Linear and fuzzy piecewise-linear signal processing with an extended DSP architecture

Demonstrates that the architecture of a high-end classical DSP can be extended to efficiently perform piecewise-linear fuzzy inferences. A systematic procedure to optimize the allocation of data through the memory hierarchy is also defined. Thus, both conventional linear processing and nonlinear fuzzy processing can be supported by the same piece of hardware that fits in a standard control or signal processing system. Implementation of advanced nonlinear systems is therefore eased and its cost made comparable with that of more conventional solutions