| Summary: | The intense requirements for high processing rates of multidimensional Digital Signal Processing systems in practical applications justify the Application Specific Integrated Circuits designs and parallel processing implementations. In this dissertation, we propose novel theories, methodologies and architectures in designing high-performance VLSI implementations for general multidimensional multirate Digital Signal Processing systems by exploiting the parallelism within those applications. To systematically exploit the parallelism within the multidimensional multirate DSP algorithms, we develop novel transformations including (1) nonlinear I/O data space transforms, (2) intercalation transforms, and (3) multidimensional multirate unfolding transforms. These transformations are applied to the algorithms leading to systematic methodologies in high-performance architectural designs. With the novel design methodologies, we develop several architectures with parallel and distributed processing features for implementing multidimensional multirate applications. Experimental results have shown that those architectures are much more efficient in terms of execution time and/or hardware cost compared with existing hardware implementations.
|
|---|
