Parallel implementation considerations for a class of signal processing algorithms

Abstract

The Burg signal-processing algorithm, which is widely used in such areas as geophysical data analysis, speech processing, and spectral analysis, is investigated in order to determine its suitability for implementation on parallel-processing architectures. The algorithm is implemented on three different machines, representing a variety of parallel architectures: Denelcor HEP, the Intel iPSC/2 hypercube, and the NASA/Goodyear massively parallel processor (MPP). The Burg algorithm is introduced, followed by a description of the parallel implementation of the algorithm, with special emphasis on the MPP, and a comparison of the performances of the three machines. It is found that the algorithm is especially well suited to the massive parallelism of the MPP. This is somewhat surprising since the MPP, with its mesh architecture, was designed (in part) to process two-dimensional images, while the Burg filter is a one-dimensional signal-processing algorithm. >