Architectural approach to alternate low-level primitive structures (ALPS) for acoustic signal processing

The input/output relationship describing a given signal-processing application in its entirety can be partitioned into a set of primitive operations which may be represented using signal-processing graph notation (SPGN) as a network of interconnected processes. A set of alternate low-level primitive structures (ALPS) can be defined to implement the operations underpinning each node in this network. This SPGN/ALPS processor network is queue driven and message based, each queue consisting of a message header and an array of data. Current systems developments have demonstrated that a small, finite set of low-level primitives is sufficient for the synthesis of the majority of acoustic-processing problems, and the paper outlines the design approach for some of these. In particular, a prime-radix DFT algorithm, FIR and IIR filter structures using constrained filter coefficients and a general purpose array address mapping element are described. In addition, system architectures based on these SPGN/ALPS networks are developed.