustics Conference(第十届西太平洋声学会议WESPAC X)
北京
英文
1-7
2009-09-21（万方平台首次上网日期，不代表论文的发表时间）