The paper describes a software tool that has been developed forproduction of speech signal processing systems, but may be easilyexpanded to include any 1 dimensional signal processing problem withmultiple inputs and outputs. The tool allows the user to define anarbitrary set of interconnected signal processing functions that areinterlinked by vectors of signal data, this kind of problem is thusexpressed in an inherently MIMD format. This user defined structure isparsed to generate a configuration description that follows acommunication based parallel processing format such as that used inOccam. This configuration may be executed either on a serial machine oron a network of transputers using both task and pipeline parallelism.The current system is capable of building a wide range of structures,for applications such as vocoders, speech enhancement, filterbank speechanalysis and speech recognition. The parallel version runs on a networkof 1825 MHz T800 transputers. Current efficiency levels are verystructure dependent, but for the problems discussed within the paper,efficiencies of around 65% are achieved
展开▼
