Architecture models for maximum likelihood direction of arrival estimation
Aprameya Satish, R.L. Kashyap · 2003
A quantitative analysis of the computational requirements of two direction-of-arrival (DOA) estimation methods, maximum likelihood (ML(RLK)) and expectation-maximization maximum likelihood, for wideband source signals is performed. Essentially, the number of arithmetic operations (complex additions and multiplications) which can be regarded as a performance measure for each of the angle estimation methods is determined analytically. Based on these observations dedicated coprocessor architecture (RISC/CISC) models are proposed for each of the methods. The idea of efficiently matching a coprocessor computer organization with a parallel algorithm, i.e., tuning the architecture to be software oriented, is utilized for the design of the underlying models. It is seen that the ML(RLK) method can also be parallelized with relative ease for high performance.>